NYSE:ICE Intercontinental Exchange, Inc.

Lifecycle & sector. ICE is a mature, capital-light financial — a quality compounder, not a deposit-taking bank. It is scored on the exchange and data lens (forward P/E, EV/EBITDA, FCF yield, recurring-revenue mix, operating margin, ROE) rather than on price-to-tangible-book. Tangible book is negative, about −$28 a share, because two decades of acquisitions — Interactive Data, Ellie Mae, Black Knight — sit on the balance sheet as $30.6bn of goodwill and $14.9bn of other intangibles. For a roll-up that is an artefact of the model, not a solvency signal. Net revenue compounds in the mid single digits with a recurring core above half, which puts it in the mature band.

Q2 2026, reported 30 July — the freshest print, and a sharp deceleration. Net revenue $2,666m, up 5% year on year. GAAP diluted EPS $1.69, up 14%; adjusted diluted EPS $1.90, up 5%. Operating margin 52%, adjusted 61%. Recurring revenue $1,353m, up 8%; transaction revenue net $1,313m, up only 2%.

The number that matters most is not in the quarter, it is in the comparison. First-half net revenue was $5,643m against $5,016m, up 12.5% — which means Q1 was up 20.4% and Q2 was up 5%. The first quarter of 2026 was a volatility windfall: the Strait of Hormuz closed in late February, oil ran to $100 in March, and everything that clears through ICE traded. That windfall is now unwinding, and it is the single most important fact about the driver score in §5. A reader who annualises the first half of this year is annualising a spike.

Why the segment split matters more than a blended read. These are three businesses with different cyclicality pointing in different directions, and a single blended multiple hides all of it. Exchanges is a volatility call option with a 75% adjusted margin. Fixed Income & Data Services is a subscription business growing 8–10% on recurring revenue with a 46% margin. Mortgage Technology is a rate-cycle bet whose GAAP margin of 8% flatters to deceive in the wrong direction — on an adjusted basis it earns 43%, and the difference is amortisation of an acquisition that closed three years ago. Nothing in this report applies one multiple to the three of them without saying so.

Moat average 79/100 — genuinely wide, one point narrower than last month. The switching-cost and pricing-power marks are set by the competitive read below, not asserted.

Net effect on the moat. Switching costs trimmed 80 → 78 and pricing power 80 → 78, both on the data-substitution vector rather than on any measured loss of customers — the tape says that line grew 9%. Cost advantage held at 68. Overall competitive threat level: moderate; share trajectory: stable — gaining in energy and ags benchmarks and in cash equities, holding in data, losing in electronic fixed-income execution, which is the one lane the acquisition addresses. The AI vector propagates to the Bear case in §11 and to the thesis-invalidation rule in §12; it does not reach Do-Not-Buy Trigger 5, because a regulated clearing house is not disrupted by a language model.

Weighted blend 69. Add a +3 embedded-optionality tilt, taken at the bottom of the sanctioned +3 to +8 range because the $6.0bn cash deal consumes the balance-sheet flexibility that would otherwise fund those options. Valuation score 72, inside the attractive band's 65–77 range.

Implied growth. Reverse the anchor and $154.31 embeds about 6.0% of five-year earnings growth against a disciplined estimate of 7.5% and a consensus of 10.35%. The price is asking for meaningfully less than the fundamentals support. That is the whole valuation case, stated in one number, and it is a modest case — not a mispricing, a small discount.

Risk-reward composite 50, up from 32 last month — the improvement is real (RSI 74 → 65.2, OBV accumulating, ATR contracting) and it is about the indicators, not the location.

Composition. Multi-timeframe trend 66 (×0.30) + risk-reward 50 (×0.20) + macro 60 (×0.20) + sentiment 58 (×0.15) + catalysts 40 (×0.15) = 56.5, recorded as 56 (rounded down, the conservative direction), which lands in the Improving band (55 and above) by a point and a half, up from 49. Timing confidence 65%, after a 10-point deduction for a high-impact release inside seven days on a High macro-sensitivity name.

Upcoming events (next 30 days)

Recent surprises (last 7 days)

The WAIT-for-event override: engaged, and non-binding. Financials is a High macro-sensitivity sector, and the FOMC minutes on 19 August are a High-impact release two trading days out — inside the three-day window, so the §8 override fires on its face. It matters less than it looks, for a mechanical reason: the override can only cap a short signal, and the technical-confirmation cap in §12 has already produced a non-BUY short. Both routes arrive at the same instruction — do not open a short-horizon position at today's price — so the report uses HOLD rather than WAIT, because HOLD carries the actionable trigger with it (“buy on confirmation”) and WAIT does not. The override is also why the next update is scheduled for 20 August rather than the default fourteen days out.

