The homepage tells the story. This page proves it. The ledger audits it.
Everything below is drawn from real captures and recorded research
records. Each chart carries its sample size, its source, the date it was computed and
an honest research status. Where the data does not exist, the gap is published rather
than papered over.
The largest single analysis behind the programme benchmarked
531,115 runner-race records against the market's own implied
probability, and returned a well-powered null: none of the structural predictors tested in that study
beat the price.
That null is published, not hidden. The grading instrument has since assessed
275 races prospectively.
At 60 minutes before the race, the favourite's implied chance was on average 4.48 percentage points away from where it would stand one minute before the off. By three minutes out, 1.43. The field average moves roughly half as far at every stage.
Prices can move in either direction. This measures the SIZE of the remaining change, not whether the horse shortened or drifted.
FavouriteField average
Favourites move roughly twice as much as the field average.
Samplen = 784 · 7,077 runner observations per checkpoint
SourceT-gate exchange receipts, 2026-07-16 to 2026-08-20
As at2026-08-20
StatusHISTORICAL
The market record behind the research
Elite Pass captures the exchange at fixed checkpoints before every race it follows, and keeps the full board each time, not just the favourite. Across 32 racing days that is 7,261 captures over 1,180 races. An eighth checkpoint at two minutes was added recently and is excluded here, because it does not span the window.
SourceT-gate exchange receipts, 2026-07-16 to 2026-08-20
As at2026-08-20
StatusHISTORICAL
The grading instrument
What the grading instrument does with a card
Across seven race days the grading instrument assessed 275 races. It issued 72 AMBER, its lower tier, meaning the evidence converges but not completely, and declined to grade the other 203. It has never issued a GREEN. GREEN is currently unreachable because one required component is not frozen, and the grading instrument will not award a grade whose inputs are incomplete. A grade is evidence convergence, not a prediction: it applies to one runner and it is not a bet.
This is the chart we would most like to show you, and the one we will not fabricate. EP Rank needs a fitted weighting between its component views. The two evidence lanes that would justify one currently stand at zero of the required sample, so the scoring function refuses to return a value at all. When it fits, it will be published here with its sample size like everything else.
DATA GAPmarket lane 0 of 181 races required · conversion lane 0 of 391 leader-rows required
Sampleno admissible sample
Sourcethe original research record
As at2026-08-21
Statusnot captured
The scoring function is deliberately locked: its weight vector is unset,
and calling it raises an error rather than returning a plausible-looking number. The two
evidence lanes that would justify a weighting currently stand at zero of their required
samples. When a fitted model exists, its output will be published here with its sample
size like everything else.
Research findings · full frames
The public findings on the
Research page, with their complete
samples, sources and statuses. The favourite analysis has its own internal report
covering the corpus, eligibility rules, how joint favourites were handled, what was
excluded and the confidence intervals.
Across 60,671 races, favourites won 35.0%, and failed to win 65.0%.
The favourite is usually the horse the market considers most likely to win. Most likely does not mean certain: across every race in our historical corpus with a complete Betfair book and a single clear favourite, the favourite lost nearly twice as often as it won.
SourceFavourite research report 1.0 (Elite Pass historical racing archive, recomputed 2026-08-21)
As at2026-08-21
StatusHISTORICAL
Shorter favourites win more often, almost exactly as often as the market expected.
Are all favourites equal? No, a 6/4 favourite and a 5/1 favourite are different propositions, and the win rates show it. But at every price band, the actual win rate lands within a percentage point of what the market's own prices implied. The shorter the favourite, the more often it wins. The market knows that too.
Actual win rateWhat the market expected
BSP up to 2.0061.4%61.5%
2.01 – 3.0040.7%41.1%
3.01 – 4.0029.8%30.0%
4.01 – 6.0023.0%22.4%
over 6.0015.0%15.9%
Samplen = 60,671 · five BSP bands
SourceFavourite research report 1.0 (Elite Pass historical racing archive, recomputed 2026-08-21)
As at2026-08-21
StatusHISTORICAL
Favourites are more vulnerable in big fields, and the market prices that in too.
In small fields the favourite wins nearly half the time; with sixteen or more runners, barely a fifth. That is a real, pre-race, structural difference, and at every field size the market's expectation tracks it. Field size alone tells you where favourites fail more often. It does not, on its own, tell you anything the price hasn't already said.
