TSX:PAAS Pan American Silver Corp.

Lifecycle and sector. Mature cyclical producer in Materials, scored on the mining metric profile: AISC margin, reserve life, balance sheet, ROIC and capital allocation. Traditional growth metrics are not the lens — the 38.4% year-on-year revenue increase is a metal-price effect, not a volume one, and I have not scored it as growth.

Competitive Moat Scorecard

Moat score: 45/100. Below the prior run's implied level, and derived from the competitive read rather than assumed. A commodity producer with no pricing power and a middle-of-the-pack cost position does not have a moat; it has assets.

Net effect on the moat: Cost Advantage trimmed to 58 on the gold-cost gap and the streaming-model comparison; Intangibles held at 45 because the permit base is as much a risk as a barrier. Switching Costs and Network Effects are not applicable and stay neutral at 50. This read propagates to the §11 bear case (a gold-cost trigger) and to the §12 thesis-invalidation rule.

ROIC & Capital Allocation

Quality: 72/100, confidence 70%. The balance sheet, the AISC margin and the capital-return framework are all strong. What holds it back is a nine-year producing reserve base, a gold segment that missed its own quarterly guidance and is now guided to the low end on volume and the high end on cost, and a moat that on honest scoring is 45.

1 — The discount rate

Growth terms are 0.0 by construction. A mining NAV discounts a finite mine plan at flat nominal cash flows over the reserve life; there is no near-term or terminal growth term. The two-stage growth formula belongs to an earnings-multiple instantiation, which L570 does not give a miner. Recorded as 0.0 rather than omitted, so a later run can see the basis instead of inferring it.

2 — The reserve base, mine by mine

Read individually from Pan American's own mine pages rather than taken from the 452.3Moz headline. That is what revealed the two facts this section turns on: 264.5Moz of the headline sits at Escobal, which has produced nothing since 2017, and 40.1Moz of it is by-product silver inside the gold segment, where its revenue is already credited against the gold cost.

All figures effective 30 June 2025 and read from the individual mine pages, with two stated exceptions: Juanicipio's ~6.05Moz of attributable output is derived (see below), and Dolores publishes no reserve figure because mining from known reserves concluded in Q3 2024. The rows reconcile to the headline: the silver reserves above sum to 452.4Moz against the 452.3Moz reported — 0.02% apart, so nothing is missing. The gold rows sum to 6,061koz against the 6,339koz reported; the 278koz difference is the Dolores residual-leach stockpile, which carries no reserve statement and is not credited anywhere in this NAV.

Confirmed absences, checked rather than assumed: La Colorada Skarn reports no reserves (309Moz indicated, PEA-stage since December 2023) and Navidad reports no reserves (632.4Moz M&I, blocked by Chubut Law 5001). Neither is inside the 452.3Moz, so neither is double-counted.

3 — The mine plan, and the by-product correction

4 — The life

Silver: 10.0 years. The four consolidated silver mines hold 147.8Moz of proven and probable reserves against 2025 output of 14.752Moz. Juanicipio's reserves are absent from the statement, which predates the 4 September 2025 MAG Silver close, so its ~6.05Moz of attributable output is assigned the same 10.0-year life — an explicit assumption, and a conservative one for the newest mine in the portfolio, commissioned in 2023. It is replaced by disclosure at the next reserve statement, expected September 2026.

Gold: 9.0 years — 6.3Moz against 700koz of guidance. All gold reserves sit at producing mines; there is no gold equivalent of Escobal.

Blended: 10.019 × 0.36727 + 9.000 × 0.63273 = 9.374 years, weighted by modelled margin contribution. Annuity factor at 9.13% = 6.1242. (The life is carried at three decimals deliberately: at a rounded 9.37 the annuity factor is 6.1224 and the NAV C$34.74, so publishing the rounded life beside an unrounded factor would look like an arithmetic slip. The ratio is 1.894 either way.)

Does blending the two streams into one annuity flatter the answer? It is a fair question, because the annuity factor is concave in the life, so a single blended annuity can be worth more than the sum of its parts. Here it is not material: discounting each stream at its own life, with the corporate charge allocated pro rata by margin, gives US$8,592.3m against the US$8,598.5m the blended annuity produces. The blend is 0.07% generous. A full mine-by-mine sum-of-parts on the silver segment, using each mine's own disclosed life and AISC, comes out higher than the blended treatment, not lower, because the blended US$17.00 guidance cost is dominated by the high-cost long-life mines while the negative-cost Juanicipio carries a longer life than Cerro Moro or San Vicente. The aggregate treatment is not the generous one.

5 — The NAV

On the net cash figure, precisely. The release states “Total Debt of $841 million” and “Cash and cash equivalents and short-term investments of $1.8 billion” — it does not print a net-cash line. The US$981m used here is therefore derived, and it implies unrounded cash and short-term investments of US$1,822m. If the true figure is US$1,800m the net cash is US$959m, the NAV falls by C$0.07 to C$34.68 and the ratio rises from 1.894 to 1.898. Immaterial to the band, but it is a derived number and is labelled as one rather than attributed to the company.

Actual C$65.82 ÷ warranted C$34.75 = 1.894×. Warranted multiple 1.00×, warranted ratio 1.894, band EXPENSIVE.

6 — The guardrail, adjudicated by hand

The Materials guardrail line is 1.5× P/NAV (or 8× EV/EBITDA). The deterministic linter's guardrail arm keys on an earnings- or book-value basis string and cannot read a P/NAV, so it is adjudicated here explicitly. P/NAV 1.894× against the 1.5× line — BREACHED. On the parenthetical expression, matched-currency EV/EBITDA is 7.75× (EV US$18,613m, from a US$19,594m market cap — 414.6m shares at the stamped C$65.82 close, converted at 1.39271, not the C$27.688bn struck on the live C$66.78 quote quoted in §15 — less US$981m net cash, over US$2,402m of TTM EBITDA) against the 8× line — not breached; the provider's unmatched 8.14× uses a different enterprise value and is not the figure used here. The primary multiple governs, so the guardrail arm fires and Gate 3 is independently triggered on top of the ≥1.40× band arm.

7 — Relative cross-checks (they order within the band; they cannot lift it out)

On that PEG, honestly: 0.62 is the provider's own forwardPriceToEarningsGrowthRatioTTM, not a figure this report computed. It implies about 15.6% forward growth against the 9.65× forward multiple, whereas the forward-versus-trailing spread (9.65× against 14.39×) implies roughly 49%. The two are struck on different bases and this run has no sourced growth rate to arbitrate between them — get_analyst_estimates failed, and a mining NAV carries no growth term at all (g_near = g_term = 0.0). Neither number is load-bearing: this lens is 10% of a block that cannot lift the band, and no growth rate enters the anchor. It is recorded rather than quietly dropped.

The Analyst Consensus lens is 15% in total, split 10% price-target signal and 5% grades. Weighted, the cross-checks come out at 37.2, and the weights sum to 100. The band caps the score below 40 in any case, so Valuation = 36 — near the top of the Expensive band, which is where the strong secondary and tertiary lenses put it, not at the bottom.

Analyst price targets — and why they are discounted

The Yahoo mean of US$65.50 implies +38.6% on the US$47.26 traded price (C$91.22 against C$65.82). I have used it rather than FMP's US$73 because FMP's coverage summary shows zero targets issued in the last month and one in the last quarter — at US$53. Both panels therefore predate the 12 August double miss and the 9.8% session that followed it. No analyst has marked to the Q2 result yet. That is why this sub-factor scores 60 rather than the 85–100 the raw upside would earn, and it is a live risk to the consensus leg of this pillar rather than support for it.

Grades: 13 buy, 10 hold, 2 sell — 52% bullish, a Buy consensus with real hesitation. No upgrade or downgrade in the last 30 days; the most recent action was B of A maintaining Buy on 9 July 2026, and before that TD Cowen upgrading Hold → Buy on 12 May 2026. All maintains, all pre-miss.

FMP ratings cross-reference: overall A− (4/5). ROE 5/5, ROA 5/5, DCF 4/5, debt/equity 3/5 — and P/E 2/5, P/B 2/5. An independent model reaching the same split verdict this section does: a high-quality business at a full price.

Valuation: 36/100, confidence 60%. The confidence haircut is honest — the mine life is an author choice the framework does not bound, the reserve statement is thirteen months old and excludes Juanicipio, and the analyst panel is stale.

The tape in one line: above a falling 50-day average, 7.2% below the 200-day, and lagging its own sector badly over three months. Neutral, not weak — and not improving.

Sentiment layer — the analyst actions in full

Nothing in the last 30 days, and nothing at all since the Q2 print. Sentiment scores 42 on news tone rather than on grade actions, because there have been no grade actions to score.

Catalyst inventory

Timing: 52/100, confidence 58%. Confidence is cut 10 points for a high-impact macro release inside 7 days on a High-macro-sensitivity name, and a further 5 because the relative-strength comparators are read off the US listing to avoid an FX artefact.

Upcoming events (next 30 days)

Recent surprises (last 7 days)

One event dominates. Materials is a High-macro-sensitivity sector and the FOMC minutes land on Wednesday 19 August, two trading days from this report, which is inside the three-trading-day window in which SKILL §8 fires a WAIT-FOR-EVENT override on the short horizon “regardless of composite score”. That is applied literally here: it sets signal_short = WAIT and it sets the next update to 20 August. This report takes the scheduling benefit of that clause, so it cannot decline the label the same clause names.

Why the minutes matter for this name specifically: the macro report of record carries a live September-hike tail at roughly 44% market-implied odds with three hawkish dissents in July, and names these minutes as its trigger. Pan American's two drivers are gold and silver, both of which are bid on fiscal debasement rather than on rate cuts — but a hawkish surprise lifts real rates and hits both regardless. With silver already −8.07% over eight weeks, that is not a risk to take blind two days before the release.

The multi-timeframe table is read off the US listing (NYSE:PAAS, in US$) because the intraday feed does not cover the Toronto line; the levels convert at C$1.39271. The Canadian listing's own daily indicators, computed separately on raw unadjusted closes, are the ones used everywhere else in this report.

The Toronto picture: close C$65.82; 50-DMA C$65.02 and falling (it was C$66.30 on 31 July); 200-DMA C$70.92 and also falling. RSI(14) 49.98 — dead neutral. MACD +1.358 against a signal of +0.733, so the histogram is still positive at +0.624 but has more than halved from +1.569 on 12 August. ATR(14) C$3.03.

The pattern that matters. This is the mirror image of the textbook “pullback in a higher-timeframe uptrend”. The monthly is still up, but the weekly has rolled into a downtrend and sits below its own 20- and 50-week averages, while the hourly is bouncing. Higher timeframe bearish plus lower timeframe rally is the setup that fades, not the one that runs. It is also exactly what the metals themselves are doing — above a falling 50-day, below the 200-day — which is not a coincidence: the equity is a geared version of the metal.

The 13 August session is the structural fact. A 9.8% single-day fall on 2.41× average volume immediately after a run from C$60.36 to C$73.10 in eight sessions. Volume that size on a down day is distribution, and it left an unfilled gap between C$68.12 and C$72.67 that is now overhead resistance. Price has since stabilised on the 50-DMA rather than through it, which is a hold rather than a recovery.

