# Ballon d’Or: the three strongest patterns

Analyzed October 18, 2026. Event 48361; fixed universe of the 19 currently active contracts. These are defensible retrospective descriptions of API price observations. The windows and candidate pair were selected after inspecting the archive; they are not preregistered tests or out-of-sample predictions. Persistence thresholds and resampling intervals describe robustness, not statistical significance.

## 1. Yamal temporarily displaced Kane, well beyond a brief crossing

**Evidence.** Among hours with all 19 candidates observed, Yamal leads in 72 consecutive hourly bins from October 1 at 10:00 to October 4 at 10:00 UTC, end exclusive. His hourly advantage over Kane has a minimum of **1.50 percentage points**, median **7.225 points**, and peak **27.75 points**. At October 1 21:00, the two prices are 60.30% and 32.55%.

The initial screening rule was a lead of at least 1 point for six consecutive complete hourly observations. Raising it to **at least 5 points for 24 consecutive observations** still retains a 42-bin Yamal episode, October 1 10:00–October 3 04:00 UTC. This is materially different from the two- and three-hour crossings on September 30. Kane’s later 42-bin return, beginning October 4 21:00, meets the original threshold but is less robust to a higher margin requirement.

**Finer sampling changes the timing claim.** New five-minute histories for Kane and Yamal produce 863 paired observations out of 864 expected bins in the nominal 72-hour window. Yamal is ahead in **854/863 (98.96%)**, but Kane is ahead in nine bins from October 4 09:15 through 09:55 UTC. The missing pair is October 1 23:35. The five-minute price-gap median is 7.10 points; maximum cross-contract timestamp difference is four seconds. This validates pairwise predominance, not five-minute leadership over all 19 candidates. It also demonstrates why the earlier hourly result must not be described as 72 hours of continuous leadership.

**Missing-data check.** Missing hours split runs; no interpolation or carry-forward is used. The focal hourly run has all 19 observations in every bin. Source timestamps within this run differ by at most 13 seconds. Its start follows a missing complete hour, so it is a resumed observed Yamal lead, not a separately established leadership change at 10:00. Tightening the archive alignment tolerance from five minutes to one minute preserves all 2,065 complete hours.

**Resolution check.** The pair has identical award rules, the same France Football source, and the same negative-risk event ID. The text concerns the same 2026 award; deadline metadata is not used to establish equivalence. Only the currently available rule version is verified.

**Market-quality check.** Repetition across many observations and finer sampling reduce concern about an isolated price sample, but do not establish historical executability. Current YES books show Kane at 0.660/0.669 and Yamal at 0.265/0.270; those October 18 quotes cannot validate October 1–4 spreads or depth. Historical book and trade-by-trade evidence is absent. The observed price reversal is supported; an executable opportunity or cause is not established.

## 2. The field narrowed sharply toward the same two candidates

**Evidence.** Use two fixed calendar windows after the other 12 named contracts resolved No on September 8: September 9–20 and October 1–17. For each UTC day, compute the median combined, unnormalized Kane-plus-Yamal price, requiring at least 20 complete hourly observations. Then compare the medians across eligible days.

| Measure | September 9–20 | October 1–17 |
|---|---:|---:|
| Complete hourly observations | 286/288 | 386/408 |
| Eligible days (at least 20 hours) | 12/12 | 14/17 |
| Median daily Kane + Yamal price sum | 78.0000% | 92.9625% |
| Median daily sum of the other 17 prices | 23.5125% | 8.3250% |

The primary increase is **14.9625 percentage points**. The reported group medians are calculated separately, so their sums need not equal a median total. Supplementally, the pair’s median daily share of the 19-price total rises from roughly **76.73% to 91.78%**; that normalization is explicitly conditional on the current 19, not the probability of the complete event universe.

**Statistical check.** Ten thousand resamples of three-day calendar blocks yield a descriptive 95% resampling stability interval of **+14.30 to +15.775 points**. Calendar blocks preserve local temporal dependence and missing-day eligibility. These are not thousands of independent hourly trials. Requiring at least 18 hours/day gives +14.65 points; requiring all 24 gives +14.8875 points. One-, three-, and five-day block variants all retain a large positive shift. The conclusion rests on the effect size and sensitivity checks, not a post-selection p-value.

**Missing-data check.** No absent observation is converted to zero. Days below the coverage requirement remain ineligible in calendar-block resampling. The result survives both relaxed and strict daily coverage rules. First observed history dates are not interpreted as listing dates: all 19 have September 2025 creation metadata despite some histories starting later.

**Resolution check.** Comparing post–September 8 windows avoids treating the 12 excluded, formerly live named contracts as zero while they were still unresolved. The current 19 have common rules, but Other and additional augmented outcomes still prevent an exhaustive-universe interpretation. This finding concerns concentration in the fixed observed set.