Actual win rateWhat the market expected
2–7 runners43.6%44.1%
8–11 runners31.9%31.8%
12–15 runners26.8%27.0%
16+ runners22.4%21.2%
Samplen = 60,671 · pre-defined field-size bands
SourceFavourite research report 1.0 (Elite Pass historical racing archive, recomputed 2026-08-21)
As at2026-08-21
StatusHISTORICAL
Good form finds better horses. The market already knows they're better.
Runners our morning rating ranks top win far more often when the market also likes them, a gradient from 41.8% down to 11.0%. Convincing, until you ask what the price alone predicted: 40.7% down to 12.0%. The gradient is the price. In a separate half-million-runner benchmark, no structural predictor we tested beat the market either.
Actual win rateWhat the market expected
Market agrees41.8%40.7%
Market neutral20.9%23.6%
Market disagrees11.0%12.0%
Samplen = 2,487 · races with a unique top-rated runner; gradient measured against BSP
Each figure published anywhere on this site
carries its own definition, source, date and governance status. Nothing is a rounded-up
marketing number.
The research module1 figure · as at 2026-08-21
n = 0EP Rank: no fitted modelThis is the chart we would most like to show you, and the one we will not fabricate. EP Rank needs a fitted weighting between its component views. The two evidence lanes that would justify one currently stand at zero of the required sample, so the scoring function refuses to return a value at all. When it fits, it will be published here with its sample size like everything else.not captured
Technical record
Sourcethe research module
As at2026-08-21
StatusDATA_GAP
Favourite research report 1.0 (Elite Pass historical racing archive, recomputed 2026-08-21)6 figures · as at 2026-08-21
n = 60,671Favourites are more vulnerable in big fields, and the market prices that in too.In small fields the favourite wins nearly half the time; with sixteen or more runners, barely a fifth. That is a real, pre-race, structural difference, and at every field size the market's expectation tracks it. Field size alone tells you where favourites fail more often. It does not, on its own, tell you anything the price hasn't already said.HISTORICAL
n = 60,671Shorter favourites win more often, almost exactly as often as the market expected.Are all favourites equal? No, a 6/4 favourite and a 5/1 favourite are different propositions, and the win rates show it. But at every price band, the actual win rate lands within a percentage point of what the market's own prices implied. The shorter the favourite, the more often it wins. The market knows that too.HISTORICAL
n = 60,671Across 60,671 races, favourites won 35.0%, and failed to win 65.0%.The favourite is usually the horse the market considers most likely to win. Most likely does not mean certain: across every race in our historical corpus with a complete Betfair book and a single clear favourite, the favourite lost nearly twice as often as it won.HISTORICAL
60,671Races in the favourite studyRaces in the recorded favourite study: every race in the Elite Pass historical racing archive with a complete Betfair book and a single clear favourite. The favourite won 35.0% and did not win 65.0%.derived from a recorded research file
22.4%Favourites that won in fields of 16 or moreShare of clear favourites that won in races with sixteen or more runners, the largest pre-defined field-size band of the favourite study.derived from a recorded research file
43.6%Favourites that won in fields of 2 to 7Share of clear favourites that won in races with two to seven runners, the smallest pre-defined field-size band of the favourite study.derived from a recorded research file
Technical record
SourceFAVOURITE_RESEARCH_REPORT_1.0 (Elite Pass historical racing archive, recomputed 2026-08-21)
As at2026-08-21
StatusDERIVED_FROM_ARTEFACT, HISTORICAL
Market conditional adjudication 2026-08-19 (recorded adjudication)2 figures · as at 2026-08-19
n = 2,487Good form finds better horses. The market already knows they're better.Runners our morning rating ranks top win far more often when the market also likes them, a gradient from 41.8% down to 11.0%. Convincing, until you ask what the price alone predicted: 40.7% down to 12.0%. The gradient is the price. In a separate half-million-runner benchmark, no structural predictor we tested beat the market either.HISTORICAL
2,487Races in the form-versus-price adjudicationRaces with a unique top-rated runner on the morning ability rating, compared against what Betfair SP alone predicted; the same sample the form-versus-price chart draws.derived from a recorded research file
Daily grades 2026-08-* (frozen the race grading specification 1.0)1 figure · as at 2026-08-21
n = 275What the grading instrument does with a cardAcross seven race days the grading instrument assessed 275 races. It issued 72 AMBER, its lower tier, meaning the evidence converges but not completely, and declined to grade the other 203. It has never issued a GREEN. GREEN is currently unreachable because one required component is not frozen, and the grading instrument will not award a grade whose inputs are incomplete. A grade is evidence convergence, not a prediction: it applies to one runner and it is not a bet.PROSPECTIVE