Levels to watch: support at C$65.16–65.97 (the current shelf, and where price is sitting), then C$62.20, C$61.23 and the late-July base at C$58.70–59.04. Resistance at C$68.12 (the gap floor), C$70.92 (the 200-DMA) and C$73.10 (the 12 August close; the intraday high was C$74.62). A close back above the 200-DMA on volume would be the first genuine change of character.

Six months of daily closes on TSX:PAAS, 17 February to 14 August 2026, with the 50-day simple moving average. The January high of C$95.39 is off the left of this window. The vertical drop at the right edge is 13 August: −9.8% on 2.41× average volume after the Q2 double miss. Note where the NAV line sits relative to the price — that gap is §4's entire argument.

Forecast:

ENTRY — Fundamental group: price below C$34.75.
→ FORECAST: unlikely on price; more likely to be met by the NAV rising.
→ BASIS: a fall to C$34.75 is −47% and below the 52-week low of C$42.43; nothing in the §11 bear case reaches it. The realistic path is the other one — the NAV moving up to meet the price. §4 quantifies it: silver at US$72.86 and gold at US$4,940 (+12.1% and +12.8% from settled spot) takes the ratio to 1.40×, and at spot metals a 13.0-year mine plan instead of 9.37 does the same. The September reserve statement is the dated event that could deliver the second.
→ CONFIDENCE: Low for the price path, Moderate for the NAV path within 12 months.

ENTRY — Technical group: 50-DMA reclaim on >1.5× volume, RSI 35–65.
→ FORECAST: catalyst-dependent, first realistic window 19–20 August.
→ BASIS: price is already above the 50-DMA, so only the volume condition is outstanding, and volume conditions cannot be time-projected — they need an event. The FOMC minutes on 19 August are the next one capable of producing a 1.5× session. RSI at 49.98 is comfortably inside the band and would stay there on a move of up to about +8%. The trap: the same event can just as easily produce the 1.5× session to the downside.
→ CONFIDENCE: Moderate that the condition resolves within two weeks; Low that it resolves upward.

ENTRY — Technical group, pullback branch: tested bounce with a higher low.
→ FORECAST: 3–6 weeks if the C$65.16–65.97 shelf holds.
→ BASIS: price has held the shelf for two sessions. A higher low requires a retest above C$65.40 and a bounce; at an ATR of C$3.03 that is roughly one to two weeks of range, and confirmation needs a further week. If C$65.40 goes, the next tested support is C$62.20 and the clock resets.
→ CONFIDENCE: Moderate. Note this path opens the entry ladder; it does not lift the Valuation-Ceiling gate, which caps the signal at HOLD independently.

ENTRY — Catalyst group: post-earnings +5% on 2× volume.
→ FORECAST: not available until early November 2026.
→ BASIS: the Q2 catalyst has been and gone at −9.78%. The next dated company catalyst capable of firing this group is Q3 results, and the substantive one before it is the annual reserve statement in September, which is not an earnings event and therefore cannot satisfy this group's conditions as written.
→ CONFIDENCE: High that it stays unmet until November.

Forecast:

EXIT — Stop-Loss: two closes below C$58.50.
→ FORECAST: unlikely in the next 4–6 weeks unless the metals break.
→ BASIS: C$58.50 is 11.1% below and 2.41 ATR away, and below the late-July base that held on the way up. Getting there needs the §11 bear trigger — silver back toward US$56 — not ordinary drift.
→ RISK TRIGGER: the 19 August FOMC minutes. A hawkish read that pushes September-hike odds well above 50% would take both metals down and this name down by more, at a beta of 1.55.

EXIT — Thesis Invalidation: 1 of 5 conditions live — the guidance softening — against a rule that needs two.
→ FORECAST: the decisive test is the September reserve statement.
→ BASIS: the two conditions that could take it to two are the gold-cost path and reserve replacement, and neither can resolve early. The gold-cost condition is not live but is close — Q2 printed US$1,984 against a US$2,050 trigger that needs two consecutive quarters, so it cannot fire before Q3 results in November. Reserve replacement cannot resolve before the September statement, and if reserve life falls below 8 years the §4 NAV falls with it and the case gets worse, not better.
→ CONFIDENCE: Moderate that it stays clear through September.

EXIT — Profit-Target: price at C$90.53 with RSI above 70.
→ FORECAST: unlikely inside 12 months except on the bull path.
→ BASIS: C$90.53 is +37.5% and above the §11 base case of C$68; only the 25%-probability bull branch reaches it. It also requires an analyst panel that has not yet marked to the Q2 miss to hold its targets.
→ CONFIDENCE: Low.

Position sizing not computed — no portfolio allocation or role was specified for this run, and the framework does not invent one. Two pieces of context that apply regardless of size:

• The §12 Conviction Ladder reads Wait (0 of 3 entry paths met), which is a sizing factor of 0×. There is no entry edge to size.
• Volatility: ATR(14) is C$3.03, or 4.6% of price per day, against roughly 0.7% for the S&P 500. Beta 1.55. The 52-week range runs C$42.43 to C$95.39 — the stock has more than doubled and then halved inside a year. A 5% position behaves like a 7.7% one in risk terms.

