Fit

Fit

Outside the framework's universe (U2 not met); contested: P2

The framework's universe test asks for a market capitalization above $10 billion. Sportradar's is $4.34 billion, and no construction of the share base or the price record in this run reaches the line. That result is reached at high confidence: two model families agreed, the trial was order-stable, and load-bearing spreads were at most 0.15. No exclusion was hit, no sensitivity flag was raised, the name-mask probe found no prior-driven risk, and one criterion — P2, free cash flow consistency — came back contested.

Sources: FY2025 Form 20-F cover page and Note 19 [1] [2], Note 19.3 treasury shares [3] and the 2026-07-27 close [4].

Universe and exclusions

U2 — scale. Not met, on a 4-0 vote across both model families, and the masked seat agreed. At December 31, 2025 the company had 221,390,294 Class A and 903,670,701 Class B ordinary shares in issue, inclusive of treasury [5]. Class B converts ten-for-one, so the gross Class A-equivalent base is 221,390,294 + 90,367,070 = 311,757,364 [6]. Treasury holds 3,939,842 Class A plus the 120,000,000 Class B parked pending cancellation, or 12,000,000 Class A-equivalents, leaving 295,817,522 net [7]. At the July 27, 2026 close of $14.68 that is $4.343 billion [8] — 43.4% of the line, $5.66 billion short of it.

The generous constructions do not close the gap either. Gross of every treasury share the figure is $4.58 billion. At the highest close in the run's price file, $19.69 on 2026-03-19, it is $6.14 billion. At the September 2021 IPO price of $27.00 per Class A share — the highest arm's-length mark verified anywhere in this corpus — it is $8.42 billion, still 16% below the line [9].

Sources: share counts from the FY2025 20-F cover page, Note 19 and Note 19.3 [10] [11] [12]; prices from the run's daily series [13]; IPO price from the 2021 prospectus [14].

The deterministic feature file could not supply this number at all — market capitalization is recorded as not computable for want of a positive period-end share count — so every figure above is built from the primary filings and the dated close, and the disagreement is recorded rather than papered over.

U1 — listing. Met, 4-0. Sportradar Group AG is a Swiss-incorporated registrant whose Class A ordinary shares, nominal CHF 0.10, are registered under Exchange Act Section 12(b) and listed on The Nasdaq Global Select Market under SRAD [15]. The security traded is the ordinary share, not a depositary receipt, so the test is satisfied on either limb. The governance fact that sits alongside it: the 903,670,701 Class B shares carry 80.3% of the votes on 29.0% of the Class A-equivalent economics [16].

The exclusion screen came back clean. All four disqualifying tests were checked against primary evidence and none was hit; the single sensitivity flag was also negative.

• X1, auto manufacture — not met. All €1,289,965k of FY2025 revenue is Betting Technology and Solutions (€1,047,070k) and Sports Content, Technology and Services (€242,895k) [17]. The named competitor set is Genius Sports, Stats Perform, Infront Sports and Media, and BetConstruct [18]. A targeted search of the revenue, segment, competition and non-current-asset disclosures found nothing on the other side.
• X2, promotion pattern — not met. The test requires cited instances on both prongs. The ownership prong is not merely unproven but contradicted: the founder-CEO held 1,840,883 Class A and 783,607,701 Class B at March 12, 2026, which is 80,201,653 Class A-equivalents of a 299,757,364 base, or 26.8% of the economics on 78.7% of the votes [19]. The counter-fact inside the same finding: he is a net seller, down from 30.2% twelve months earlier [20], over a period in which officers and directors had no Section 16(a) reporting obligation. On the delivery prong the record shows one clean guidance break — an August 2023 reaffirmation followed on November 1, 2023 by a revision to €870-880m [21] — and one long-term target reset, from the IPO-era at-least-20% growth and 25-30% margin to at-least-15% CAGR and a 27% margin at the April 2025 Investor Day [22]. One break and one reset is not the repeated pattern the exclusion names.
• X3, structural decline — not met. Revenue rose in all seven fiscal years on record, from €380.4m in FY2019 to €1,290.0m in FY2025, and the feature's consecutive-decline counter is zero. The strongest surviving fact against that reading sits on the input side, not the customer side, and is treated in the year-10 gate below.
• X4, consensus-saturated story — not met. The consensus prong is plainly satisfied: 16 strong buy, 2 buy, 4 hold and no sell across 22 analysts, mean target $21.15 against the $14.68 close, 44.1% implied [23]. The valuation prong is not. Equity value is 2.81x to 3.37x FY2025 revenue across a 1.00 to 1.20 dollar-per-euro range, and about 4.2x on an enterprise basis that adds the entire €1,529.2m licence fee payable [24]. The test needs both prongs; it has one.
• S1, China dependence — not flagged. No document in the corpus states a China revenue, asset, customer or supplier figure. Note 5 names Switzerland as domicile plus every country above 10% of revenue — Malta and the United States — and states that no individual country exceeded 10% of the total, which caps China revenue below €129.0m and caps China non-current assets inside the €109.5m "other countries" line [25]. The bound is inferred from a regional aggregate and a negative disclosure threshold, not read from a stated number.