Ruling on Tail 4 — September hike, ~44% implied, trigger is these minutes. This is the one macro tail that reaches ICE through two channels at once, and they point opposite ways, so a generic “rate-sensitive financial” framing is wrong in both directions. Take the hawkish branch first. Rate volatility drives interest-rate futures and options volume: the Financials line inside Exchanges grew 21% in Q2, the fastest of any transaction line. It also drives collateral: clearing margin deposits are $114.6bn against $76.8bn at December, and ICE earns interest on the balance, which is why “OTC and other” grew 15%. Against that, Mortgage Technology loses. Thirty-year mortgage rates are already 6.77%; a hike pushes originations further down, and origination technology is $197m a quarter.

Now size the two legs honestly, because the tempting comparison is the wrong one. It would be easy to set the whole Exchanges segment ($1,464m a quarter) against the whole of Mortgage Technology ($557m) and call it a three-to-one tilt. That is not the right comparison: most of Exchanges is not rate-driven at all — Energy $518m, Ags and Metals $87m, cash equities $140m, data and connectivity $287m and listings $129m are levered to other things entirely. The genuinely rate-sensitive lines are Financials at $192m plus the collateral-interest share of OTC and other at $111m — call it $200–300m a quarter of upside exposure. Against that sits origination technology at $197m, which is the genuinely rate-cyclical downside, with servicing software at $226m running the other way, because high rates keep loans on servicers' books for longer.

So the like-for-like read is roughly $200–300m of upside exposure against $197m of downside, with $226m countercyclical behind it — balanced on the transaction lines and tilted mildly positive once servicing is counted. That is a much smaller edge than the segment headline suggests, and it should be. It is enough to say a hawkish print is net neutral-to-mildly-positive for ICE's revenue and negative for its share price on the day, because the market trades ICE as a financial rather than as a volatility long. It is not enough to call ICE a rate-hike beneficiary. And the hawkish leg carries the same caveat the driver section applies to the whole name: energy revenue fell 13% with the strait already shut, so volatility does not reliably convert into volume here either. The conclusion that survives is the modest one — ICE is not a one-directional bet on the minutes, which is why they are not a reason to sell, and why the unresolved event is a reason not to open the short horizon at $154 two sessions beforehand.

What the last week actually said, and it cuts against the hike. The two largest High-impact surprises in the window were both dovish misses — PPI flat against +0.2% expected on 13 August, and retail sales contracting 0.6% against +0.1% expected on 14 August — alongside an in-line CPI whose core rate is disinflating, 2.6% to 2.5%. The hawkish evidence in the same window is thinner: a strong Empire State print, which is Medium impact, and rising one-year inflation expectations. That matters for how to read the minutes: the ~44% implied odds were set before this run of data, and the data since has leaned the other way. It does not change the ruling set out above — the point of that ruling is that ICE is not a one-directional bet on the outcome — but a report that listed only the hawkish prints would be telling the reader the opposite of what the week contained.

The other three armed tails, named individually so nobody has to infer them. S&P concentration / AI earnings-quality unwind — armed but receding, and not inherited here (§2). Private-credit crack — building, not armed, so no inheritance is available; ICE has no direct private-credit exposure in any case. Hormuz closure escalation — live, with Brent at 88.52 against a $100 trigger, so live-but-not-triggering. For ICE that tail is genuinely two-sided and this quarter it was negative: energy net revenue fell 13% with the strait already shut.

The weighted read is 66 — mostly bullish — but the composition matters more than the number. The monthly chart is in an uptrend and price sits just under its own 20-month average of $163.16, so the multi-year structure survived this year's drawdown intact. The weekly is the awkward one: price is above the 20-week average of $147.57 and below the 50-week at $155.67, which is a weakening configuration rather than the outright downtrend the data provider's label suggests — price is above both weekly exponential averages, so the label is not consistent with its own inputs and this report scores the configuration rather than the label. The daily is where the decision lives: above a 50-day at $141.20 by 9.3%, and fourteen cents under a 200-day at $154.45.

The pattern this makes is the textbook one for the framework's higher-timeframe bullish, lower-timeframe extended case — except run in reverse. Normally you get an uptrend with a pullback to buy. Here you have a recovery with no pullback to buy, seven weeks and 25% off the low, arriving at resistance. The intraday charts are all making higher highs on volume that is almost nil (0.65× the 20-day average on the daily, and thinner still intraday), which is what a grind into a level looks like rather than a break through one. Volume is the missing ingredient, and it is the ingredient the entry rule in §12 is waiting for.

The weekly support ladder is $159.21 / $152.38 / $144.18 / $143.17 / $121.79, so the nearest weekly level below spot is $152.38 — 1.3% away, and the level the §12 entry rule is waiting on.

* The tool tags the weekly bar support_breakdown; on the settled data price is above both weekly exponential averages, so the label is carried here for transparency rather than relied on. † The daily support shown is the 6 August low, derived by hand — the tool's own daily support list runs $137.76 and below, which is too far from spot to be useful. Both are listed in §15.