{
  "ticker": "PAAS.TO",
  "company": "Pan American Silver Corp.",
  "brand": "",
  "currency": "CAD",
  "reporting_currency": "USD",
  "exchange": "TSX",
  "exchange_ticker": "TSX:PAAS",
  "isin": "CA6979001089",
  "api_ticker": "PAAS.TO",
  "us_ticker": "PAAS",
  "finder_ticker": "PAAS",
  "finder_exchange": "\ud83c\udde8\ud83c\udde6 TSX",
  "analysis_status": "on-going",
  "gics_sector": "Materials",
  "sector": "Materials",
  "sub_industry": "Silver + Gold Mining",
  "country": "Canada",
  "lifecycle_stage": "mature_cyclical_producer",
  "date": "2026-08-17",
  "version": "v6",
  "price_at_rating": 65.82,
  "price_asof": "2026-08-14 close - the last bar published in the daily price series at run time. The 17 Aug session had not yet appeared in the series when this report was struck; the 17 Aug quote showed C$66.78 (+1.5% on a 65.82 previous close) and is noted rather than used, because a stamped price should come from a bar in the series.",
  "usd_cad_rate": 1.39271,
  "usd_cad_rate_asof": "2026-08-14",
  "shares_outstanding_m": 414.6,
  "shares_outstanding_basis": "Market cap C$27.688bn / the C$66.78 live quote = 414.6m, cross-checked against the Q2 2026 weighted-average diluted count of 420.345m less 4.352m repurchased in-quarter and 2.456m repurchased after quarter-end (implying ~415.6m). The two agree to 0.24%; the reconciled figure is used rather than the provider market cap taken on trust.",
  "eps_trailing": 4.57,
  "eps_trailing_basis": "TTM net income US$1,380.3m / 420.3m weighted-average shares = US$3.2841, x 1.39271 = C$4.5738. Quarterly EPS sum 0.44 + 1.09 + 1.08 + 0.70 = US$3.31 corroborates.",
  "trailing_pe": 14.39,
  "forward_pe": 9.65,
  "signal_short": "WAIT",
  "signal_medium": "HOLD",
  "signal_long": "HOLD",
  "primary_signal": "HOLD",
  "composite_short": 52,
  "composite_medium": 53,
  "composite_long": 58,
  "quality_score": 72,
  "valuation_score": 36,
  "timing_score": 52,
  "driver_score": 66,
  "driver_margin_weights": "63.27% gold / 36.73% silver, from the section-4 mine plan at the base deck (gold margin US$1,505.0m, silver margin US$873.6m). CROSS-CHECKED against the revenue mix, which is 69.5% gold / 30.5% silver at the same deck (gold US$2,800.0m, silver US$1,227.2m), and against Q2 actuals of roughly 55% gold / 26% silver / 19% base metals on attributable revenue. Margin is the correct weight for a DRIVER score because the equity is geared to the cash margin rather than the top line; silver's margin share exceeds its revenue share precisely because its cost ratio is lower. Both measures agree that this is majority-gold and neither supports scoring it as a pure silver name. An earlier draft used 58/42 from an uncorrected silver volume; the corrected weights do NOT change any blended driver score (Short 60.4, Medium 65.7, Long 71.7 - identical to 3 significant figures).",
  "driver_score_short": 60,
  "driver_score_medium": 66,
  "driver_score_long": 72,
  "quality_detail": {
    "industry_benchmark_name": "AISC Margin (Mining)",
    "industry_benchmark_value": 61.3,
    "industry_benchmark_score": 89,
    "moat_score": 45,
    "roic_percentile_vs_peers": 72,
    "capital_allocation": 70,
    "management_skin_in_game": 45,
    "aisc_silver_q2_2026_usd_oz": 17.8,
    "aisc_silver_h1_2026_usd_oz": 12.64,
    "aisc_silver_fy2026_guidance_usd_oz": "15.75-18.25",
    "aisc_gold_q2_2026_usd_oz": 1984,
    "aisc_gold_h1_2026_usd_oz": 1918,
    "aisc_gold_fy2026_guidance_usd_oz": "1700-1850 (guided to the HIGH end)",
    "silver_production_q2_2026_moz": 6.47,
    "silver_production_fy2026_guidance_moz": "25.0-27.0",
    "gold_production_q2_2026_koz": 165.9,
    "gold_production_fy2026_guidance_koz": "700-750 (guided to the LOW end)",
    "net_cash_usd_m": 981,
    "cash_and_sti_usd_m": 1822,
    "total_debt_usd_m": 841,
    "total_debt_provider_lease_inclusive_usd_m": 841,
    "total_debt_note": "CORRECTED AFTER AUDIT. The Q2 2026 release states 'Total Debt of $841 million is primarily related to Senior Notes and lease obligations' and 'Cash and cash equivalents and short-term investments of $1.8 billion'. So US$841m IS the company figure and it is lease-inclusive by the company's own description; get_yahoo_quote returns the identical 841,000,000. An earlier draft of this report carried US$717m of debt and US$1,698m of cash from a secondary summary, described US$841m as a provider artefact, and claimed 1,698 - 717 = 981 'reconciles exactly' - which it does only because the same US$124m of leases had been removed from BOTH sides. The true identity is 1,822 - 841 = 981. NET CASH OF US$981m WAS AND REMAINS CORRECT, so the NAV, enterprise value and FCF yield are unaffected; what was wrong was the provenance and the direction of the lease-inclusive rule.",
    "buyback_q2_2026_shares_m": 4.352,
    "buyback_q2_2026_price_usd": 51.46,
    "buyback_postquarter_shares_m": 2.456,
    "buyback_postquarter_price_usd": 44.36,
    "reserves_pp_silver_moz": 452.3,
    "reserves_pp_gold_moz": 6.339,
    "reserves_pp_gold_used_in_nav_moz": 6.3,
    "reserves_pp_gold_note": "The company reserves page states 6,339koz. The NAV uses 6.3Moz, which shortens the gold life from 9.06 to 9.00 years - conservative, and against this report's own conclusion rather than for it.",
    "reserves_pp_silver_producing_moz": 187.8,
    "reserves_pp_silver_producing_silver_segment_moz": 147.8,
    "reserves_pp_silver_gold_segment_byproduct_moz": 40.1,
    "reserves_per_mine_reconciliation": "Per-mine P&P silver read individually from the company mine pages sums to 452.4Moz against the 452.3Moz headline (0.02% apart): La Colorada 90.7, Huaron 40.7, San Vicente 11.4, Cerro Moro 5.0, Escobal 264.5, El Penon 22.1, Shahuindo 16.4, Minera Florida 1.6. Gold sums to 6,061koz against the 6,339koz the company reports: Jacobina 3,127, Shahuindo 977, Timmins 846, El Penon 626, Minera Florida 271, Cerro Moro 150, La Colorada 64; the 278koz difference is the Dolores residual-leach stockpile, which carries no reserve statement and is credited nowhere in this NAV. The NAV uses 6.3Moz rather than 6.339Moz, shortening the gold life from 9.06 to 9.00 years - conservative, and against this report's own conclusion.",
    "reserves_asof": "2025-06-30",
    "reserve_life_years": 9.374,
    "reserve_life_source": "Margin-weighted blend of a 10.02-year SILVER-SEGMENT stream (147.8Moz of P&P silver at La Colorada, Huaron, San Vicente and Cerro Moro / 14.752Moz produced by those same four mines in 2025 - a matched numerator and denominator, with Juanicipio assigned the same life because the 30 Jun 2025 statement predates the MAG acquisition) and a 9.00-year GOLD stream (6.3Moz / 700koz FY2026 low-end guidance), weighted 36.73% / 63.27% by modelled margin at the base deck = 9.374 years. Escobal's 264.5Moz is excluded from both cash flow and life, and the 40.1Moz of gold-segment by-product silver is excluded from the silver stream because its revenue is already credited inside the gold AISC. CORRECTED IN ROUND 2: this field previously carried the superseded 9.1-year / 20.6Moz / 42.05-57.95 basis.",
    "reserve_freshness_check": "The statement is effective 30 June 2025 and is therefore 13.5 months old, with the next due around September 2026. Two material consequences, both disclosed in section 4: (1) it predates the 4 September 2025 MAG Silver close, so Juanicipio's reserves are absent while its ~6.05Moz of attributable output IS in the production denominator - the matched-basis correction puts the silver-segment life at 10.02 years on 147.8Moz / 14.752Moz; (2) La Colorada Skarn and Navidad were checked individually and report NO reserves, so they are correctly absent from the 452.3Moz rather than double-counted."
  },
  "valuation_detail": {
    "nav_basis": "COMPUTED BOTTOM-UP at r = 9.13% over a 9.374-year margin-weighted producing reserve life, metals held at a base-case deck of US$59.00/oz silver and US$4,000/oz gold, AISC struck at the company's own FY2026 guidance steer (silver midpoint US$17.00, gold HIGH end US$1,850), net cash credited UNDISCOUNTED, Escobal carried at a 25%-risked US$766m, and La Colorada Skarn, Navidad and all reserve conversion carried at ZERO. BY-PRODUCT CORRECTION APPLIED: the silver stream carries 20.80Moz, NOT the 26.0Moz of total attributable guidance. The gold-segment AISC of US$1,850/oz is stated net of by-product credits which include ~5.2Moz of silver from El Penon, Dolores, Shahuindo and Minera Florida; charging those ounces to the silver segment and taking their revenue again added ~US$248m/yr of margin that does not exist. An earlier draft of this report made exactly that error and published 1.786 instead of 1.894. The correction is validated independently: struck at spot the corrected model gives US$1,650m of annual after-tax cash flow against the company's own H1 2026 attributable FCF of US$832m = US$1,664m annualised, 0.8% apart; the uncorrected model gave US$1,807m, 8.6% too high.",
    "nav_r_struck_usd_per_share": 24.95,
    "nav_r_struck_cad_per_share": 34.75,
    "nav_annuity_factor": 6.1242,
    "nav_after_tax_cashflow_usd_m": 1404.0,
    "nav_pv_operations_usd_m": 8598.5,
    "nav_escobal_risked_usd_m": 766,
    "nav_net_cash_usd_m": 981,
    "nav_net_cash_basis": "DERIVED, not company-stated. The Q2 release prints total debt of US$841m and cash + short-term investments of '$1.8 billion' but no net-cash line. US$981m implies unrounded cash + STI of US$1,822m. At a literal US$1,800m the net cash is US$959m, the NAV C$34.68 and the ratio 1.898 - immaterial to the band.",
    "nav_total_usd_m": 10345.5,
    "nav_tax_rate_pct": 37.0,
    "nav_tax_note": "The Q2 2026 effective rate of 36.98%, not the 27.71% TTM rate, because the company RAISED full-year cash tax guidance to US$585-635m on 12 August. At 30% the NAV is C$37.96 and the ratio 1.734 - still Expensive. (CORRECTED IN ROUND 2: this field previously carried C$40.71 / 1.617 from the superseded pre-by-product model.)",
    "nav_corporate_charge_usd_m": 150,
    "nav_at_spot_deck_cad_per_share": 39.82,
    "nav_at_spot_deck_ratio": 1.653,
    "nav_escobal_zero_cad_per_share": 32.18,
    "nav_escobal_zero_ratio": 2.045,
    "nav_escobal_50pct_cad_per_share": 37.33,
    "nav_escobal_50pct_ratio": 1.763,
    "nav_escobal_basis": "20Moz/yr at the historic run-rate over a 13.2-year life, US$59 silver against a US$16/oz AISC (double the 2017 level), ~30% Guatemalan tax and royalty, US$150m restart capital, deferred 4 years, probability-weighted at 25% = US$766m = C$2.57/share = 7.4% of NAV. CROSS-CHECKED by an unrelated method: 264.5Moz at a conservative in-situ US$3.00/oz = US$794m, agreeing to 3.7%. The BAND is invariant from 0% to 50% probability (2.045 / 1.894 / 1.763, all Expensive) but the DNB adjudication is NOT: at a 0% probability the ratio is 2.045, above the 2.0x at which DNB Trigger 2(a) fires. See dnb_adjudication for why 0% is not the basis adopted.",
    "ratio_crosses_120_at_silver_usd_oz": 81.72,
    "ratio_crosses_120_at_gold_usd_oz": 5540,
    "ratio_crosses_140_at_silver_usd_oz": 72.86,
    "ratio_crosses_140_at_gold_usd_oz": 4940,
    "ratio_crosses_120_at_life_years_deck": 33.88,
    "ratio_crosses_120_at_life_years_spot": 18.81,
    "ratio_crosses_140_at_life_years_deck": 18.22,
    "ratio_crosses_140_at_life_years_spot": 12.96,
    "boundary_binding_lever": "THE DECK BINDS, NOT THE COST - and this is the mirror image of ABX.TO in the same batch, where the 1.20 boundary fell INSIDE the company's own AISC guidance band so cost was the weak point. The 1.40 boundary is a surface in two variables and is published as one: at Q2 actual costs (Ag US$17.80 / Au US$1,984) the ratio is 1.975 and 1.40 needs Ag US$74.48 / Au US$5,049; at the guidance point used here (US$17.00 / US$1,850) the ratio is 1.894 and 1.40 needs Ag US$72.86 / Au US$4,940; at the BEST end of the entire guidance range (US$15.75 / US$1,700) the ratio is still 1.806 and 1.40 needs Ag US$70.94 / Au US$4,810. Run it as a cost-only path and the combined cost base must fall 57.4% - to a silver AISC of US$7.24/oz and a gold AISC of US$788/oz - which is unreachable. So the cost lever moves the ratio only across 1.806-1.975 over its whole guidance range and never reaches 1.40, while a 9-15% move in both metals does. The deck is the live lever; the cost assumption is not the weak point on this name.",