Pattern match

The reader contract carries four recognition setups: cyclicals at the bottom, and large banks in particular; a high dividend yield paired with a high free cash flow yield on a business that is not going away; healthcare or insurance forecasting errors that reprice out; and quality tech monopolies or duopolies on a fear dip. Sportradar fits none of the four squarely.

The nearest lens is the fourth, and it is worth running because the surface features match: a data and rights layer with few credible suppliers, priced 25.44% below its March 2026 high after a specific, testable fear — two short-seller reports alleging revenue from unlicensed operators [26]. Three of that pattern's own checks fail. The market structure is an oligopoly with four named rivals rather than a monopoly or duopoly [27]. The scale is a fifth of the pattern's precedents and below the framework's own universe line. And the pattern assumes a business whose cash generation is visible through the dip; here the adjusted yield is 1.55% to 1.66% against an 8.5% bar. The second setup is unavailable outright — no dividend has ever been declared [28] — and the first and third describe balance-sheet and regulated-repricing businesses this is not.

The pillar ledger

Every criterion below is the tally's aggregate. Reference lines are stated as arithmetic; the thresholds are the framework's, not a score.

Source: the run's deterministic fit tally, aggregating four jury seats plus a name-masked seat over the pillar claims ledgers.

Year-10 gate

P1 — not met, 4-0 across both families, masked seat agreeing. Trimmed-mean probability 0.685, spread 0.15. The gate asks whether year-10 revenue and year-10 adjusted free cash flow will both be higher than today's, with very high conviction, and it resolves genuine doubt to not met by construction.

What supports the durable reading is contracted and disclosed. Unsatisfied performance obligations stood at €2,122.6m, of which €1,357.0m is scheduled for 2026 recognition [29]. Two thirds of revenue sits on fixed-fee recurring contracts of one to five years [30], no customer exceeds 10% of revenue and the top ten are 27% [31], and the end market the company cites is projected to compound 9% a year from $127.1bn in 2025 to $192.7bn in 2030 [32]. The disqualifier is not triggered: seven consecutive years of revenue growth, zero consecutive decline years.

The strongest surviving counter-facts sit in the same treatment, and they are what moved the jury. That backlog is front-loaded — 63.9% of it lands inside 2026 and 5.7% beyond 2028 [33]. The only disclosed expansion metric fell to 109% in 2025 from 127% in 2024, on a top-200 cohort that shrank from 83% to roughly 79% of revenue, and the filing gives no decomposition of that fall between pricing, contraction and attrition [34]. And the input side re-priced: sport rights expense ran 24.41%, 31.85% and 31.34% of revenue across FY2023 to FY2025, a 6.93-point re-basing in one renewal round of which only 0.51 points has come back [35]. That reinvestment is financed on vendor credit: €1,529.2m of licence fee payables for capitalized sport rights [36], against total loans and borrowings of €62.9m — a ratio of 24.3x [37]. Against that, the cost is already inside the reported numbers, €270.2m of it amortization rather than a cash call [38], and free cash flow still rose from €58.4m in FY2023 to €174.7m in FY2025 [39]. The company also names its own year-10 hazards: disintermediation by leagues, and prediction markets that "may reduce the size, scale, or growth of traditional sports betting markets" [40]. Full treatment in The Barrier and the Bill.

The framework's other conviction source is unavailable by construction: the business was founded in 2001 and has been public since September 2021, so there is no 30-to-50-year record against which to test through-cycle durability.

Consistency

P2 — contested. Two seats read not met, two read cannot determine; the masked seat read not met. This is the only criterion where the two model families diverged, and it is treated in full under Contested and undetermined below.