Levels that matter: $154.45 is the 200-day and the immediate decision point; $158.55 is the 30 July swing high and the first real resistance above it; $147.47 is the 6 August low and the top of the support shelf; $141.20 is the 50-day and the level a failure here would test.

Six months of raw daily closes to 17 August 2026 (126 sessions, unadjusted, so the moving-average comparisons are not distorted by the dividend), with the 50-day average in orange. The shape is the story: $175 in late January, $121.79 intraday on 29 June — a 30% drawdown on no deterioration in ICE's own numbers — then a 25% recovery that has run straight into the 200-day average at $154.45.

Probability-weighted 12-month value: $175.30 — 0.25 × $210 + 0.55 × $176 + 0.20 × $130, or +13.6% against the 17 August close. The weights sum to 100 and the central case is the most probable, as the framework requires. The distribution is wider than it looks for a 0.93-beta infrastructure business: the gap between the two tails is 51.8% of the share price — the bull sits 61.5% above the bear — and almost all of that width is multiple rather than earnings.

Forecast:

Technical group, RSI leg. RSI (14) needs to come back under 65 from 65.2. It has fallen from 74 on 31 July at roughly 2 points a week on a flat-to-slightly-higher tape, so a single quiet session or a shallow pullback does it. Forecast: within 1–5 trading days. Confidence: High. The trap is that RSI can also fall by the stock dropping, which is not the entry you want — the condition to watch is RSI under 65 with price holding the $152.38 weekly shelf named below.

Technical group, volume-or-support leg — the harder of the two. It has two branches and both are currently shut. The breakout branch needs 6.08m shares against today's 2.64m; volume rules cannot be time-projected because they are event-driven, and nothing between here and the 29 October print reliably produces that except a Fed surprise. Forecast: catalyst-dependent. Confidence: Low. The pullback branch is the realistic one: it needs a tested bounce off weekly support at $152.38 — 1.3% below spot — with a confirmed higher low. A shallow dip that holds that level satisfies both this leg and the RSI leg at the same time. Forecast: 1–3 weeks, and the 19 August minutes are the most likely cause. Confidence: Moderate.

Fundamental group, price-below-fair-value leg. Met today with $9.69 of headroom. It stops being met above $164, which the base case reaches within twelve months — so this path closes as the thesis works, which is the normal way round. Forecast: remains met for 3–6 months on the current trajectory. Confidence: Moderate.

Catalyst group. Cannot trigger before 29 October 2026. Both failing legs are earnings-print conditions and there is no print in between. Forecast: 29 October, or not at all. Confidence: Moderate — ICE has beaten consensus in recent quarters, but a +5% single-day move on a 0.93-beta infrastructure name is a high bar it has not cleared in either of the last two prints.

Net. Both failing Technical legs resolve on the same event, which is what makes this a short wait rather than an open-ended one: a shallow pullback into the $152.38 weekly shelf that holds would put RSI back under 65 and give the tested higher low at once, flipping the group to met, the ladder to Full-Size and the short signal to BUY. The 19 August minutes are the most likely trigger, and they are the reason this report re-runs on 20 August rather than in a fortnight. The alternative — a high-volume break above $158.55 — would also do it, but on today's 0.65× volume that is the less probable path.

Forecast:

Stop-loss at $146.50. Forecast: unlikely inside 4–6 weeks absent a catalyst. Confidence: Moderate. It is 5.1% below spot and 2.2 ATR away, which on a 2.3% daily range is roughly two bad sessions — not remote. The realistic trigger is a hawkish surprise in the 19 August minutes taking the stock back through the $147–150 shelf, which is exactly the shelf it bounced off ten days ago.

Profit-target trim at $180.00. Forecast: 6–12 months on the base path. Confidence: Moderate. It needs +16.6% and an RSI above 70 at the same time, so it is a momentum-overshoot condition rather than a price condition. On the base case the stock gets there without the RSI, in which case do nothing.

Thesis invalidation. Forecast: no leg live, and none within one quarter on current trajectory. Confidence: Moderate. The leg to watch is not the guidance one, it is the competitive one — the Fixed Income Data and Analytics growth rate, reported quarterly, is the number that tells you whether the AI de-rating is a mood or a mechanism. It was +9% in the second quarter. Two prints below +5% and the case changes shape.

No portfolio allocation or role was specified for this analysis, so a position size in percentage terms is not computed — the framework does not assume one. What can be given is the risk arithmetic underneath it.

Conviction ladder: Half-Size (1 of 3 entry paths met). One valid rationale is open — the fundamental one — which the framework treats as a starter position rather than a full one. That is a 0.5× factor on whatever base allocation a reader uses.