    "boundary_note": "The published weak point. At settled spot metals a 12.96-year mine plan instead of 9.37 clears the Expensive band. Pan American reports 1,130.6Moz of M&I silver outside reserves, but 632.4Moz of that is Navidad (legally blocked) and 309Moz is the La Colorada Skarn (PEA-stage, no reserves) - leaving roughly 189Moz at or adjacent to the producing silver mines, about 1.3x the 147.8Moz silver-segment reserve base, plus 9.9Moz of M&I gold against 6.3Moz of reserves (1.6x, essentially all at producing mines). Full conversion would take both streams to roughly 23 years, well past the boundary. So the verdict is about three and a half years of reserve conversion away from its opposite, and the September 2026 reserve statement is the dated test.",
    "ev_ebitda_matched": 7.75,
    "ev_ebitda_provider_unmatched": 8.14,
    "ev_ebitda_provider_unmatched_note": "The matched figure is EV US$18,613m (market cap US$19,594m less US$981m net cash) over TTM EBITDA of US$2,402.1m = 7.75x, against the 8x Materials line - NOT breached. The provider's 8.137x uses a different enterprise value. The guardrail fires on the PRIMARY multiple, P/NAV 1.894x against 1.5x.",
    "guardrail_adjudication": "Adjudicated BY HAND because the linter's guardrail arm keys on an earnings- or book-value basis string and cannot read a P/NAV. Materials line = 1.5x P/NAV (or 8x EV/EBITDA). P/NAV 1.894x >= 1.5x - BREACHED, so Gate 3 fires on the guardrail arm as well as on the >=1.40x band arm. EV/EBITDA 7.75x < 8x - not breached.",
    "fcf_yield": 8.25,
    "fcf_ttm_usd_m": 1535,
    "attributable_fcf_q2_2026_usd_m": 344,
    "attributable_fcf_h1_2026_usd_m": 832,
    "historical_valuation_decile": 8,
    "price_to_book": 2.75,
    "optionality_tilt_applied": 0,
    "optionality_tilt_note": "ZERO, and arithmetically unavailable rather than discretionary: SKILL L465 caps an Expensive-band score below 40, so a tilt on 36 would push it outside the band. The optionality is instead carried as the explicit statement that C$31.07/share - 47% of the price - is what is PAID for un-reserved ounces, not a discount received.",
    "prior_anchor_note": "DISTINCT FROM THE ABX.TO / NEM CORRECTION THIS WEEK: on those two names the prior anchor had been struck on an EARNINGS MULTIPLE when SKILL L570 gives miners P/NAV primary, so the lens was wrong. Pan American's prior anchor was already labelled P/NAV; the defect was that it was never COMPUTED. The 7 August calibration recorded actual_multiple 0.7 / warranted 1.0 / ratio 0.7 / val_band attractive. That P/NAV was SELECTED, not computed - its own section 4 described it as 'a discount to NAV at spot metals (~0.7x)' with no NAV build. This run computes it from the mine-by-mine reserve disclosure. Do not read 0.7 -> 1.894 as a deterioration in the business; it is the correction of an unsourced anchor."
  },
  "timing_detail": {
    "mtf_confluence": 60,
    "risk_reward_score": 41,
    "position_risk_score": 45,
    "relative_strength_score": 28,
    "sentiment_score": 42,
    "news_tone_score": 30,
    "macro_overlay_score": 58,
    "catalyst_clustering_score": 50,
    "dynamic_macro_weight": 0.2,
    "atr14": 3.03,
    "atr_pct_of_price": 4.6,
    "rsi14": 49.98,
    "sma50": 65.02,
    "sma50_slope": "falling - C$66.30 on 2026-07-31 to C$65.02 on 2026-08-14",
    "sma200": 70.92,
    "macd": 1.358,
    "macd_signal": 0.733,
    "macd_hist": 0.624,
    "range_52w_position_pct": 44.2,
    "high_52w": 95.39,
    "high_52w_date": "2026-01-26",
    "low_52w": 42.43,
    "low_52w_date": "2025-08-19",
    "earnings_day_move_pct": -9.78,
    "earnings_day_volume_ratio": 2.41,
    "volume_ratio_convention": "All volume ratios in this report use the INCLUSIVE trailing 20-session average (the 20 sessions ending on the day being measured). On that single convention 13 Aug is 2,474,400 / 1,027,740 = 2.41x and 14 Aug is 1,045,700 / 1,047,160 = 1.00x. An earlier draft mixed conventions, quoting 2.61x for 13 Aug on a prior-20 basis alongside 1.00x for 14 Aug on the inclusive basis; both were individually correct and inconsistent side by side.",
    "beta": 1.55
  },
  "val_band": "expensive",
  "actual_multiple": 1.894,
  "warranted_multiple": 1.0,
  "warranted_ratio": 1.894,
  "val_multiple_basis": "P/NAV - the Materials/Miners primary multiple per SKILL L570 - with the NAV struck at the framework's own discount rate r = 9.13%, exactly as step 6 (L490) specifies for a miner. NAV is COMPUTED BOTTOM-UP from Pan American's mine-by-mine reserve disclosure, not selected: silver-segment production of 20.80Moz/yr (total attributable guidance 26.0Moz LESS ~5.2Moz of gold-segment by-product silver, whose revenue is already credited inside the gold AISC) at a base-case deck of US$59.00/oz less a US$17.00/oz segment AISC, plus gold of 700koz/yr (the company's own low-end steer) at US$4,000/oz less a US$1,850/oz segment AISC (its own high-end steer), less US$150m of non-AISC corporate, project, care-and-maintenance and net interest, taxed at the 37.0% Q2 effective rate = US$1,404.0m/yr after tax; discounted over a 9.374-year margin-weighted producing reserve life (silver 147.8Moz / 14.752Moz = 10.019yr at 36.727% of margin; gold 6.3Moz / 0.700Moz = 9.000yr at 63.273%) at an annuity factor of 6.1242 - the life is carried at three decimals because at a rounded 9.37 the factor is 6.1224 and the NAV C$34.74, so pairing a rounded life with an unrounded factor would read as an arithmetic slip; the ratio is 1.894 either way = US$8,598.5m; plus a 25%-risked Escobal at US$766m and net cash of US$981m undiscounted. Total US$10,345.5m / 414.6m shares = US$24.95 = C$34.75. Actual C$65.82 / C$34.75 = 1.894x against a warranted 1.00x -> ratio 1.894 -> EXPENSIVE band. Blending the two streams into one annuity is worth US$8,598.5m against US$8,592.3m for discounting each at its own life with the corporate charge allocated pro rata - 0.07% generous, immaterial. Guardrail adjudicated by hand: P/NAV 1.894x against the 1.5x Materials line - BREACHED; EV/EBITDA 7.75x against the 8x line - clear. Gate 3 fires on BOTH the >=1.40x band arm and the guardrail arm.",
  "discount_rate_r": 9.13,
  "risk_free_10y": 4.63,
  "risk_free_source": "FRED DGS10 at 2026-08-13 = 4.63%, giving r = 9.13%. Two alternatives checked and both inert: the 14 Aug DGS10 print of 4.68% (which matches the price and FX stamp date) gives r = 9.18%, AF 6.1114, NAV C$34.69, ratio 1.897; the 12 Aug macro report's 4.70% gives r = 9.20%, AF 6.1063, NAV C$34.67, ratio 1.899. The 4.63% basis is the most conservative of the three and all three are Expensive.",
  "g_near": 0.0,
  "g_term": 0.0,
  "anchor_growth_basis": "Both zero, and deliberately so: a mining NAV discounts a finite mine plan at flat nominal cash flows over the reserve life, so no growth term enters this anchor. The two-stage growth formula applies to an earnings-multiple instantiation, which SKILL L570 does not give a miner. Recorded as 0.0 rather than omitted so a later run can see the basis rather than infer it. This also makes the DNB Trigger 2(a) growth carve-out structurally unavailable - which is why the 1.894 vs 2.00 margin (5.3%), not a growth argument, is what keeps this name out of a hard Do-Not-Buy.",
  "clean_pe": 16.31,
  "clean_pe_at_forward_tax_rate": 18.71,
  "clean_peg": 0.62,
  "nonop_pct_of_net_income": 16.8,
  "nonop_basis": "Non-operating income EXCLUDING interest = US$231.5m against TTM net income of US$1,380.3m. Above the 15% line, so net margin, P/E and ROE were recomputed on operating earnings (PEG was NOT - the 0.62 carried in clean_peg is the provider's forward PEG, not an operating-earnings recomputation, and section 4 says so): TTM operating income US$1,685.1m x (1 - 0.2771) = US$1,218.2m = US$2.898/share = C$4.037, giving a clean P/E of 16.31x against the reported 14.39x. TAX-BASIS NOTE: the clean multiple is struck at the 27.71% TTM effective rate because it normalises HISTORIC TTM earnings, whereas the NAV is struck at the 37.0% Q2 rate because it discounts FORWARD cash flows and the company raised full-year cash tax guidance on 12 August. Struck instead at 37%, clean EPS is C$3.52 and the clean P/E is 18.71x - richer, not cheaper. Recorded but NOT load-bearing either way: the anchor is P/NAV, so no earnings multiple sets the band.",
  "fcf_yield": 8.25,
  "roe": 22.4,
  "fair_value_est": 34.75,
  "stop_loss": 58.5,
  "target_price": 68,
  "fair_value_vs_target_note": "These measure different things and both are needed. fair_value_est C$34.75 is the r-struck NAV - the anchor that sets the band and the Fundamental entry condition. target_price C$68 is the section 11 BASE-CASE 12-month path, which assumes the market keeps paying about 1.58x that NAV at spot metals. A HOLD with a base case above spot is coherent: the framework's HOLD is a statement about price versus proven value, not a forecast of a fall.",
  "scenario_base_target": 68,
  "scenario_bull_target": 95,
  "scenario_bear_target": 46,
  "scenario_probabilities": {
    "bull": 25,
    "base": 55,
    "bear": 20
  },
  "scenario_weighted_target": 70.35,
  "hard_gate_state": "caution",
  "gates_triggered": [
    "Valuation Ceiling"
  ],
  "gates_caution": [
    "Section-8 WAIT-FOR-EVENT override: FOMC Minutes 2026-08-19 falls two trading days from this report, inside the 3-trading-day window for a High-macro-sensitivity Materials name. Applied literally - it sets signal_short = WAIT and next_update_date = 2026-08-20. Recorded as a caution rather than a hard gate because it caps the Short horizon only.",
    "Earnings quality (step 7b): non-operating income excluding interest = 16.8% of TTM net income, above the 15% normalisation line. Clean P/E 16.31x vs reported 14.39x. Not load-bearing - the anchor is P/NAV.",
    "Regulatory: Escobal on care and maintenance since 2017 with the ILO 169 consultation in Phase 2 of 3, and Navidad blocked by Chubut Law 5001. Adjudicated as NOT a Gate-5 binary - Escobal is 7.4% of NAV at the 25% risking used, so neither resolution moves the stock 20% alone."
  ],
  "do_not_buy_triggers": [],