Dislocation and yield

P3a — identifiable event. Met, 4-0. The trigger is dated, named and carries a mechanism: short-seller reports published on April 22, 2026 alleging revenue from illegal gambling operators, on which the shares fell 22.57% in a single session, from $16.84 to $13.04, on 27,482,052 shares [41] [42]. Counter-fact in the same treatment: 38.8% of the $7.34 peak-to-trough decline preceded the event, with the stock falling 14.47% from $19.69 on March 19 to $16.84 on April 21 and no dated cause anywhere in the corpus for that leg. The event is also an outside allegation rather than a company disclosure of worse numbers.

P3b — capitulation. Met, three seats to one cannot determine. The reference line is a fall-leg volume multiple of at least 2x the trailing median. The deterministic feature prints 1.6253x, below the line — and its denominator is one trading session, because the run's price file begins on 2026-03-18, a single session before the peak, rather than the 180 sessions the definition requires. Measured on the tape that does exist, the event session traded 12.31x the 23-session pre-event median and 11.30x the 90-session median, and the five sessions from the event to the trough averaged 5.79x [43]. Both readings are on the page; the feature's figure was reported, not replaced.

P3c — yield against the bar. Not met, 4-0. On the framework's own basis — free cash flow less stock-based compensation less the trailing five-year average of cash spent on acquisitions — FY2025 adjusted free cash flow is €174.7m − €54.9m − €57.3m = €62.5m [44]. Against a market capitalization of roughly $4,343m that is an adjusted yield of 1.55% to 1.66% across the run's plausible currency range, 684 to 695 basis points below the 8.5% fortress bar and roughly 2,340 basis points below the 25% levered bar. The three-year average is worse: €17.0m of adjusted free cash flow across FY2023 to FY2025, a 0.42% to 0.45% yield, some 806 basis points from the bar, because FY2023 was negative €38.5m.

Which bar applies is itself two-sided, and the answer does not change the result. On the reported loans-and-borrowings line the company is net cash by €302.4m — borrowings of €62,852k are entirely lease liabilities against €365,295k of cash, with a €220.0m revolver undrawn [45] — which selects the 8.5% fortress bar [46] [47]. Including the €1,530.9m of interest-bearing sport-rights payables, which cost €85.6m of cash interest in FY2025, net debt is €1,228.5m and 4.14x FY2025 adjusted EBITDA of €296.8m, which selects the 25% levered bar [48] [49]. The measured yield clears neither. The fortress signature the framework looks for — a stable 3.5% to 4% historical yield that suddenly jumped toward the bar — does not exist here, because adjusted free cash flow was negative in FY2023 and €26.8m in FY2024. Full treatment in Rights Bought on Credit.

P3d — forward path. Not met, 4-0. Trimmed-mean probability 0.285, spread 0.12. Consensus free cash flow of €227.0m, €296.8m, €375.5m and €492.5m for FY2026 to FY2029, less growing share-based compensation and the trailing acquisition average, gives adjusted yields of 3.6-4.1%, 5.4-6.2%, 7.2-8.4% and 9.9-11.6% on today's capitalization. The 8.5% bar is first cleared in FY2029 — one year outside the three-year window.

Source: derived from reported financials — consensus free cash flow from the run's estimates record [50], adjusted using share-based compensation and cash acquisitions from the FY2025 consolidated statements of cash flows [51]; the band spans the tested currency and share-compensation treatments.

The counter-fact belongs in the same breath. On the most generous permissible treatment — share-based compensation frozen at FY2025's €54.9m and a weak dollar — FY2028 computes to 8.39%, 11 basis points short, which is inside the estimation error of the compensation assumption itself. And the repurchase authorization is a genuine mechanism: $1.0 billion, equal to 23.0% of the market capitalization, which if fully executed at prices near $14.68 would lift the FY2028 figure above 9% [52]. Against that, total liquidity was €542m at March 31, 2026, so most of that programme would have to be funded out of the same forward cash flow being counted in the numerator. The skeptic pass weakened the pillar's own point estimate of 0.25 to a qualitative near-miss, on the ground that the corpus does not support a deterministic figure; the tally's 0.285 is the jury's aggregate, not the ledger's.