Technical record
SourceDAILY_GRADES_2026-08-* (frozen the race grading specification 1.0)
As at2026-08-21
StatusPROSPECTIVE
T-gate exchange receipts, 2026-07-16 to 2026-08-205 figures · as at 2026-08-20
n = 784How much movement was still left in the market?At 60 minutes before the race, the favourite's implied chance was on average 4.48 percentage points away from where it would stand one minute before the off. By three minutes out, 1.43. The field average moves roughly half as far at every stage.HISTORICAL
n = 1,180The market record behind the researchElite Pass captures the exchange at fixed checkpoints before every race it follows, and keeps the full board each time, not just the favourite. Across 32 racing days that is 7,261 captures over 1,180 races. An eighth checkpoint at two minutes was added recently and is excluded here, because it does not span the window.HISTORICAL
1.43ppFavourite's average remaining movement at 3 minutes outAt 3 minutes before the off, the mean gap between the favourite's implied chance and where it stood one minute before the off.derived from a recorded research file
4.48ppFavourite's average remaining movement at 60 minutes outAt 60 minutes before the off, the mean gap between the favourite's implied chance and where it stood one minute before the off.derived from a recorded research file
784Races in the market-movement studyRaces with a timestamped exchange receipt at every checkpoint from 60 minutes out to the final minute, over which the favourite's remaining movement was measured.derived from a recorded research file
Technical record
SourceT-gate exchange receipts, 2026-07-16 to 2026-08-20
As at2026-08-20
StatusDERIVED_FROM_ARTEFACT, HISTORICAL
Archive sweep panel 2026-08-17 §2(a), independently re-run by two reviewers; paired on the same 1,997 races / 18,529 runners with proportional overround normalisation8 figures · as at 2026-08-17
0.925 to 1.023Betfair SP calibration 95% confidence interval95% confidence interval for the Betfair SP calibration slope. It CONTAINS 1.000, so the exchange price is not statistically distinguishable from perfect calibration on this sample. z against 1 = -1.05.from a recorded research file
0.974Betfair SP calibration slopeCox calibration slope of Betfair Starting Price implied probability against realised win frequency. 1.000 means the price is exactly right on average. Measured 0.9737 and 0.9741 by two independent re-runs; the programme's standing value is 0.9745. Published rounded to three decimals because the standard error (0.025) makes the fourth decimal meaningless.from a recorded research file
0.2%Betfair SP built-in marginMean overround on Betfair Starting Price: 1.0018, i.e. the implied probabilities in a race sum to about 100.2%. The 0.2% is the market's residual imbalance, not a charge; exchange commission is separate and is levied on winnings.from a recorded research file
1,997Races in the paired price comparisonRaces carrying both a Betfair Starting Price and a bookmaker Starting Price, compared against each other rather than against separate samples. Pairing is what makes the comparison a comparison.from a recorded research file
18,529Runners in the paired price comparisonRunner-race records within the 1,997 paired races.from a recorded research file
1.041 to 1.148Bookmaker SP calibration 95% confidence interval95% confidence interval for the bookmaker SP calibration slope. It EXCLUDES 1.000. This is the contrast that matters: on identical races the exchange price is indistinguishable from calibrated and the bookmaker price is measurably not.from a recorded research file
1.095Bookmaker SP calibration slopeThe same statistic computed on the same races for the bookmakers' Starting Price. Measured 1.0948 and 1.0952. z against 1 = +3.46.from a recorded research file
19.2%Bookmaker SP built-in marginMean overround on bookmaker Starting Price across the same races: 1.1917, i.e. implied probabilities sum to about 119.2%. That excess is the margin.from a recorded research file
Technical record
SourceARCHIVE_SWEEP_PANEL_2026-08-17 §2(a), independently re-run by two reviewers; paired on the same 1,997 races / 18,529 runners with proportional overround normalisation
As at2026-08-17
StatusGOVERNED_ARTEFACT
Held-out validation of a stand-alone draw rule, threshold fixed before the test (draw case study, Elite Pass research register)1 figure · as at 2026-07-07
477Held-out races in the stand-alone draw testRaces the stand-alone draw rule had never seen when it was reported, with the threshold fixed on development data before the test. The sample behind the draw case study's held-out null.derived from a recorded research file
Technical record
SourceHeld-out validation of a stand-alone draw rule, threshold fixed before the test (draw case study, Elite Pass research register)
As at2026-07-07
StatusDERIVED_FROM_ARTEFACT
Daily grades 2026-08-22 (recorded race day) + the evidence strength specification 1.0 Amendment A11 figures · as at 2026-08-22