Catalyst modifier: 0.7×. Catalyst clustering scores 40, in the 20–50 band, because three high-impact macro releases fall inside fourteen days. The framework reduces size by 30% for that, and the reduction compounds with the ladder factor: a starter position, taken smaller still.

Volatility context. ATR (14) is $3.57, or 2.3% of the share price — a typical daily range, and contracting from 2.8% two weeks ago. Beta is 0.93, so a position in ICE behaves marginally less volatile than the index in risk terms, which is the usual profile for regulated infrastructure. That said, the maximum drawdown over the past twelve months was 33% — $181.72 to $121.79 — which is a useful corrective to reading a sub-1 beta as safety. A low beta measures co-movement with the index; it does not measure how far a single-name narrative can take a stock on its own.

Risk per share. Entry at $154.31 with the stop at $146.50 is $7.81, or 5.1%. On a 1% portfolio risk budget that implies a position of roughly 20% of capital, which is far too large on any other ground — the binding constraint here is concentration and the ladder, not the stop distance.

Staggered entry. For a medium- or long-horizon position the framework suggests spreading entry over tranches rather than a single fill: a first tranche once the RSI condition clears, a second into the $147–150 shelf, and a third into the 50-day at $141.20 if it is offered. This is illustrative portfolio arithmetic, not advice.