Balance sheet and self-help

P4a — outlast and allocation headroom. Met, 4-0. There is no debt maturity schedule to build. The entire loans-and-borrowings balance is lease liabilities [53], the €220.0m revolving facility is undrawn, and the company was in covenant compliance at December 31, 2025 and expects to remain so [54]. No refinancing event can force a paydown that displaces repurchases. The obligation with any power to crowd them out is the rights bill, not debt: €319.4m of licence payables fall due within twelve months, up 79% from €178.3m, and €1,209.9m beyond [55]. In the most recent quarter liquidity fell from €585m to €542m while €91m of repurchases ran, so the buyback is currently drawing on the balance sheet rather than being covered by the €44m of quarterly free cash flow [56] [57].

P4b — repurchase engine. Not met, 4-0. Repurchases are executed cash, not authorization language: €9.0m, €28.7m and €105.2m across FY2023 to FY2025 on the financing line of the audited cash-flow statement, plus €91m in Q1 2026 alone [58]. What three years of that bought is the point the criterion turns on, and the answer depends on which share-count basis governs. Point-in-time, Class A-equivalents net of treasury fell from 297,930,657 to 295,817,522, a decline of 0.71%. On the weighted-average diluted series the company itself reports for earnings per share, the same base rose from 317,013k to 319,365k, up 0.74% [59] [60]. Either way the two-year movement is under 1% in absolute value: no demonstrated multi-year reduction, and no unambiguous rise either, so the hard-fail rule on a rising share count did not fire. The deterministic feature that would settle it is null. The offsetting fact is that the issuance is compensation and licensor equity rather than acquisition currency — €59.9m of share-based compensation expensed in FY2025 [61] against €105.2m of gross treasury spend [62], and that of the 5,127,074 shares bought in FY2025, 4,130,572 ran under the announced programme at an average $22.10 [63]. Full treatment in Guidance and Control.

P4c — dividend cover. Not applicable, 4-0. The criterion activates when a dividend yield of roughly 4% or more is material to the return case. Sportradar has never declared or paid a dividend since incorporation in 2021 and does not anticipate one [64]. The yield is 0%, so there is no cover ratio, no downturn record and no commitment language to test. This is an absence of content, not an absence of data.

Diagnosis

P5 — temporary versus permanent. Met, 4-0. Probability that the impairment is temporary: 0.66, spread 0.05, carried from the adversarial trial. The trial ran three independent judges over two opposing cited briefs; reading order barely moved the answer, with a temporary-first mean of 0.66 against a permanent-first mean of 0.685, a gap of 0.025.

The supporting fact is a guidance delta of exactly zero. The FY2026 outlook table published March 3, 2026 and the table published April 28, 2026 — six days after the short reports — are line for line identical: €1,557-1,582m revenue at 23-25% constant currency, €390-400m Adjusted EBITDA, 25.0-25.3% margin, free cash flow conversion above 56% [65] [66]. Three months later FY2026 consensus revenue of €1,556.2m still sits 0.05% under the guided floor. Nothing in the reported operating record broke: the last printed quarter grew revenue 11% to €347m, Adjusted EBITDA 12% to €66m and free cash flow to €44m [67].

Here is the qualification that belongs with it. That intact consensus is the weakest available witness to its own resilience — the 21-analyst FY2026 revenue high is also at the guidance floor, and exactly one of 22 published price targets sits below the market [68]. No published estimate line in the corpus has been rebuilt on a reduced revenue base, which is absence of analysis rather than evidence of an intact valuation. The quarter above closed on March 31, three weeks before the reports, so the corpus holds no post-allegation operating observation at all; the first is the Q2 print scheduled for August 3, 2026. And the size of the contested base is unmeasured from outside: the reports estimated 20-40% of revenue, the chief executive put it at "5% to 12%, 13%" on the April 28 call from a client-by-client build the company has not disclosed [69] [70]. Full treatment in Priced for an Allegation.

Instrument context

I1 — not verifiable, 4-0. This criterion never blocks a pillar verdict; it drives the watchlist overlay, which does not apply here because the overlay attaches only to a fitting or leaning-toward-fitting result. Listed options carrying a January 21, 2028 expiry existed as of the July 27, 2026 quote — 543 days, or 17.8 months, clearing the 12-month floor and just short of the 18-month target — and implied volatility was retrievable from a dated source at a 72.53% 30-day mean and a 62.47% 180-day mean, the 30-day reading sitting above the framework's 60-70 elevated band with results six days out. The open-interest and spread prong could not be verified at either qualifying expiry: only the front-month August 21, 2026 chain was served by any reachable source. The skeptic pass marked the whole claim unverifiable because its support is external and dated rather than a corpus page, and this tab does not link it as a filing citation.