  "dnb_adjudication": "None fire, and 2(a) is the CLOSEST CALL IN THE REPORT rather than a formality. Trigger 2(a) needs >=2.0x warranted or >=1.5x the guardrail line (2.25x P/NAV, not reached). The anchor ratio is 1.894, so it does not fire - but by only 5.3%, and the growth carve-out is STRUCTURALLY UNAVAILABLE because a mining NAV has g_near = g_term = 0.0, so the margin is the entire defence. ONE PUBLISHED SENSITIVITY CROSSES THE LINE: valuing Escobal at exactly zero gives 2.045, which would meet 2(a). That basis is not adopted, for three stated reasons - Escobal's 264.5Moz is proven-and-probable RESERVE, disclosed in the same 30 Jun 2025 statement this NAV relies on for the OTHER mines' lives, so treating that statement as authoritative for life while valuing its largest single reserve at nil is a basis mismatch. (Escobal itself is excluded from both the cash flow and the life - see reserve_life_source; the mismatch is about trusting one statement selectively, not about double-counting Escobal's ounces); the 25%-risked US$766m is corroborated to within 3.7% by a conservative in-situ US$794m computed by an unrelated method; and the basis actually ADOPTED is what the trigger is measured against, not the floor case. THE ADJUDICATION RESTS ON THOSE THREE REASONS. An earlier draft added a FOURTH - that a specification producing a hard DO NOT BUY on a net-cash producer at an 8.25% FCF yield would be mis-specified for the asset class - which is an appeal to outcome rather than an application of the rule; it has been WITHDRAWN from the adjudication and RESTATED AS AMENDMENT CLAUSE (e) in framework_amendment_proposed, where it is written out in full rather than merely referred to. Trigger 2(b) needs the Expensive band PLUS a live de-rating catalyst; the macro report's four tails were each tested - AI-concentration and private-credit have no cohort relevance to a silver miner (and private credit is 'building, NOT armed' in any case), Hormuz is supportive rather than adverse for precious metals, and the September-hike tail - which the macro state file records with status 'live', reserving the word 'armed' for the AI tail - fails 2(b) on the COHORT TEST rather than on its status. SKILL L1085 requires the tail to materially apply to THIS name: real cohort exposure, not membership of a broad category. A market-wide rate tail is the precise analogue of the 'it's a tech stock' case that clause exists to exclude; if a live rate tail counted, every Expensive-band Materials and Financials name would become a Do-Not-Buy on any live rate tail. The secondary point is that its own stated trigger, the 19 Aug minutes, has not happened, and the standing rule is that an armed-but-not-triggering tail is a Valuation-Ceiling cap rather than a prohibition. CORRECTED AFTER AUDIT: an earlier draft led with 'armed but not triggering', which re-labelled the macro file's own wording; this report now uses 'live' throughout and rests the adjudication on the cohort test. The relative arm does not fire either, and the warrant matters: it needs the primary multiple in the TOP DECILE (90th percentile+) of its own five years AND no growth acceleration. The historical valuation decile is 8 (80th percentile), below the bar; and growth is accelerating on a FORWARD measure - forward P/E 9.65x against trailing 14.39x implies roughly 49% forward EPS growth. An earlier draft cited trailing revenue growth for this forward test, which was the wrong warrant even though the conclusion holds. Triggers 1, 3, 4 and 5 are all clear: net cash, no sustained negative revisions, no insider selling spike, no structural business-model threat.",
  "entry_groups_met": 0,
  "entry_conviction": "Wait",
  "short_entry_confirmed": false,
  "short_hold_reason": "gate",
  "short_cap_reason": "Two independent grounds, and the published WAIT label comes from the first. (1) SECTION-8 WAIT-FOR-EVENT OVERRIDE: Materials is a High-macro-sensitivity sector and the FOMC minutes fall on Wed 19 Aug 2026, two trading days from the 17 Aug report date. SKILL L1894 has it firing 'regardless of composite score' and L1577 names the same 3-day window as normative for scheduling - this report relies on that clause to set next_update_date = 2026-08-20, so it cannot decline the label the same rule names. (2) SHORT TECHNICAL-CONFIRMATION CAP (short_entry_confirmed = false, the longer-lived of the two): the Technical entry group is UNMET - the 14 Aug close of C$65.82 is above the 50-DMA of C$65.02 but volume was 1,045,700 against a 20-day average of 1,047,160 = 1.00x versus the >1.5x required, and the pullback branch is unavailable because 13 Aug printed a LOWER low (C$65.40) on 2.41x volume. The Catalyst group is also unmet: the post-earnings move was -9.78%, not the required >+5%. Trigger to lift the technical cap: a close above C$65.02 on >1.5x average volume with RSI inside 35-65, OR a tested pullback into the C$65.16-65.97 shelf with a higher low. NOTE that lifting the technical cap would NOT produce a BUY - the base matrix returns HOLD on all three horizons from Quality 72 x Valuation 36 (Expensive) before any cap is applied, and the Valuation Ceiling caps it again. The Section-8 window clears on 20 Aug; the valuation cap is what actually governs this name.",
  "short_hold_reason_note": "'gate' is recorded because the Section-8 event override is the cap that PRODUCES THE PUBLISHED WAIT LABEL, and the reason code should track the label-setting cap. 'technical_pending' is equally live and is recorded in full inside short_cap_reason so no information is lost. Unlike ABX.TO this run, the technical cap is NOT the binding constraint here either - the Valuation Ceiling is, and it caps all three horizons rather than the short alone.",
  "exit_groups_live": 0,
  "exit_action": "Hold",
  "framework_amendment_proposed": "THE SKILL DOES NOT SPECIFY THE MINE LIFE OVER WHICH A MINER'S NAV IS STRUCK, and on this name that single unbounded author choice decides the report. L490 pins r and says only 'commodity price held at the base-case deck'; L570 adds 'use mid-cycle commodity prices for normalisation'. Neither bounds the life. Here the disclosed producing reserve life is 9.37 years and, at settled spot metals, a 12.96-year life would take the ratio below 1.40 and lift the Gate-3 cap on all three horizons - while Pan American holds roughly 189Moz of M&I silver at its producing mines (once Navidad and the La Colorada Skarn are excluded) and 9.9Moz of M&I gold against a 6.3Moz reserve base. So a three-and-a-half-year difference in one unstated assumption is the difference between HOLD everywhere and a live signal. The same silence also leaves the SEGMENTATION unspecified, which is how an earlier draft of this very report double-counted gold-segment by-product silver and published 1.786 instead of 1.894. It also means the framework will systematically read underground miners (short, continuously-replaced reserve lives - El Penon carries 5.4 years of gold reserves and has operated continuously since 1999) as more expensive than open-pit miners (Barrick: a 27.6-year reserve life, ratio 1.218, FULL band) for reasons of mine geometry rather than price. PROPOSAL: (a) require the life to be struck on the disclosed reserve base with MATCHED asset sets in numerator and denominator, stating every exclusion by name; (b) require every miner report to publish the life AND the metal prices at which the ratio crosses 1.40 and 1.20, so the indeterminacy is visible rather than buried; (c) rule explicitly on whether measured-and-indicated resources at PRODUCING mines may enter the NAV and at what haircut - the current silence means each report re-decides it; (d) require the segment-AISC / by-product convention to be stated explicitly, because a by-product ounce credited inside another segment's cost and then revenued again is a silent 6% error on the NAV; and (e) RECALIBRATE DNB TRIGGER 2(a) FOR A P/NAV BASIS. The 2.0x-warranted threshold reads naturally against an earnings multiple, where 2x fair value is plainly extreme. Against a P/NAV struck on proven reserves alone it is not: on this name a ratio of 1.894 - a net-cash producer generating an 8.25% FCF yield at 7.75x EV/EBITDA and 9.65x forward earnings - sits within 5% of a hard prohibition, and the Escobal-at-zero floor case crosses it outright at 2.045. Miners routinely trade at 1.5-2.5x a reserves-only NAV precisely because reserves understate mine life, so a threshold that treats 2.0x as extreme will fire on ordinary miners. PROPOSAL: either set a separate, higher 2(a) threshold for a P/NAV instantiation, or require the trigger to be tested against an NAV that includes a bounded measured-and-indicated allowance. Raised, not self-applied: this report applies the 2.0x threshold as written, does not fire on the adopted basis by 5.3%, and publishes the crossing sensitivity in full rather than suppressing it.",
  "competitive_share_trajectory": "stable",
  "competitive_threat_level": "moderate",
  "economic_alignment_stance": "Trend-Following",
  "economic_alignment_conviction": 74,
  "economic_alignment_pressure": "Tailwind",
  "economic_alignment_source": "watchlist-signal",
  "economic_alignment_amplified": false,
  "economic_alignment_note": "SOURCE UPGRADED from the prior run's sector-map: PAAS.TO now appears directly in the 12 Aug macro report's Economic Watchlist Forecast at O / SO / SO. Pressure is Tailwind, but it amplified NOTHING - the base signal is HOLD on every horizon and HOLD never amplifies, and independently the Expensive band bars amplification outright (L1300/L494). The macro report's own caveat on this ticker is carried into sections 5 and 11 rather than smoothed away: 'silver is ~46% below its January high and below its 200-day average. Very high beta cuts both ways.'",
  "macro_report_date": "2026-08-12",
  "driver_commodity_trend": "TWO DRIVERS, weighted 63% GOLD / 37% SILVER by modelled margin contribution - this is a gold producer with a silver name, and scoring it as a pure silver play would misread it. Raw settled closes, auto_adjust=False, 14 Aug 2026, quoted from the futures not the ETFs. SILVER SI=F US$64.99: ABOVE a FALLING 50-DMA of 61.79 but BELOW the 200-DMA of 70.58; +15.97%/4wk, +7.16%/6wk, -8.07%/8wk. GOLD GC=F US$4,380.40: ABOVE a FALLING 50-DMA of 4,158.01 but BELOW the 200-DMA of 4,486.27; +9.16%/4wk, +6.51%/6wk, +0.49%/8wk. Both are a rebound INSIDE a correction, not a clean structural tailwind, and neither is available to be described as one. STEP-2b SHORT CAP DOES NOT FIRE: the settled 4- and 6-week windows are positive on both metals and neither is below its 50-DMA. It makes no practical difference - the short driver lands at 60 (Neutral) on its own components, which is where the cap would have put it. What the negative 8-week silver leg earns is a place in the section 11 bear case and the section 12 invalidation floor as a LIVE near-term risk. AISC margins: silver 72.6% of spot against a Q2 AISC of US$17.80; gold 54.7% against US$1,984 - but gold AISC is up 23% YoY (US$1,611 -> US$1,984) and guided to the HIGH end, so the gold margin is compressing faster than the metal is moving, which is why gold scores BELOW silver on current state despite the heavier weight. Components - Silver: historical 60, current 62/70/78, forward 60/68/76 -> 61/67/73. Gold: historical 63, current 60/66/74, forward 58/65/72 -> 60/65/71. Blended -> Short 60 (Neutral), Medium 66 (Tailwind), Long 72 (Tailwind). Medium and long clear the 65 amplification threshold and amplify NOTHING, because the base signal is HOLD and the Expensive band bars amplification on every horizon regardless.",
  "overall_confidence": 58,
  "confidence_quality": 70,
  "confidence_valuation": 60,
  "confidence_timing": 58,
  "confidence_driver": 55,
  "analyst_consensus_target": 91.22,
  "analyst_target_high": 130.92,
  "analyst_target_low": 73.81,
  "analyst_target_median": 90.53,
  "analyst_target_upside_pct": 38.6,
  "analyst_target_basis": "Yahoo panel in USD (mean 65.50, median 65.00, high 94.00, low 53.00, n=8) converted at 1.39271. FMP's consensus of US$73 was NOT used: its summary shows 0 targets in the last month and 1 in the last quarter (at US$53), so it is stale-weighted. CRITICAL CAVEAT - both panels predate the 12 Aug double miss and the -9.8% session; no analyst has marked to the Q2 result. A recency discount is applied and the consensus sub-factor scores 60 rather than the 85-100 the raw upside would earn.",
  "analyst_grades_consensus": "Buy",
  "analyst_bullish_pct": 52,
  "analyst_coverage_count": 25,
  "analyst_coverage_basis": "25 = the total FMP grades-consensus panel (13 buy / 10 hold / 2 sell). The 7 Aug calibration recorded 13, which is the BUY count rather than the panel size; the bullish percentage of 52% is unchanged on either reading.",
  "recent_upgrades_30d": 0,
  "recent_downgrades_30d": 0,
  "fmp_rating": "A-",
  "fmp_overall_score": 4,
  "relative_strength_vs_spy": 4.7,
  "relative_strength_vs_sector": -12.7,
  "relative_strength_note": "1-month: PAAS +7.48%, SPY +2.8%, GDX +20.15% (vs SPY +4.7pp, vs the miner complex -12.7pp). 3-month: PAAS -21.57%, SPY +3.77%, GDX -4.24% (vs SPY -25.3pp, vs GDX -17.3pp). Pan American is a LAGGARD inside a sector that is working, and that is the uncomfortable number in the Timing pillar.",
  "dividend_ttm_usd": 0.62,
  "dividend_declared_usd": 0.184,
  "dividend_forward_cad": 1.025,
  "dividend_forward_yield_pct": 1.56,