{
  "ticker": "ICE",
  "date": "2026-08-17",
  "version": "v6",
  "brand": "",
  "company": "Intercontinental Exchange, Inc.",
  "currency": "USD",
  "exchange": "NYSE",
  "exchange_ticker": "NYSE:ICE",
  "isin": "US45866F1049",
  "api_ticker": "ICE",
  "finder_ticker": "ICE",
  "finder_exchange": "NYSE",
  "analysis_status": "on-going",
  "status_badge": "On-Going",
  "sector": "Financials",
  "gics_sector": "Financials (XLF)",
  "lifecycle_stage": "mature / capital-light exchange & data compounder",
  "user_context": {
    "horizon": null,
    "allocation_pct": null,
    "portfolio_role": null
  },
  "price_at_rating": 154.31,
  "price_asof": "2026-08-17 close",
  "eps_trailing": 7.08,
  "trailing_pe": 21.8,
  "signal_short": "HOLD",
  "signal_medium": "BUY",
  "signal_long": "BUY",
  "primary_signal": "BUY",
  "short_entry_confirmed": false,
  "short_hold_reason": "technical_pending",
  "short_cap_reason": "Short-horizon technical-confirmation cap: the base matrix gives BUY (Quality 79 High / Valuation 72 Attractive / Timing 56 Improving), but neither the Technical nor the Catalyst entry group is met, so signal_short is capped at HOLD - buy on confirmation. Two of the three Technical legs fail: the volume-or-weekly-support condition and the RSI ceiling (65.2 against 65.0). The quality-starter override does not apply, and it is disqualified TWICE OVER, not once: the override needs BOTH a Fair (40-64) valuation and a Neutral (40-54) timing score, and this run has neither - Valuation is 72 (Attractive) and Timing is 56 (Improving). Either alone takes the base row off the High/Fair/Neutral HOLD line the override is scoped to, so the base signal is BUY and the technical cap governs.",
  "wait_for_event_override": "Section-8 WAIT-FOR-EVENT override engaged and non-binding: Financials is a High macro-sensitivity sector and the FOMC minutes (19 Aug, High impact) fall 2 trading days out, inside the 3-day window. The override can only cap, and the technical-confirmation cap has already produced a non-BUY short signal, so it changes nothing. It is the reason next_update_date is 20 Aug rather than the +14d default.",
  "entry_conviction": "Half-Size",
  "entry_groups_met": 1,
  "exit_groups_live": 0,
  "exit_action": "Hold",
  "quality_score": 79,
  "valuation_score": 72,
  "timing_score": 56,
  "driver_score": 61,
  "economic_alignment_conviction": 68,
  "economic_alignment_stance": "Trend-Following",
  "economic_alignment_pressure": "Tailwind",
  "economic_alignment_source": "watchlist-signal - ICE's own line in the Economic Watchlist Forecast (macro 2026-08-12): Short O / Medium O / Long O",
  "macro_report_date": "2026-08-12",
  "overall_confidence": 65,
  "quality_detail": {
    "industry_benchmark_name": "Exchange/Data quality composite",
    "industry_benchmark_score": 84,
    "moat_score": 79,
    "roic_percentile_vs_peers": 56,
    "capital_allocation": 62,
    "management_skin_in_game": 72
  },
  "moat_score": 79,
  "valuation_detail": {
    "fcf_yield": 4.8,
    "fcf_yield_basis": "adjusted FCF ~$5.05bn annualised from the H1 2026 $2.6bn, over EV $105.4bn",
    "implied_growth_rate": 6.0,
    "consensus_growth_rate": 10.35,
    "historical_valuation_decile": 4,
    "forward_pe": 19.13,
    "forward_pe_fy2": 17.57,
    "ntm_pe": 18.13,
    "trailing_pe": 21.8,
    "ev_ebitda": 15.1,
    "ev_ebitda_provider": 15.2,
    "ebitda_ttm_usd_m": 6981,
    "dividend_yield": 1.35,
    "price_to_tangible_book": null
  },
  "timing_detail": {
    "mtf_confluence": 66,
    "risk_reward_score": 50,
    "relative_strength_vs_spy": -4.56,
    "relative_strength_vs_sector": -12.71,
    "relative_strength_window": "3-month, 15 May to 17 Aug 2026; on 1 month ICE is +10.5% vs SPY +4.0% and XLF +2.4%",
    "catalyst_clustering_score": 40,
    "dynamic_macro_weight": 0.2,
    "daily_rsi": 65.2,
    "atr_14": 3.57
  },
  "relative_strength_vs_spy": -4.56,
  "relative_strength_vs_sector": -12.71,
  "warranted_multiple": 20.37,
  "actual_multiple": 19.13,
  "val_multiple_basis": "clean forward P/E on FY2026 consensus adjusted diluted EPS of $8.07 (FMP, 8 analysts); ICE's adjusted basis already excludes acquisition-intangible amortisation and the fair-value gains on equity investments",
  "discount_rate_r": 0.0913,
  "risk_free_10y": 0.0463,
  "g_near": 0.075,
  "g_term": 0.03,
  "warranted_ratio": 0.94,
  "val_band": "attractive",
  "sector_guardrail_line": 30.0,
  "driver_commodity_trend": "N/A - ICE monetises commodity VOLUME and volatility, not the price level; the Step-2b commodity cap does not apply to a venue operator. Context pulled anyway: Brent 88.52, WTI 82.40, NatGas 2.73, all above their 50-DMA except gas (settled 14 Aug).",
  "nonop_pct_of_net_income": 20.9,
  "clean_pe": 24.07,
  "clean_peg": 2.33,
  "competitive_share_trajectory": "stable",
  "competitive_threat_level": "moderate",
  "analyst_consensus_target": 183.25,
  "analyst_target_high": 208,
  "analyst_target_low": 163,
  "analyst_target_median": 180,
  "analyst_target_upside_pct": 18.8,
  "analyst_grades_consensus": "Buy",
  "analyst_bullish_pct": 91.7,
  "analyst_coverage_count": 36,
  "recent_upgrades_30d": 0,
  "recent_downgrades_30d": 0,
  "fmp_rating": "B",
  "fmp_overall_score": 3,
  "fair_value_est": 164,
  "stop_loss": 146.5,
  "target_price": 176,
  "scenario_base_target": 176,
  "scenario_bull_target": 210,
  "scenario_bear_target": 130,
  "scenario_probabilities": {
    "bull": 25,
    "base": 55,
    "bear": 20
  },
  "gates_triggered": [],
  "gates_caution": [