What a 3x-in-3-years would require

The framework's target test values normalized adjusted free cash flow at the applicable bar yield and asks what consensus would have to concede to get there. The tally records the arithmetic as unavailable: "Re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing." Both inputs are genuinely missing rather than unfavourable.

The bar is undetermined because the balance-sheet class is. The deterministic feature returns "unknown", and the two defensible readings of the same balance sheet — net cash of €302.4m on the reported borrowings line, or net debt of €1,228.5m at 4.14x adjusted EBITDA once the sport-rights payables are counted — select 8.5% and 25% respectively, a three-fold spread in the applicable line. Normalized adjusted free cash flow is not computable on the framework's own definition: the feature file carries no share-based compensation for any year FY2019 to FY2025 and no complete consecutive five-year acquisition window, and reconstruction from the primary cash-flow statements closes only three years, FY2023 to FY2025, at −€38.5m, €26.8m and €62.5m. A three-year mean of €17.0m is not a normalized figure; it is a trailing average across an investment step-up.

Base-rate context from prior drawdown episodes is also unavailable in this run. The price file holds 90 sessions beginning 2026-03-18 and is flagged as partial history, so no cycle high, no prior peak-to-trough episode and no recovery base rate can be measured — the drawdown labelled against a three-year high is in fact computed over that 90-session window. What can be stated is the observed path: 37.28% peak to trough, and 25.44% below the peak at the July 27 close after an 18.87% recovery off the low [71].

Contested and undetermined

One criterion was contested. Nothing was returned as cannot-determine at the tally level.

P2 — free cash flow consistency. Contested: two seats not met, two seats cannot determine, across the two model families; the name-masked seat read not met.

The criterion asks whether the rolling five-year average of adjusted free cash flow is stable, with any negative episode explained by a business-model-inherent cycle recurring every five to eight years. Both readings rest on the same fact: that rolling average cannot be computed.

The not-met reading. The metric is unavailable, but the question settles anyway on the years that do compute. Adjusted free cash flow runs −€38.5m, €26.8m and €62.5m across FY2023 to FY2025 — one negative year in three, a mean of €17.0m, a peak-to-trough range of 5.96x that mean [72]. On unadjusted free cash flow the seven-year record has a coefficient of variation of 86%, including €1.5m in FY2021 and €5.5m in FY2022 — years in which revenue grew 38.6% and 30.1%, so those troughs cannot be the inherent cycle the criterion's escape hatch names. They are capital-intensity timing.

The cannot-determine reading. The designated metric is missing and the substitutes do not replace it. Revenue growth and an investment-timing explanation answer a different question from the one the criterion asks, and three years is not five.

The named missing datapoints, recorded by the seats that returned cannot determine: "five consecutive adjusted-FCF years needed to compute rolling 5-year adjusted FCF stability", and "five consecutive adjusted-FCF years; FY2015-FY2018 cash-acquisition inputs needed for the rolling 5-year adjusted-FCF stability metric". The gap is structural, not clerical: the framework's trailing five-fiscal-year acquisition term for FY2021 needs cash-acquisition figures back to FY2017, and the earliest audited consolidated cash-flow column that exists for this company anywhere is FY2019, in the IPO registration statement.

The fact that cuts the other way, in the same treatment: cash paid for intangible assets has fallen as a share of revenue in every year, 24.1% in FY2019 to 17.3% in FY2025, and absolute cash intangible capex was flat year over year in FY2025 on revenue up 16.6% — which is what a step-up that is genuinely finishing looks like [73].

Because the universe test decides the overall result, P2's contested status does not change it. It is recorded here as contested rather than resolved.

Provenance

Source: the run's deterministic fit tally and its skeptic ledger.

How hard the verdict was pressed, in short. Every criterion was voted independently by four seats drawn from two different model families, plus a fifth seat that saw the same evidence with the company's name removed — and the masked seat reached the same verdict on every gate, which is what makes the prior-driven-risk flag negative. Each verdict-critical claim was then re-derived from its cited pages by a separate skeptic: of 61 claims, 17 were pressed beyond triage, and the two that did not survive intact were a probability point estimate downgraded to a qualitative near-miss and a claim that no adversarial trial existed, refuted by the trial artifacts themselves.