  "dividend_action": "US$0.184 declared 2026-08-12 IS a genuine raise from the US$0.180 paid in May and February 2026 - verified against the payment history with get_stock_dividends, not a search summary. Ex and record 24 Aug, paid 4 Sep. The trailing four payments (0.18 + 0.18 + 0.14 + 0.12 = US$0.62) reconcile exactly to the provider's dividendPerShareTTM. One secondary source reported the declaration as US$0.18; the tool figure is used.",
  "q2_2026_results": {
    "reported": "2026-08-12 after close",
    "revenue_usd_m": 1124,
    "revenue_prior_year_usd_m": 812,
    "revenue_yoy_pct": 38.4,
    "revenue_attributable_usd_m": 1268,
    "revenue_attributable_prior_year_usd_m": 811,
    "revenue_note": "US$1,124m consolidated and US$1,268m attributable, both read from the Q2 2026 release's own highlights tables (Revenue $M: 2026 1,124 / 2025 812; attributable 1,268 / 811). FMP's get_income_statement returns US$1,095.5m on a different consolidation basis; the COMPANY figure is used. Round 2 of the audit restored this after round 1 wrongly called US$1,124m unsourced - see post_audit_fixes_round2.",
    "net_earnings_usd_m": 305,
    "eps_basic_usd": 0.72,
    "adjusted_earnings_usd_m": 308,
    "adjusted_eps_usd": 0.73,
    "consensus_adjusted_eps_usd": 0.84,
    "eps_surprise_pct": -13.1,
    "revenue_consensus_usd_m": 1131.05,
    "revenue_actual_fmp_usd_m": 1103.3,
    "revenue_surprise_basis_note": "THE -2.5% REVENUE MISS IS COMPUTED ON FMP'S OWN PAIR (get_earnings_calendar: revenueActual 1,103.3 vs revenueEstimated 1,131.05), NOT on the company figure. Do not compute a surprise from revenue_usd_m (1,124, the company's consolidated highlights table) against revenue_consensus_usd_m (1,131.05, FMP) - those are two different bases and the result would be a spurious -0.6%. The company figure is the one used everywhere else in this report; the FMP pair is used only for the consensus comparison, because a surprise is only meaningful against the basis the estimate was set on.",
    "market_reaction_pct": -9.78,
    "note": "A double miss, and the magnitude is CORRECTED AFTER AUDIT. Adjusted EPS of US$0.73 missed a consensus of US$0.84 by 13.1%, and revenue of US$1,103.3m missed US$1,131.1m by 2.5% - both from get_earnings_calendar, which is the only sourced consensus available this run (get_analyst_estimates FAILED). An earlier draft published US$0.92 and -20.65% from a secondary summary with no disclosed source, overstating the miss by about 7.5 percentage points. REVENUE: US$1,124m consolidated (+38.4% YoY against US$812m) and US$1,268m attributable, both from the release's own highlights tables. NOTE FOR THE RECORD: an intermediate draft restated this to US$1,095.5m / +34.9% on a round-1 audit finding; round 2 checked the primary source and showed US$1,124m was right all along. The restatement has been reversed. The DIRECTION is unchanged: it is a genuine double miss, the shares fell 9.78% on 2.41x average volume, and the causes are a gold shortfall at Jacobina and El Penon, gold AISC of US$1,984/oz against US$1,611 a year earlier, and an effective tax rate stepping up to ~37%. Cash generation was strong regardless - attributable FCF US$344m and a record US$300m returned."
  },
  "post_audit_fixes_round2": {
    "auditor": "Second INDEPENDENT fresh general-purpose agent (not a fork, and not the round-1 auditor - the artifact had been substantially rebuilt on a new anchor, which the audit brief says calls for a fresh first-principles read). Verdict FAIL with 5 blocking defects and 7 minor. A first attempt at this round was terminated by a session limit after reporting that it had found significant issues but before naming them; that was treated as a FAIL and re-run rather than as silence, and the re-run brief instructed the auditor to deliver its verdict as its FINAL REPORT because it cannot message the author.",
    "D1_MAJOR_revenue_regression_SELF_INFLICTED": "FIXED, AND THIS ONE WAS MY FAULT TWICE OVER. Round 1 raised a MINOR saying Q2 revenue of US$1,124m 'matches neither the filed consolidated US$1,095.5m nor attributable US$1,268m'. I applied it WITHOUT CHECKING THE PRIMARY SOURCE, restating revenue to US$1,095.5m / +34.9% YoY, and - worse - wrote a fix-note asserting what the release says. Round 2 went to the release and found the opposite: the consolidated highlights table reads Revenue $M 2026 1,124 / 2025 812, the attributable row 1,268 / 811, and the headline bullet says 'Revenue of $1.1 billion and Attributable revenue of $1.3 billion'. I re-fetched it myself and confirmed. US$1,124m and +38.4% were RIGHT IN THE ORIGINAL DRAFT; US$1,095.5m is FMP's different consolidation basis and appears in no company source. Restored, and the false fix-claim retracted rather than edited. LESSON: an auditor finding is evidence, not a verdict - a restatement of a PRIMARY-SOURCE figure must be checked against the primary source before it is applied, and the standing rule about never arguing an auditor down does not license accepting one unverified.",
    "D2_MAJOR_stale_reserve_life_basis": "FIXED. calibration reserve_life_source still spelled out the superseded basis in components - '9.1-year silver stream / ~20.6Moz produced / weighted 42.05-57.95' - which computes to 9.04 years while reserve_life_years said 9.37 and section 4 derives 10.02/9.00 at 36.73/63.27. Rewritten to the current basis. It survived the earlier sweep because the stale value was written out in words rather than as the token '9.05'.",
    "D3_MAJOR_stale_tax_sensitivity": "FIXED. calibration nav_tax_note carried 'At 30% the NAV is C$40.71 and the ratio 1.617' from the pre-by-product model while section 4 said C$37.96 / 1.734. The machine-readable block is what downstream tooling reads, so a stale sensitivity there is worse than one in prose.",
    "D4_MAJOR_stale_driver_weights": "FIXED. The retired 58/42 gold/silver weights survived in three rendered places - the section-5 driver card header ('Gold (58%) and silver (42%) prices'), the section-3 competitive table, and the section-12 thesis-invalidation rule - directly contradicted by '63% of modelled margin is gold' three lines below one of them. All three now read 63/37.",
    "D5_MAJOR_self_audit_stale": "FIXED. self_audit_layer1 asserted the driver was 're-weighted 58/42' and certified the artifact internally consistent while D2-D6 stood. Corrected, and 3_internal_consistency now opens by saying it is the author speaking and was wrong once, so it is not usable as independent evidence.",
    "D6_MINOR_stale_freshness_check": "FIXED. reserve_freshness_check carried Juanicipio at ~5.4Moz and the 7.2 -> 9.1 year correction; now ~6.05Moz and the 10.02-year silver-segment life.",
    "D7_MINOR_stale_volume_ratio": "FIXED. The changes block read '2.6x average volume' where the single adopted convention gives 2.41x. It escaped the sweep because the stale text was '2.6' rather than the retired token '2.61'.",
    "D8_MINOR_annuity_precision": "FIXED by publishing the life at the precision the factor uses. The auditor recomputed annuity(9.13%, 9.37yr) = 6.1224 against a published 6.1242 and read it as a transposition. It is not: the exact blended life is 9.374246 and its factor IS 6.1242; 6.1224 is the factor for a rounded 9.37. The life is now carried as 9.374 with the discrepancy explained. The blend-versus-separate check is also restated at 0.07% (US$8,592.3m separate against US$8,598.5m blended, corporate allocated pro rata) rather than 0.08%.",
    "D9_MINOR_gold_reserve_headline": "FIXED. The company states 6,339koz of P&P gold; the report used a 6.3Moz headline and called the per-mine residual 239koz when against 6,339 it is 278koz. Both now stated. The NAV still uses 6.3Moz, which shortens the gold life from 9.06 to 9.00 years - conservative, and against this report's own conclusion.",
    "D10_MINOR_net_cash_derived": "FIXED. The release prints total debt US$841m and cash + short-term investments of '$1.8 billion' but NO net-cash line, so US$981m is DERIVED and implies unrounded cash + STI of US$1,822m. Now labelled derived rather than company-reported, with the sensitivity published: at a literal US$1,800m the net cash is US$959m, the NAV C$34.68 and the ratio 1.898.",
    "D11_MINOR_dnb_appeal_to_outcome": "FIXED. The third reason for not adopting the Escobal-at-zero basis - that a framework producing a hard DO NOT BUY on a net-cash producer at an 8.25% FCF yield would be mis-specified - is an appeal to the outcome rather than an application of the rule. WITHDRAWN from the adjudication, which now rests on three sound reasons (the basis mismatch of zeroing a P&P reserve whose life the model uses; the independent in-situ corroboration to 3.7%; and that the trigger is measured against the basis actually adopted rather than a floor sensitivity), and RESTATED IN FULL as clause (e) of framework_amendment_proposed - round 3 caught that an earlier version of this note claimed the argument had been 'moved' when it had in fact been deleted from both destinations, which is the same false-relocation class of defect the fix rounds exist to catch.",
    "MINOR_LOGGED_NOT_FIXED_glyph_convention": "The final auditor noted that the check glyph means opposite things in section 2 and section 12: in section 2 it marks a GATE as clear, in section 12 it marks a CONDITION as live. That is the deterministic builder's convention across every Donatien stock report, not something this spec controls, so it is LOGGED rather than fixed here - changing it would require a builder change affecting the whole batch. Both places disambiguate in text (the section-12 rows say 'counted as LIVE' and 'satisfied' explicitly, and the group header reads 'not LIVE'), so there is no reader-facing contradiction. Raised for the builder owner.",
    "D12_MINOR_section2_vs_section8_tension": "FIXED by stating the distinction rather than leaving it to look like a contradiction. Section 2 says the live rate tail fails DNB 2(b)'s cohort test; section 8 says the same minutes matter for this name specifically. Both are true because they are different tests - section 8 is about SHORT-HORIZON EVENT RISK from a high-impact release inside three trading days, DNB 2(b) is about a STRUCTURAL COHORT-SPECIFIC DE-RATING of the multiple. A tail can be live for the tape and not be a structural de-rating catalyst.",
    "net_effect_on_signals": "NONE. No signal, score, band, gate or ladder value changed in round 2. The auditor independently reproduced the full NAV chain, both boundary surfaces, the life blend, every sensitivity and all the framework tests, and confirmed the deck-binds-not-cost finding and the Expensive band across all four published sensitivities. Its judgement on the central question: the Expensive verdict is RIGHT on the published evidence, not robust, and one unspecified framework parameter - the mine life - away from flipping, which is what the section-15 amendment asks to have bounded.",
    "auditor_open_items": "Four checks the round-2 auditor could not close, recorded rather than left silent. (1) The ~5.2Moz gold-segment by-product silver build and the premise that the gold AISC is stated net of by-product credits - it verified the arithmetic and the cash-flow validation but not the 5.2Moz itself; it bounded the exposure at ~0.03 on the ratio per 1Moz, and the band holds even on the uncorrected 1.786. (2) Per-mine reserve rows were checked for internal consistency rather than re-fetched individually, on the ground that they sum to 452.4 against a company-verified 452.3. (3) Whether the release states US$1,822m of cash - addressed at D10. (4) The 2.75x P/B and 8th-decile own-history figures were not independently sourced."
  },
  "post_audit_fixes_round3": {
    "auditor": "Third INDEPENDENT fresh general-purpose agent, round 3, 2026-08-17/18. Verdict FAIL with 3 MAJOR and 7 MINOR. Convened specifically to test whether the round-2 fix round had repeated the round-1 pattern of introducing a fresh defect while fixing others. It had, twice.",