1,024Races in the EP Score backtestUnique-leader races across 45 race days (2026-06-29 to 2026-08-22) used to test the score's band behaviour, with features taken only from pre-race files and outcomes joined separately. Higher bands won more often at their price, not better than it: actual over expected stayed at or below 1.0 in every band, so no edge is claimed.recomputed from source
57Races assessed on the example race dayEvery eligible race on the 2026-08-22 card entered the grading instrument. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.derived from a recorded research file
1Races reaching full convergenceRaces on that race day scoring in the 90 to 100 band, which requires the convergence component. Exactly one of 57. A retrospective example from one recorded race day, scored after the fact while the score was being validated. It is not a prospective read, not a prediction and not a claim of edge.recomputed from source
41Races published as NO READRaces on that race day where the evidence did not converge enough to earn a read. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.recomputed from source
16Races that qualified and were scoredRaces on that race day passing every qualification gate. The other 41 were published as NO READ, which is a result rather than a gap. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.recomputed from source
50Base points, race qualifiedFixed base awarded when a race passes the qualification gate: unique morning-rating leader, population reconciled, rating gap present. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.recomputed from source
15Convergence points in the exampleAwarded when both role-reading sources identify the same single leader and that horse also leads the morning rating. The two role sources are one source type, never independent confirmation. This is an early-stage measure. A retrospective example from one recorded race day, scored after the fact.recomputed from source
11Runners in the example raceDeclared runners in the Redcar 16:50 race on the recorded 2026-08-22 race day. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.derived from a recorded research file
90EP Score of the example raceEP SCORE 1.0 total for the Redcar 16:50 race on the recorded 2026-08-22 race day: base 50 (race qualified) + separation 15 + convergence 15 + structure and pace 10. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.recomputed from source
15Separation points in the exampleOne point per full rating point between the top-rated and second-rated runner, capped at 25. The example race carried a 15-point gap. A retrospective example from one recorded race day, scored after the fact while EP SCORE 1.0 was being validated. It is not a prospective read, not a prediction and not a claim of edge. Prospective pre-race reads begin with the live product.recomputed from source
10Structure and pace points in the exampleAwarded when the ratings structure is clear and pace opposition is none or partial. This is an early-stage measure. A retrospective example from one recorded race day, scored after the fact.recomputed from source
Technical record
SourceDAILY_GRADES_2026-08-22.json (sealed card-day artefact) + the evidence strength specification 1.0 Amendment A
As at2026-08-22
StatusDERIVED_FROM_ARTEFACT, RECOMPUTED
Favourite research report 1.0 (Elite Pass historical racing archive, recomputed 2026-08-21). Corpus 2008-09-03 to 2026-06-26.5 figures · as at 2026-08-21
0.997Favourites: actual wins against the market's expectationActual favourite wins divided by the wins implied by the favourites' own Betfair SP across the study (21215 actual against 21285.3 expected). A value near 1.000 means the outcomes tracked the prices; it is a statement about calibration, not about profit.derived from a recorded research file
65.0%Clear favourites that did not winThe complement of the favourite win rate over the same 60671 races. 'Did not win' includes second, last and every position between; it is not a claim about how they were beaten.derived from a recorded research file
35.0%Clear favourites that wonShare of the 60671 races with a single clear Betfair SP favourite in which that favourite won. Measured 0.34967. It is a WIN RATE, not a profit rate: the two are different questions and only the first is answered here.derived from a recorded research file
34.6% to 35.3%Favourite win rate, 95% confidence interval95% confidence interval for the favourite win rate on this sample (0.34589 to 0.35348). The interval is narrow because the sample is large; it describes sampling uncertainty and nothing else.derived from a recorded research file
300Races excluded for joint favouritesRaces where two or more runners shared the lowest Betfair SP. They are held out of the favourite study rather than resolved by a tie break, because choosing one of two equal prices would invent a favourite the market did not name.derived from a recorded research file
Technical record
SourceFAVOURITE_RESEARCH_REPORT_1.0 (Elite Pass historical racing archive, recomputed 2026-08-21). Corpus 2008-09-03 to 2026-06-26.