The falsifier ledger

These are the standing conditions that would change the reading. The first five are the framework's own templates; the rest are the name-specific versions nominated by the trial judges and the jury seats, several of which converged on the same August 3 print and the same licensing mechanism from different directions.

• adjusted FCF or EBITDA declines where flat-or-better was underwritten
• revenue declines for a third consecutive year
• capital allocation pivots to debt paydown over repurchases
• share count inflects upward
• the industry repricing cycle fails to materialize where industry-wide mean reversion was underwritten
• Q2 results on 3 August 2026 cut or qualify the FY2026 EUR1,557-1,582m revenue / EUR390-400m Adjusted EBITDA guide, or net retention falls below 100% - the guide requires 23-25% cc growth against Q1's actual 11% reported / 16% cc, so the reaffirmation is the load-bearing claim and it is testable within days.
• Q2 results on 3 August 2026 cut the FY2026 EUR1,557-1,582m revenue or EUR390-400m Adjusted EBITDA guidance, or net retention falls below 100%.
• Q2 2026 or FY2026 reporting cuts the €1,557-€1,582m revenue or €390-€400m Adjusted EBITDA guide, or net retention falls below 100%.
• Any formal enforcement, licensing or suitability action opened against Sportradar itself by a US state regulator, the UKGC, the MGA or the DOJ - the mechanism by which permanent pruning actually occurs, and the only thing that converts the FY2025 Annual Report's own 'may ultimately result in a loss of revenue' language into realised loss.
• A formal enforcement, licensing or suitability action is opened against Sportradar itself (UK, Malta, a US state regulator, or DOJ) by year-end 2026.
• A regulator or licensing authority opens a formal action requiring Sportradar to cease, restrict, or remediate disputed customer relationships.
• The FY2026 20-F or class-action discovery disclosing gray-market exposure materially above the 12-13% ceiling management gave on the Q1 call - i.e. any independent evidence moving toward Muddy Waters' 20-40%, since today's residual $674m of lost value roughly capitalises the 13% band in full and the incremental downside lives only above it.
• Audited disclosure or discovery shows illegal/gray exposure materially above management's low-to-mid-single-digit estimate or above the 12% adverse-simulation bound.
• The FY2026 20-F contradicts the 12%-or-below bound - restatement, auditor emphasis-of-matter, or disclosed exposure materially above management's figure - or the CEO's pledged $10m personal purchase goes unexecuted.
• A Tier-1 operator or a major rights holder (MLB/NBA/UEFA/ATP) terminating or declining renewal on revenue-quality grounds, or the CEO's pledged $10m personal purchase going unexecuted after the trading window opened.
• Disclosed customer pruning or a counterparty defection: a Tier-1 operator or any of MLB/NBA/UEFA/ATP terminating or declining renewal, or exited gray-market clients exceeding ~5% of revenue.
• A Tier-1 operator or major rights partner terminates, declines renewal, or reprices on terms that leave rights expense/payables structurally mismatched to revenue.

Data gaps

What this run could not answer, as recorded by the tally.

The feature file could not compute the verdict-bearing numbers. Market capitalization is null for want of a positive period-end share count, and adjusted FCF, adjusted FCF yield, the yield baseline, float retirement years, balance-sheet class, FCF stability and the share-count trend are null in consequence. Every market-capitalization, yield and share-count figure on this tab is reconstructed from primary filings instead. Three disagreements are recorded rather than silently substituted: share-based compensation is on the face of every cash-flow statement despite the feature reporting it missing; borrowings and cash are both on the FY2025 balance sheet despite the balance-sheet class reporting them missing; and the canonical facts file reports zero cash acquisitions for all seven years against audited figures of €198.4m (FY2021), €56.2m (FY2022), €12.8m (FY2023), €27.1m (FY2024) and a €7.9m inflow (FY2025) — an error worth about 142 basis points of yield.

Adjusted free cash flow closes only three years. Five consecutive adjusted-FCF years are needed for the stability metric; FY2015-FY2018 cash-acquisition inputs do not exist anywhere, the earliest audited cash-flow column being FY2019. Two defensible definitions of share-based compensation also exist and the framework does not disambiguate them — the cash-flow add-back of €54,877k or the €59,918k expensed in the income statement — giving FY2025 adjusted FCF of €62.5m or €57.5m. Sign and ranking are unaffected in all three years.