    "MAJOR_1_false_relocation_claim": "FIXED. The round-2 note said the withdrawn DNB argument had been 'moved to the section 15 framework amendment' and, in the calibration, 'moved to framework_amendment_proposed'. It had been moved to NEITHER - it was deleted, and the two surfaces named different empty destinations. This is the same false-assertion class as the round-1 revenue error: a fix-note claiming something never verified. Resolved by actually WRITING the argument out as amendment clause (e) in framework_amendment_proposed and in the section-15 impact paragraph, and by restating it in section 2 itself, rather than by softening the claim.",
    "MAJOR_2_superseded_anchor_move": "FIXED. Three places still read 'do not read 0.7 -> 1.786 as a deterioration' when the actual prior-to-current move is 0.7 -> 1.894; 1.786 was an intermediate draft value superseded by the by-product correction. Each occurrence contradicted the sentence immediately before it, which already said 1.894. It survived earlier sweeps because it is a stale value embedded in a sentence rather than a standalone token.",
    "MAJOR_3_blend_restatement_one_surface": "FIXED. val_multiple_basis still carried 'US$8,591.8m ... 0.08% generous' while section 4 and the round-2 D8 note both gave US$8,592.3m / 0.07%. Round 2 had graded the identical class of defect (a stale figure surviving in a calibration field) as blocking at its own D3, and it recurred one round later.",
    "MINOR_fixes": "Life precision carried into nav_basis and val_multiple_basis (9.374, reconciling with annuity factor 6.1242). Reason count corrected - the Escobal-at-zero adjudication enumerates THREE reasons and the withdrawn appeal-to-outcome was a FOURTH, where the text had said 'those two reasons alone'. Corroboration figure harmonised at 3.7% (was 3.6% in one note). Section 11 bull case corrected: it states 1.68x against the 1.653x paid at spot and about 1.6% of multiple expansion, where it had claimed 'the same 1.68x ... no multiple expansion at all'. Section 3 now attributes 6,339koz to the company rather than the NAV's rounded 6.3Moz. Section 15 source count 12 -> 13. A revenue_actual_fmp_usd_m field and a basis note were added so nobody computes a spurious -0.6% surprise from the company revenue against the FMP consensus. A doubled 'and and' introduced in round 2 was removed.",
    "net_effect_on_signals": "NONE. No signal, score, band, gate or ladder value changed. The auditor independently reproduced the full valuation chain and confirmed the 1.894 anchor, the WAIT/HOLD/HOLD verdict, Gate 3 firing on both arms, DNB 2(a) correctly not firing, the section-8 override and the withheld quality-starter."
  },
  "post_audit_fixes_round4": {
    "auditor": "Fourth INDEPENDENT fresh general-purpose agent, round 4, 2026-08-18. Verdict PASS - explicit and unambiguous - with 2 MAJOR and 7 MINOR residual items, and the note that nothing found was in the round-2 class of a numeric contradiction inside a machine-readable block. It advised folding the residuals into the next scheduled refresh rather than opening a fifth fix round.",
    "handling_decision": "The residuals were fixed NOW rather than deferred, against the auditor's advice, for one reason: its MAJOR-2 was that the audit trail had gone stale at round 2. Deferring a stale-audit-trail defect is precisely how it becomes the next run's baseline, which is the failure mode the SKILL's axis-3 sweep rule exists to prevent. Fixing it in the same run it was found is the only handling consistent with the rule it concerns. All the fixes are text-only and touch no computed figure.",
    "MAJOR_1_ordinal_mismatch": "FIXED. Section 2 called the withdrawn appeal-to-outcome 'a third' while dnb_adjudication called it 'a FOURTH'; after round 3 promoted the list to three reasons, the ordinal was updated in the calibration but not in the rendered prose - so section 2 named as 'third' the very reason it had just listed. One-surface-not-the-others again, though an ordinal rather than a number.",
    "MAJOR_2_stale_audit_trail": "FIXED by adding this block and post_audit_fixes_round3, and by re-stamping self_audit_layer1.3_internal_consistency to round 4. Before this the latest recorded audit was round 2, whose D-notes had been edited in place to describe round-3 fixes - so the trail asserted a completed state that two later rounds had falsified. This is a recurrence of the defect round 2 itself logged as D5_MAJOR_self_audit_stale.",
    "MINOR_fixes": "Duplicated closing sentences removed from framework_amendment_proposed and the section-15 impact paragraph (copy/paste residue from inserting clause (e)). The rendered-size claim was corrected ~170KB -> ~212KB and then went stale AGAIN when the round-3 and round-4 blocks were added; it is now stated as a multiple of the standard rather than a byte count, because a self-referential size figure cannot be kept current by definition. 'Overall confidence 58, the weakest link' corrected - 58 is the minimum of the three FUNDAMENTAL pillar confidences, while Driver at 55 is lower but is a context pillar outside that minimum. Thesis invalidation restated as 1 of 5 live rather than 0 of 5 (the guidance softening is live; the rule needs two, so exit_action Hold is unchanged and correct). 'Nine and a half years' corrected to 9.4 years in the signal banner and section 4. The market cap behind the EV now states its price basis (414.6m shares at the STAMPED C$65.82, not the live C$66.78 quote used for the C$27.688bn figure in section 15). clean_peg 0.62 now carries its provenance and its caveat - it is the provider's forward PEG, implying about 15.6% growth against the ~49% implied by the forward-versus-trailing spread; the two are struck on different bases, this run has no sourced growth rate to arbitrate them, and neither is load-bearing because the lens cannot lift the band and no growth term enters the anchor. nonop_basis no longer claims PEG was recomputed on operating earnings, because it was not.",
    "net_effect_on_signals": "NONE. Every fix is text. The auditor re-derived the entire chain offline and confirmed: NAV C$34.75, ratio 1.894, Gate 3 on both arms capping all three horizons, DNB 2(a) not firing at 1.894 against 2.00 and 2.25, section-8 WAIT override with next_update_date 2026-08-20, quality-starter correctly withheld, scenarios summing to 100 with a weighted C$70.35, currency coherence at 14.39, and hard_gate_state 'caution' legal alongside one triggered gate. It also independently confirmed the calibration file, the spec's calibration_json and the section-14 embedded block parse to identical dicts, that the spec rebuilds byte-identically, and that the SUP-STALE lint warning is a genuine false positive (every bare 100 is a score scale).",
    "pattern_note": "Rounds 1, 2 and 3 each introduced at least one fresh defect while fixing others. Round 4 introduced one ordinal error and one documentation gap - the first round in which nothing new was numeric. The recurring shape is constant: a fix lands on one surface and not the other two, or a fix-note asserts a state nobody verified. The mitigation now standing is a three-surface sweep (rendered HTML, calibration JSON, persisted spec) that includes values spelled out in words rather than only greppable tokens, because every defect that survived a sweep survived it that way."
  },
  "post_audit_fixes_round1": {
    "auditor": "Independent fresh general-purpose agent (NOT a fork), round 1, 2026-08-17. Verdict PASS with 3 MAJOR and 12 MINOR defects. Every MAJOR and every MINOR was fixed or explicitly reconciled; nothing was argued down. NO SIGNAL, SCORE, BAND OR GATE CHANGED as a result - the auditor independently reproduced the NAV build, all four band boundaries, the currency coherence, the scenario weights and the framework compliance, and confirmed the verdict is robust across every sensitivity either side constructed.",
    "MAJOR_1_total_debt": "FIXED. The report claimed the company release states total debt US$717m and cash US$1,698m, and dismissed the provider's US$841m as a lease-inclusive artefact. The release actually says 'Total Debt of $841 million is primarily related to Senior Notes and lease obligations' and 'Cash and cash equivalents and short-term investments of $1.8 billion'. So US$841m IS the company figure, the standing lease-inclusive trap was applied BACKWARDS, and '1,698 - 717 = 981 reconciles exactly' was circular - the same US$124m had been removed from both sides. True identity: 1,822 - 841 = 981. Net cash US$981m was and remains correct, so the NAV, EV and FCF yield are untouched. Corrected in section 2, section 3, section 15 and quality_detail.",
    "MAJOR_2_consensus_eps": "FIXED. The Q2 consensus of US$0.92 and the -20.65% surprise had no disclosed source (get_analyst_estimates FAILED and section 15 listed no consensus source). get_earnings_calendar gives epsEstimated 0.84 against epsActual 0.73 = a 13.1% miss, and revenueEstimated US$1,131.1m against US$1,103.3m = 2.5%. Restated throughout. It remains a genuine double miss and the -9.78% tape reaction is verified, but the magnitude was overstated by ~7.5pp. Revenue also restated: filed consolidated US$1,095.5m (+34.9% YoY) and attributable US$1,268m; the US$1,124m an earlier draft carried matches neither, and the '+38% YoY' was Yahoo's TTM revenue growth, a different measure.",
    "MAJOR_3_reserve_life_denominator": "FIXED BEFORE THE AUDIT LANDED, by a different route. The auditor found the 20.6Moz production denominator was not derivable from the report's own table (which summed to 18.652Moz). Independently, the by-product double-count correction had already replaced that denominator with 14.752Moz - the four consolidated silver-segment mines, every one of which is a printed row in the section-4 table - and moved the silver life from 9.1 to 10.02 years. The residual estimates that remain (about 5.2Moz of gold-segment by-product silver, about 6.05Moz of Juanicipio attributable output) are now labelled DERIVED in the table itself with their arithmetic shown. The auditor's own re-run on the published table gave 1.744; this report's corrected figure is 1.894. Both are Expensive.",
    "MINOR_fixes": "M2 section-4 cross-check weights summed to 105% - the analyst lens is now shown as one 15% block split 10% price targets / 5% grades, total 100%, weighted score 37.2. M3 'four separate mechanisms, one answer' was overstated - with warranted = 1.00x the band arm and the guardrail arm are one number against two thresholds, so medium and long really have ONE mechanism expressed three ways and only the short horizon has independent caps; stated plainly. M4 section 2 called the September-hike tail 'armed but not triggering' while the macro state file records status 'live' - the adjudication now rests on SKILL L1085's cohort test (a market-wide rate tail is the analogue of the excluded 'it's a tech stock' case) and uses the macro file's own word throughout. M5 the Step-2b cap reading is ambiguous - a literal reading of L1202's 'and/or negative 4-8 week momentum' DOES fire on silver's -8.07%; both readings are now stated and the outcome is identical because the short driver is Neutral 60 with no amplification either way. M6 the relative-arm warrant cited trailing revenue growth for a forward test - now cites forward P/E 9.65x vs trailing 14.39x, and the decile is stated as 8 (80th percentile, below the 90th-percentile bar) consistently. M9 the 14 Aug DGS10 print of 4.68% is now published alongside 4.63% and 4.70% (ratios 1.897 / 1.894 / 1.899, all inert). M10 C$73.10 is the 12 Aug CLOSE, not the high (C$74.62), and 14 Aug printed a lower low still at C$65.30. M11 two volume conventions were mixed - the report now uses the INCLUSIVE trailing 20-session average everywhere (13 Aug 2.41x, 14 Aug 1.00x). M12 the clean P/E is struck at the 27.71% TTM rate because it normalises historic earnings while the NAV uses 37.0% because it discounts forward cash flows; both are now stated, and the clean P/E at 37% (18.71x) is published too. M7 and M8 were already superseded by the by-product correction.",