    "Gate 1 - Financial Distress: leverage steps up on the MarketAxess deal, and the financing is now PERMANENT rather than bridge paper. Net debt $18.78bn today on ICE's stated $19.8bn of debt less $1.07bn unrestricted cash. On 11 and 13 August ICE filed a 424B5 prospectus supplement for $3.75bn of senior notes (4.700% due 2029 $1,250m, 4.900% due 2031 $1,100m, 5.150% due 2033 $650m, 5.400% due 2036 $750m; net proceeds ~$3.71bn, settling on or about 20 August) and its as-adjusted capitalisation table shows a $2,000m term loan and commercial paper rising 1,218 to 1,714, taking total debt 19,846 to 26,046. The filing states the bridge commitments are reduced by the offering proceeds and by any term-loan commitments. So the $6.2bn 364-day bridge is being taken out - but only PARTLY at fixed coupons: $3.75bn of the ~$6.2bn is fixed-coupon term debt at 4.70-5.40% running to 2036, while the $2,000m term loan and $1,714m of commercial paper are floating or short-dated. That leaves about 14% of the $26,046m as-adjusted total in floating or short-dated form, which is what makes DNB Trigger 1's 'predominantly floating-rate or maturing within 18 months' leg clearly false - do NOT characterise the whole structure as fixed. ICE guides to 3.4x GROSS leverage at close, back to 3.0x or below within 18-24 months; the as-adjusted table implies ~3.6x NET on ICE standalone ($26,046m less $1,067m of cash over $6,981m of trailing EBITDA) - which is not in conflict with the 3.4x GROSS guidance, because that figure is struck pro-forma for MarketAxess earnings in the denominator. Trigger is above 5x, so it does not fire. The residual sizing note is roughly $310m a year of incremental interest - about $187m of that the note coupons, the rest the $2,000m term loan and the extra commercial paper - which takes interest coverage from about 8x trailing to roughly 5.5x pro-forma, still far above the 1.5x trigger. The sharper residual is structural: the 2033 and 2036 notes carry a SPECIAL MANDATORY REDEMPTION at 101% if the merger agreement terminates or misses the outside date.",
    "Gate 4 - Accounting / Earnings quality: non-operating items were 20.9% of H1 2026 net income ($452m of fair-value gains on equity investments plus $43m of unconsolidated-investee earnings, against $2,371m). Above the 15% flag, below the ~30% gate, and every multiple in this report is struck on ICE's adjusted basis, which excludes both. Share count is falling, not rising.",
    "Gate 5 - Regulatory / Binary: ICE is the ACQUIRER of MarketAxess ($167.00 cash, ~$6.0bn). Antitrust review and the MarketAxess shareholder vote are open, and a fairness investigation is public. A $6.0bn bolt-on against an $86.6bn market capitalisation cannot move ICE more than 20% either way, so the binary test fails and the gate does not trigger."
  ],
  "do_not_buy_triggers": [],
  "dnb_trigger_4_ruling": "Insider-selling spike: does NOT fire. Four Form 4s in the last 60 days (EDGAR submissions API + ownership XML, pulled on the audit round). Only Douglas Foley (SVP HR & Administration, 5 Aug, 7,300 shares at $148.875, ~28% of the pre-trade holding, no plan cited) is a non-planned disposal above 25%. Lynn Martin (President NYSE Group, 16 Jul) exercised and sold 15,882 under a plan, holding exactly unchanged at 54,420. Mayur Kapani (CTO, 12 Aug) exercised 7,299 and sold 4,271 under a plan, taking his holding 64,960 -> 67,988, i.e. UP 3,028 shares (+4.7%) - he finished owning more stock, not less; an earlier draft wrongly called this unchanged. Martha Tirinnanzi (director, 13 Aug) sold 141 under a plan. One qualifying disposal against a threshold of three, from four filings. Does not fire.",
  "hard_gate_state": "caution",
  "systemic_tail_inheritance": "Not inherited. ICE is not in the armed S&P-concentration / AI-earnings-quality cohort - its earnings are not levered to the AI capex or monetisation trade, and the H1 fair-value gains are on market-structure stakes, not AI holdings. The separate, idiosyncratic 'AI disrupts financial data' de-rating is live and is carried in the Bear case and the thesis-invalidation rule instead.",
  "tail4_ruling": "Fed hikes in September (live, ~44% implied, trigger = the 19 Aug FOMC minutes) cuts BOTH ways for ICE, and the tempting comparison is the WRONG one. Setting the whole Exchanges segment ($1,464m a quarter) against the whole of Mortgage Technology ($557m) overstates the asymmetry roughly fivefold, because most of Exchanges is not rate-driven at all - Energy $518m, Ags and Metals $87m, cash equities $140m, data and connectivity $287m and listings $129m are levered to other things. The genuinely rate-sensitive lines are Financials at $192m (+21% in Q2) plus the collateral-interest share of OTC and other at $111m - call it $200-300m a quarter of upside exposure, helped by clearing margin deposits at $114.6bn against $76.8bn at December. Against that sits origination technology at $197m, the genuinely rate-cyclical downside with 30-year rates already at 6.77%, with servicing software at $226m running countercyclically because high rates keep loans on servicers' books. So the like-for-like read is roughly $200-300m of upside exposure against $197m of downside with $226m countercyclical behind it: balanced on the transaction lines and tilted mildly positive once servicing is counted. A hawkish print is therefore net NEUTRAL-TO-MILDLY-POSITIVE for revenue and negative for the share price on the day, because the market trades ICE as a financial rather than as a volatility long. That is NOT enough to call ICE a rate-hike beneficiary, and the hawkish leg carries the same caveat as the driver pillar: energy revenue fell 13% with the strait already shut, so volatility does not reliably convert into volume here. Dovish reverses both, smaller in each direction. The conclusion that survives is the modest one - ICE is not a one-directional bet on the minutes.",