The price history is 90 sessions. It begins 2026-03-18 and is flagged partial. There is no 52-week or multi-year high, no fiscal-year-end close before 2026, and therefore no measured per-year yield history and no company trading-multiple history. The 180-day pre-peak median volume the specified volume-spike multiple requires does not exist; the printed 1.63x divides by a single session. The technical-analysis files return "unavailable" and an illiquidity flag for the same reason, and neither should be read as evidence about the shares' tradability — they are listed on Nasdaq and traded 1.90m shares a session on average over the 20 sessions to July 27.

No exchange rate is staged anywhere in this run. Every comparison between a dollar market capitalization and a euro reporting base carries an unstated assumption; yields are reported across a 1.0769 to 1.15 band, and the price-to-sales multiple across 1.00 to 1.20. The conclusion is invariant across the range for the current yield and for the sales multiple; it is material to the forward path alone, where the same range moves the FY2028 figure from 6.9% to 8.8% and straddles the bar.

The retention fall cannot be decomposed. The metric is defined net of contraction and attrition and excluding new customers, but no split between pricing, contraction and attrition is disclosed, no customer count has been published since FY2022, no revenue bridge separates price from volume in any period, and no account is named anywhere as won from or lost to a competitor.

The contested revenue base is unmeasured from outside. No filing discloses revenue by market-legality status and no period shows revenue actually lost on that basis. The short reports themselves and the company's April 28 Form 6-K response are not in the corpus, so both the allegations and the rebuttal are known only second-hand. In most European jurisdictions the company's business-to-business supply requires no licence, so no external party can test a European client list against a supervisory register [74]. The drop-through on any lost revenue is also undisclosed: the honest band runs from 23.0% to 100%, which moves the implied permanent revenue loss on the event window from 13.78% to 3.17%.

Consensus exists in one vintage. The panel was updated 2026-07-23, three months after the event, with no revision-history fields, so the before-and-after consensus delta is not computable and the unchanged March and April guidance tables are used as the substitute. Coverage thins sharply — 21 estimates at FY2027, 12-13 at FY2028, two at FY2029 — and no estimate count or share-based compensation estimate is published for the free-cash-flow line at any year.

There is no post-event operating observation. The last reported period ended March 31, 2026, three weeks before the reports; the first test is August 3, 2026.

Positioning and peer context could not be run. Short interest is unavailable in every file, so days-to-cover and any change in positioning around April 22 are not computable. No peer or index price series exists, so the 14.47% pre-event drift could not be tested against a sector move. External web research was unavailable this pass — the research provider returned an insufficient-credit error on three attempts — so nothing beyond the corpus could confirm the drift's cause. The share of the recovery attributable to the company's own repurchases cannot be separated either: a $250m enhanced programme was announced on the trough session and €91m ran in Q1 alone, with no daily or monthly execution disclosure.

Instrument data is thin at the tenor that qualifies. No reachable dated source published open interest or bid/ask for the January 2028 or July 2027 expiries, no implied-volatility rank or percentile history was retrievable, and no term-structure point between 30 and 180 days was available, so the band judgment rests on two point observations dated 2026-07-27 with results six days out.

Two share-count ambiguities remain open. No share count is published for any date after December 31, 2025, and roughly 6-7 million further shares are implied retired since — about 2% of the count, which would raise every yield here by about 2% relative. And the FY2025 Item 16E table is internally inconsistent: the column header attributes 4,130,572 shares to the announced programme while its own footnote states the same 4,130,572 were purchased outside any programme. The narrative and Note 19.3 reading is used here; a reader taking the footnote literally would conclude that essentially none of FY2025's repurchases were programme purchases [75].

Two structural absences, recorded as such. No roll-forward of the sport-rights licence payable is disclosed, so the FY2025 increase cannot be split cleanly between obligations assumed with the IMG ARENA acquisition and obligations newly incurred. And no filing confirms execution of the chief executive's stated intention to purchase $10 million of shares personally, nor could one be expected: directors and officers had no Section 16(a) obligation until March 18, 2026.

Checked and unremarkable

No scout memo closed as routine or empty in this run. All eight — accounting and cash quality, business economics, capital allocation, competition and moat, history and track record, industry, people and governance, and valuation and expectations — returned load-bearing findings, and each is carried at full strength in the evidence chapters rather than summarised here.

Playbook version

Rendered against playbook fcf-dislocation, version 4, as frozen for this run.