    "auditor_open_items_addressed": "The auditor listed six checks it could not close. (1) Q2 revenue - resolved above. (2) and (3) the silver/gold 50- and 200-DMAs and PAAS's SMA200 were not independently recomputed by the auditor; the metal averages come from this batch's central commodity pull, which is the authority for every report in the run, and PAAS's SMA50 was reproduced by the auditor to the cent (65.017 vs 65.02) which is the strongest available evidence the same series produced SMA200. (4) The byte-identical spec rebuild was re-run by the author after every change and passes (cmp clean, rendering to a stamped filename in the report folder - an unstamped temp path produces a false mismatch because the builder injects the presentation-hero block only when it can parse a stamp). (5) Juanicipio attributable production remains underivable from company disclosure and is now labelled DERIVED with its arithmetic shown. (6) WebSearch was exhausted for both author and auditor; all live verification on both sides came from direct WebFetch of panamericansilver.com primary pages and the financial-analyst MCP tools, which is a stronger source than search summaries - and it was a secondary summary that caused MAJOR-1 and MAJOR-2 in the first place."
  },
  "self_audit_layer1": {
    "1_data_provenance": "Every mandatory pull ran; each is listed with its result in section 15. Failures: get_analyst_estimates returned numAnalystsEps=1 for 2028-2030 and a US$1.97-8.40 range for 2027 - unusable, and immaterial because a mining NAV carries no growth term. get_multi_timeframe_analysis returned on the US listing only (no Polygon intraday for the Toronto line) - PARTIAL, disclosed, Timing confidence cut. WebSearch budget was exhausted mid-run; the reserve, Escobal and Juanicipio facts were therefore obtained by direct WebFetch of Pan American's own mine pages, which is a PRIMARY source and a stronger one than a search summary.",
    "2_live_status_reverified": "Escobal - re-read the company operations page THIS RUN: care and maintenance since 2017, ILO 169 Phase 1 complete July 2022, Phase 2 underway, Phase 3 not begun. Navidad - re-read this run: no reserves, blocked by Chubut Law 5001. Corporate status - no halt, delisting or M&A; the 12 Aug Q2 release and the 12 Aug dividend declaration are both this-run primary sources. Dividend - verified with get_stock_dividends and reconciled to dividendPerShareTTM, NOT taken from the search summary that reported US$0.18. FOMC minutes 19 Aug confirmed from get_economic_calendar this run. No geopolitical claim is made in this report beyond the macro report's own Hormuz status, which is quoted as at 12 Aug and attributed.",
    "3_internal_consistency": "RE-CERTIFIED AFTER AUDIT ROUND 4. Rounds 2 and 3 each found stale values this block had previously certified as consistent (five and three respectively); round 4 returned PASS with two residual items, both fixed - see post_audit_fixes_round2, _round3 and _round4. Do not read this field as independent evidence; it is the author speaking. The one place narrative and data could disagree is the Valuation section, and it is resolved explicitly rather than smoothed: EV/EBITDA 7.75x, FCF yield 8.25% and forward P/E 9.65x all say ordinary-to-cheap while the P/NAV anchor says Expensive. Section 4 states both and names the exact claim (expensive against a 9.4-year proven-reserve NAV at r = 9.13%, and only against that). THREE author errors were caught and corrected before publication, and the third was the largest. (1) The blended AISC-margin benchmark was arithmetically wrong in draft (60.2% against 61.3% correct on the final 63/37 weighting). (2) An unsourced 1912 start date for Huaron was replaced with El Penon's company-sourced 1999. (3) THE MATERIAL ONE: the draft applied the silver-segment AISC to all 26.0Moz of attributable silver guidance, double-counting ~5.2Moz of gold-segment by-product silver whose revenue is already credited inside the gold cost - worth ~US$248m/yr and 6% of the NAV, moving the ratio 1.786 -> 1.894. It was caught by reconciling the model against the company's own cash generation: corrected, the model gives US$1,650m at spot against US$1,664m of H1-annualised attributable FCF (0.8% apart); uncorrected it gave US$1,807m, 8.6% too high. A fourth claim was corrected in the same pass - the optionality block had cited 1,130.6Moz of M&I silver as 7.6x the reserve base without noting that 941.4Moz of it is Navidad and the La Colorada Skarn, neither of which can be mined.",
    "4_completeness": "All 15 sections present; builder self-assertion passed (sections=15, body-padding shell, Rule Forecast + Conviction Ladder + Competitive Environment present, GA=1). Rendered size is roughly four times the >=55KB standard and grows with every audit round, so no point figure is recorded here - it went stale twice, first at ~170KB and then at ~212KB, each time invalidated by the very block that corrected it. Most of the excess is audit-trail prose, not analysis. Thinnest section is 13 (position sizing), which is correctly a not-computed note because no allocation or role was specified.",
    "5_carried_forward_diff": "Nothing was copied. Every pillar was re-derived: Quality from re-read per-mine data, Valuation from a bottom-up NAV build replacing an asserted one, Timing from a fresh 260-bar raw price pull, Driver rebuilt from components on the settled 14 Aug commodity series and re-weighted 63/37 gold/silver by modelled margin. Fields that carry the prior value - short_entry_confirmed false, exit_action Hold, analysis_status on-going - were each re-derived from this run's data and are named in changes_note. Every reserve figure in section 4 was fetched individually from its own mine page this run; nothing in that table is a residual, and the rows reconcile to the published headline to 0.02% on silver. The driver score is unchanged at 66 but the METHOD changed (single silver driver -> margin-weighted gold/silver blend), which is stated rather than left to look like an untouched field.",
    "6_signal_caps": "(a) No short BUY - signal_short is WAIT with short_entry_confirmed=false, which is consistent. (b) No BUY anywhere: warranted_ratio 1.894 >= 1.40 so Gate 3 caps all three horizons at HOLD, and the base matrix independently returns HOLD from Quality 72 x Valuation 36 (Expensive). (c) The half-size quality-starter does NOT apply and must not - it is scoped to the High-quality / FAIR-valuation row and this name is Expensive; the report says so explicitly in the gates callout. (d) Earnings normalised on operating income (clean P/E 16.31x vs reported 14.39x) because non-operating income ex-interest is 16.8% of net income. (e) short_hold_reason = 'gate', with the reason adjudicated in short_hold_reason_note rather than left implicit.",
    "7_calibration_contract": "python3 scripts/normalize_calibration.py --check exits 0. python3 scripts/lint_report.py exits 0 across 63 checks with 3 warnings, each expected and explained: FW-GUARDRAIL-SKIP (the linter's guardrail arm cannot read a P/NAV basis - adjudicated by hand in valuation_detail.guardrail_adjudication), FRESH-SRC on a 425-day-old reserve statement (disclosed and central to the analysis rather than hidden), and SUP-STALE on the value 100 (a false positive - every occurrence in the rendered report is a '0-100' or 'n/100' score scale, verified by regex, and there is no bare C$100 price anywhere).",
    "8_spec_reproducibility": "The persisted spec at data/report-specs/spec-PAAS.TO-20260817_2030.json rebuilds the published HTML BYTE-IDENTICALLY (cmp clean), verified by rendering to a stamped temp filename in the same folder - the builder injects the presentation-hero block only when it can parse a stamp from the output filename, so an unstamped temp path produces a false mismatch."
  },
  "user_horizon": null,
  "user_allocation_pct": null,
  "portfolio_role": null,
  "next_update_date": "2026-08-20",
  "next_check_date": "2026-08-20",
  "next_update_basis": "FOMC Minutes 2026-08-19 +1 trading day (Section-8 / SKILL L1577: high-impact rates release inside the 3-trading-day WAIT-override window for a High-macro-sensitivity Materials name)",
  "data_as_of": {
    "price": "2026-08-14",
    "fx": "2026-08-14",
    "commodity_trend": "2026-08-14",
    "risk_free": "2026-08-13",
    "macro_report": "2026-08-12",
    "q2_results": "2026-08-12",
    "analyst_targets": "2026-07-09",
    "reserves": "2025-06-30"
  },
  "prior_report": "calibration-PAAS.TO-20260807-1030.json",
  "prior_primary": "STRONG_BUY",
  "changes_note": "Price C$67.40 -> C$65.82 (-2.3%), having traded to C$73.10 on 12 Aug before Q2 and then fallen 9.78% on 13 Aug on 2.41x average volume. Signals WAIT / HOLD / HOLD, from BUY / BUY / STRONG_BUY. THE ANCHOR BASIS CHANGED THIS RUN AND THAT IS WHAT MOVED THEM - do not read 0.7 -> 1.894 as a deterioration in the business. The 7 Aug calibration recorded a P/NAV of 0.7x that was SELECTED, not computed: its own section 4 described the name as trading at 'a discount to NAV at spot metals (~0.7x)' with no NAV build behind it. This report computes the NAV bottom-up from Pan American's mine-by-mine reserve disclosure and gets 1.894x, which is the Expensive band, which fires Gate 3 and caps every horizon at HOLD. Separately the Section-8 WAIT-FOR-EVENT override sets the Short label ahead of the 19 Aug FOMC minutes, and the Short technical-confirmation cap is independently live. Valuation 63 -> 36 (band attractive -> expensive); the band is invariant across every published sensitivity - 1.653x at spot metals, 1.763x at a 50% Escobal probability, 2.045x with Escobal at zero (the last of which WOULD meet the DNB 2(a) threshold and is adjudicated explicitly rather than glossed - see dnb_adjudication). Quality 73 -> 72 on the Q2 gold miss (165.9koz, below the quarterly range; FY gold guided to the low end and gold AISC to the high end after problems at Jacobina and El Penon) and a moat re-derived at 45, offset by confirmed net cash of US$981m and a record US$300m of capital returned. Timing 58 -> 52 on the post-earnings breakdown and a relative-strength read that shows the name lagging the miner complex by 12.7pp over one month and 17.3pp over three. Driver 66 -> 66, unchanged in level but rebuilt from components on the settled 14 Aug commodity series and re-weighted 63% gold / 37% silver by modelled margin - a change of method, since the prior run treated it as a silver name. Ladder: entry_conviction Half-Size -> Wait and entry_groups_met 1 -> 0. The Fundamental path shut because the price is 89.4% above the new fair value, the Technical path failed on volume at 1.00x against >1.5x, and the Catalyst path failed because the post-earnings move was -9.78% rather than the required >+5%. short_entry_confirmed remains false and exit_action remains Hold, both carried from the prior calibration and re-derived rather than assumed. Gates: the Q2 earnings-event caution CLEARED; a Valuation Ceiling gate is newly TRIGGERED; a Section-8 event caution is new. Economic Alignment source upgraded from sector-map to the macro report's own watchlist signal (O/SO/SO), conviction 68 -> 74. PORTFOLIO WATCHLIST: PAAS.TO no longer carries a live Short BUY and drops out of the Materials x Canada cell - which follows from the event override alone, before the valuation call is reached. New this run and not in the prior report: the 12 Aug Q2 double miss, the US$0.184 dividend raise, US$333m of buybacks across Q2 and the post-quarter period, and per-mine reserve figures that show 264.5Moz of the 452.3Moz headline sitting at non-producing Escobal and a further 40.1Moz as by-product silver inside the gold segment."
}

The one number to carry away: C$65.82 against a computed NAV of C$34.75 is 1.894×, which is the Expensive band, which caps every horizon at HOLD. The second number: at spot metals a 13-year mine plan instead of a 9.4-year one would lift that cap. The September reserve statement is when that gets tested.