  "post_audit_fixes_20260817": "THREE ROUNDS, then a fourth re-audit. Round 2 (both auditors re-run on the corrected artefact) returned FAIL again on ONE failure mode repeated twice - a fix that landed in the HTML prose but not in the OTHER surface carrying the same claim. (a) This tail4_ruling field was byte-identical to its pre-fix text, still asserting the repudiated whole-segment comparison and the superseded 'net positive-to-neutral' wording, while section 8 said the opposite - and both audit blocks were vouching for the fix. The calibration is the publisher's source, so a video would have voiced the repudiated version. Now rewritten to match section 8 exactly. (b) The insider fix left two contradicting references: section 3 skin-in-the-game still said no insider pull was available, and the section 15 impact block still deducted confidence for an unverified insider position - both corrected. Round-2 MINORs: the ~$187m residual interest was the note coupons ONLY, restated to ~$310m including the term loan and commercial paper, with pro-forma interest coverage restated 6x to 5.5x; the calibration Gate 1 said ~3.4x NET when the as-adjusted table gives ~3.6x net (the 3.4x is GROSS, pro-forma for MarketAxess earnings); the new section 8 paragraph said 'the ruling below' when the ruling renders above it; 'just raised' was softened in the section 12 buy box but not the sell box; and the section 8 window headings were reconciled in the subtitle. Round 3 (coordinator-relayed): the round-2 fixes verified clean on both surfaces, plus four further corrections. (i) Mayur Kapani's holding did NOT stay unchanged - he exercised 7,299 and sold 4,271, going 64,960 -> 67,988, UP 3,028 shares (+4.7%); both section 2 and section 15 had recited the numbers and drawn the wrong conclusion from them. Benign in direction (a net purchase cannot make an insider-selling trigger more likely) but it misstated a filing this report cites. (ii) 'Fixed-rate money out to 2036' was an over-claim on THREE surfaces including the calibration: only $3.75bn of the ~$6.2bn is fixed-coupon term debt, and DNB Trigger 1 is now argued from ~14% floating-or-short-dated against $26,046m, which holds cleanly. (iii) 'One of three is not a cluster' restated as 'one qualifying disposal against a threshold of three, from four filings'. (iv) A commented-out block of superseded insider text was left sitting in the published HTML and has been excised. Round 4 (fresh auditor): it reproduced the whole valuation from SKILL L451 and confirmed the anchor independently (warranted 20.365 -> 20.37x, ratio 0.9394 -> 0.94, attractive) and confirmed the earlier 'correction' to 19.115/1.0008 was fabricated - the band is settled, do not re-litigate it. Its findings: (A) section 3 skin-in-the-game STILL carried the repudiated Kapani claim - the THIRD failure of that one fix, and on a surface two previous sweeps had claimed to complete; it also said 'two' unchanged where only Martin is, and called all four filings 'disposals' when Kapani's is a net purchase. Now rewritten. (B) This field opened 'TWO ROUNDS' while narrating a third, and the section 15 audit paragraph presented round 2 as terminal - both reconciled to the rounds actually run. (C) The section 8 subtitle said 'two Medium-impact prints' when the table carries three. (D) relative_strength_vs_spy was -4.6 against a table implying -4.56; restated with vs_sector -12.71. LESSON RECORDED: the calibration is a SEPARATE SURFACE from the prose, and axis 3 of the sweep must grep BOTH - never let post_audit_fixes or section 15 assert a fix not re-read in the file it names. Round 1 findings follow. MAJOR: (1) acquisition financing described as unrefinanced 364-day bridge paper when the 11-13 Aug 424B5 had termed it out ($3.75bn of notes at 4.700-5.400% plus a $2,000m term loan; total debt 19,846 -> 26,046 as adjusted) - Gate 1, DNB Trigger 1, the balance-sheet sub-score and the section-15 freshness claim all rewritten; (2) section 8 recent-surprises table omitted the two largest High-impact prints in the window, both dovish (PPI 0.0 vs +0.2 on 13 Aug; retail sales -0.6 vs +0.1 on 14 Aug) while carrying two Medium-impact hawkish ones - table rebuilt on the High-impact set; (3) section 2 enumerated DNB Triggers 1, 2, 3 and 5 and silently skipped 4 - now ruled explicitly on Form 4 evidence; (4) section 9 showed weekly support $143.17 while sections 7 and 12 relied on $152.38 - corrected to the nearest weekly level below spot; (5) Tail-4 compared the whole Exchanges segment against the whole of Mortgage Technology, overstating the asymmetry about fivefold - rewritten on the genuinely rate-sensitive lines ($200-300m of upside exposure against $197m of downside, $226m countercyclical). MINOR: g_near rounding disclosed (7.5% used, 7.766% exact, conservative); timing composite stated as 56.5 recorded 56; EBITDA basis pinned to TTM $6.98bn so leverage and EV/EBITDA reconcile; payout basis stated; the 6% defensive-cap sensitivity corrected to ratio 1.001 and relabelled unmistakably as a sensitivity rather than the headline figure, since it tips into the fair band not the attractive one; three further tool-label overrides disclosed; VIX, the hike odds and the 2-year yield added as sources; the open-interest +18% figure re-attributed to earnings coverage rather than the release; the last-month target average flagged as a single data point; the CFO quote re-sourced to the release rather than the call; 'just raised' softened. NO SIGNAL MOVED on any finding.",