**Market-quality check.** Both focal contracts currently have two-sided books. Nine of the other 17 currently have no YES bid; tiny positive indicators can therefore coexist with little near-price sellability. Historical trading volume, spreads, and depth are not available as synchronized series, so the reduction in their price sums does not demonstrate capital or attention moving to Kane and Yamal. It is a change in observed relative prices.

## 3. The 19-price total is persistently high, but does not establish inconsistency

**Evidence.** Restricting the checksum to September 9 onward avoids the period before the 12 omitted named contracts resolved No. There are **928 complete hours out of 957**, with no computed total in the 29 incomplete hours.

| Post–September 8 checksum | Result |
|---|---:|
| Median 19-price total | 101.35% |
| Range | 96.75–103.10% |
| Complete hours above 100% | 833/928 (89.8%) |
| Above 102% | 181/928 |
| Below 98% | 9/928 |

The persistence is visible at the daily level too: 35 of 36 eligible days (at least 20 complete hours) have a daily median above 100%. This is a repeated price-total property, not an isolated outlier. The ±2-point screen is an exploratory tolerance, not an arbitrage boundary.

**Missing-data check.** Only complete synchronized observations enter these counts; the realized timestamp span across candidates is at most 53 seconds across the full complete archive. One-minute eligibility gives exactly the same complete sample as five-minute eligibility. Nothing is renormalized to 100%.

**Resolution check.** The event has **89 records: 19 active/open, 12 closed and resolved No, and 58 inactive/open**. The inactive group includes Other, Gavi, and 56 alphabetic placeholders. The event is explicitly `negRiskAugmented=true`. The rule sends the result to Other if no winner is declared by December 31, 2026 at 11:59 PM ET. Thus the 19 are not a guaranteed exhaustive partition. Omitted residual outcomes can help explain a total below 100%; they do not, by themselves, explain an above-100 total.

All 89 descriptions match, but end-date metadata differs among the active contracts: 13 show January 1, 2027 UTC and six show October 31, 2026 UTC. Use the rule text, not the metadata date, to decide what is being compared. Historical rule revisions have not been reconstructed.

**Market-quality check.** Fresh October 18 YES books show nine of 19 contracts with no bid. Those nine account for **$7.92 million of $10.77 million** in reported Gamma liquidity, illustrating that headline liquidity is not near-price YES depth. Their current 0.0005 indicators add 0.45 point to the price sum; that does not explain every historical excess.

Current asks sum to **102.1%**. Existing bids sum to **99.0% only if absent bids are assigned zero for arithmetic**; that convention is not an executable sell quote for all 19 contracts. Across the ten two-sided books, spreads total 2.2 points. Across all 19, the zero-to-ask arithmetic envelope spans 3.1 points and contains 100%. These book reads span approximately 23 seconds and are not an atomic event snapshot. Gamma and CLOB snapshots are kept separate because prices moved between them.

The current evidence demonstrates why a sum of price indicators can differ from a sum of executable quotes. It does not identify the cause of every past deviation, establish a complete payoff bundle, or support an arbitrage claim. Historical spread/depth validation remains unfinished.

## Recommended visualization

**One zero-centered Kane–Yamal lead-margin timeline, focused on September 28–October 7.** Plot `Yamal price − Kane price` in percentage points using actual hourly observations. A thin zero line shows leadership crossings; a subtle ±1-point band shows narrow leads. Use one compact strip to distinguish six-hour-qualified runs from brief crossings. Leave missing bins blank, and show all-19 hourly coverage alongside the scrubbed value.

Annotate the strong October 1–3 Yamal advantage and the October 4 recrossing. The latter should explicitly distinguish the five-minute pairwise observation at 09:15 from the first Kane-leading hourly bin at 10:00. A scrubber should reveal source timestamps, both observed prices, the gap, and coverage. Avoid a 100%-normalized stack, which would conceal the checksum finding, and avoid drawing all 19 equally prominent lines.

This communicates the strongest discovery: a substantial but temporary change in the leading candidate, separated from small boundary crossings. This recommendation adds no product, account system, trading controls, or backend.

## Evidence and reproducibility

- Original hourly inputs: [aligned-hourly.json](aligned-hourly.json), [histories-full.json](histories-full.json), and [event-raw.json](event-raw.json).
- New audit evidence and scripts: [validation/README.md](validation/README.md).
- Main numerical audits: [concentration-robustness.json](validation/concentration-robustness.json), [lead-sensitivity.json](validation/lead-sensitivity.json), [fine-summary.json](validation/fine-summary.json), [post-sept8-sums.json](validation/post-sept8-sums.json), [rules-quality-audit.json](validation/rules-quality-audit.json).

Primary references: [event rules and records](https://gamma-api.polymarket.com/events/48361), [augmented negative risk](https://docs.polymarket.com/concepts/negative-risk), [price and order-book concepts](https://docs.polymarket.com/concepts/prices-orderbook), [price-history endpoint](https://docs.polymarket.com/api-reference/markets/get-prices-history). Historical observations are described as API prices; their exact midpoint/trade construction is not independently established.