  "delta_vs_prior": "Versus the 31 July 2026 report. Price $152.85 -> $154.31 (+1.0%). Short signal cut BUY -> HOLD; medium and long both held at BUY. The short cut is a mechanism change, not a deterioration, and the half-size quality-starter that produced last month's short BUY is now disqualified on BOTH of its two legs, not one. The override requires a Fair (40-64) valuation AND a Neutral (40-54) timing score. Valuation moved to 72, which is the attractive band; timing moved to 56, which is the improving band. Either move alone would have taken the base row off the High/Fair/Neutral HOLD line the override is scoped to. On the valuation leg the prior report also mislabelled the band: it recorded 'fair' with a score of 63 on a warranted ratio of 0.99, and 0.99 falls in the attractive band (0.80-1.00), so that score sat outside its own band. With the base row now High/Attractive/Improving the base signal is BUY and the technical-confirmation cap governs instead - and the counterfactual is worth stating: on the OLD valuation of 63 the row would read High/Fair/Improving, which is also a BUY, so the technical cap would still have produced a short HOLD. Valuation score rose 63 -> 72, mostly on a restated earnings basis: the prior 'adjusted NTM EPS ~$7.45' does not reconcile to H1 2026 adjusted diluted EPS of $4.25 actual or to the FY2026 consensus of $8.07, so the actual multiple restates 20.5x -> 19.13x. Timing score rose 49 -> 56 (RSI 74 -> 65.2, OBV improving, price reclaimed the 50-day). Quality trimmed 80 -> 79 and the moat trimmed 80 -> 79; capital allocation trimmed 65 -> 62 on the MarketAxess premium. Driver score fell 64 -> 61: VIX has fallen from 20.66 on 29 July to 15.19, and net revenue growth decelerated from +20.4% in Q1 to +5% in Q2. Economic Alignment is the largest single correction: the prior run read ICE off the XLF sector map at Neutral / conviction 52 when ICE carries its OWN line in the macro Economic Watchlist Forecast (Outperform on all three horizons), so stance moves Neutral -> Trend-Following, pressure Neutral -> Tailwind and conviction 52 -> 68, and the value is now written under the canonical key economic_alignment_conviction rather than the legacy econ_conviction. Hard gate state moves clear -> caution on three caution gates (1, 4 and 5), none triggered; gates_triggered is empty in both runs. The conviction ladder is unchanged at Half-Size on 1 of 3 entry groups, exit action unchanged at Hold, short_entry_confirmed unchanged at false. The largest factual gap closed: the 31 July report never mentioned the MarketAxess acquisition, agreed 29 July, the day before it published.",
  "prior_report_corrections": "Two claims in the 31 July report are corrected here from the primary 10-Q and press release. (1) It said Q2 GAAP was 'deflated by a ~-$700m non-operating charge'. ICE's statement of income shows Q2 other income, net of +$74m and total other income/(expense), net of -$104m, of which -$205m is interest expense. H1 GAAP was INFLATED by $452m of fair-value gains on equity investments, essentially all in Q1 - the opposite sign. (2) It described itself as the 'first-ever Donatien report on ICE' when the 16 July 2026 report exists.",
  "framework_amendment_proposed": "SKILL section 8 says a High-impact release within 3 trading days on a High macro-sensitivity sector 'triggers a WAIT-FOR-EVENT short-term override regardless of composite score', but it does not say what string signal_short should carry, and the override is not listed as a step in the canonical override chain (Base Matrix -> Amplification -> Short technical cap -> Short quality-starter -> Hard Gates -> Do-Not-Buy). This report emits HOLD, because the technical-confirmation cap - which IS in the chain - independently produced a non-BUY short, and HOLD carries the actionable trigger that WAIT does not. Both independent auditors accepted the reading as legitimate and disclosed rather than an escape hatch, and both noted the SKILL should specify the emitted value. Raising it for a central decision; nothing downstream distinguishes the two, since Portfolio-Watchlist tile selection keys only on BUY versus non-BUY.",
  "next_update_date": "2026-08-20",
  "next_check_date": "2026-08-20",
  "next_update_basis": "FOMC minutes 2026-08-19 +1 trading day. Financials is a High macro-sensitivity sector and the minutes are a High-impact release inside the 3-day window, which is the one case the scheduling rule counts a recurring macro release as impactful. Earlier than the +14d default of 2026-08-31; Q3 results are 2026-10-29 (company-confirmed)."
}

Read across the pillars: a high-quality business (79) at a modestly attractive price (72) with improving but unconfirmed timing (56), a driver that has gone quiet (61) and an economy the macro report says is behind it (68, Trend-Following, Tailwind). Amplification needed the driver at 65 or above and a Tailwind pressure; only the second condition holds, so nothing is amplified and the base signals stand. Three gates read caution and none is triggered; no Do-Not-Buy trigger fires. The short horizon is capped at HOLD by the technical-confirmation rule with a single leg outstanding, and the next update is set for 20 August, the trading day after the FOMC minutes.