As at2026-08-21
StatusDERIVED_FROM_ARTEFACT
Benter Phase 2 results, out-of-sample section, 2026-07-141 figure · as at 2026-07-14
73Held-out races in the form-stack validationRaces the form stack had never seen when its out-of-sample likelihood was reported. The sample the published combined-model and market benchmark figures are computed on.from a recorded research file
Benter Phase 2 results, settled-data section, 2026-07-143 figures · as at 2026-07-14
8Race-days in the form-stack studyDistinct race-days covered by the form-stack study, 2 to 9 July 2026. The report states this is far below the sample its own power audit requires.from a recorded research file
317Settled races joined for the form-stack studyRaces settled and exactly joined for the form-stack study. The population the study was drawn from, not the sample its out-of-sample figures were computed on.from a recorded research file
2,828Settled runners joined for the form-stack studyRunners in the settled join behind the form-stack study. A population figure, never the inferential sample.from a recorded research file
Elite Pass historical racing archive (27 source exports)1 figure · as at 2026-08-21
71,626Races in the historical research corpusDistinct historical races held in the research corpus behind the programme's large-n studies: 71,626 races across 78 British and Irish courses, 2002-03-27 to 2026-06-28, keyed date|track|time. An inventory figure, what the archive holds. The convenience-corpus caveats travel with it: several exports truncated at 60,000 rows and the top five courses supply around 44% of complete-field races, so it is not a representative sweep of two decades of racing.recomputed from source
Technical record
SourceElite Pass historical racing archive (27 source exports)
As at2026-08-21
StatusRECOMPUTED
Place dataset to place accumulator verdict (completed study)1 figure · as at 2026-08-21
633,474Historical runner-race records analysedRunner-race records carried through the completed place-market study, the largest single analysis the programme has run. 633,474 rows, each a unique (date, track, horse) key, drawn from a research corpus of 71,626 races across 78 British and Irish courses, 2002-2026. The corpus is a convenience corpus, weighted toward all-weather and major courses. It is not a representative sweep of two decades of racing, and the figure is a count of records analysed, not of races.recomputed from source
Bookmaker benter console 2026-08-19 (dayclose/), stated verbatim as '2,772 races / 25,119 runners with full labelled T15 to T1 to BSP ladder (results workbooks)'2 figures · as at 2026-08-19
2,772Races with a full pre-off price ladderRaces carrying a complete labelled price ladder from T15 through T1 to Betfair SP, taken from the results workbooks. This is a SAMPLE DEFINITION, not a result: it says which races had a full ladder to measure, and nothing about what was found in them.from a recorded research file
25,119Runners with a full pre-off price ladderRunner records inside those races. Adjacent corpora exist (2,737 races / 25,140 runners) and are DISTINCT populations; External claim audit 2026-08-27 records that they must never be spliced, and test_site enforces it.from a recorded research file
Technical record
SourceBOOKMAKER_BENTER_CONSOLE_2026-08-19 (research/dayclose/), stated verbatim as '2,772 races / 25,119 runners with full labelled T15 to T1 to BSP ladder (results workbooks)'
As at2026-08-19
StatusGOVERNED_ARTEFACT
T-gate capture receipts (tgates_*)2 figures · as at 2026-08-20
7,261Pre-race market capturesTimestamped pre-off exchange captures across 1,180 races and 64,967 runner-price observations, 32 racing days, 2026-07-16 to 2026-08-20. Each capture keeps the full back/lay board, not one price. Genuine exchange observations, never Industry SP, forecast or tissue prices. Excludes the recently added two-minute checkpoint, which does not span the window.recomputed from source
1,180Races with a timestamped market receiptDistinct races for which Elite Pass holds its own timestamped pre-off exchange capture, a real back/lay board recorded at fixed checkpoints before the off. 32 racing days, 2026-07-16 to 2026-08-20, complete days only. These are genuine exchange observations, not Industry SP, forecast or tissue prices.recomputed from source
Technical record
SourceT-gate capture receipts (tgates_*.jsonl)
As at2026-08-20
StatusRECOMPUTED
Daily grades 2026-08-{06,07,16,17,19,20,21}1 figure · as at 2026-08-21
275Races assessed by the grading instrumentRaces the frozen grading instrument has assessed across 7 race days (6, 7, 16, 17, 19, 20 and 21 August 2026). It issued 72 AMBER and 203 No Read, and has never issued a GREEN. A grade applies to ONE runner in a race; it is not a ranking of the field, not a probability and not a bet.recomputed from source
Forensic audit 1.0 correction A3, de-vigged baseline, 30,088 races; frozen in programme b endpoint resolution 1.0 as the static rating-rank hypothesis, NULL beyond BSP1 figure · as at 2026-08-29
0.998Top-rated horses: actual winners against expectedAcross the 30,088-race study, Elite Pass's own top-rated horses won 0.998 times as often as a de-vigged Betfair Starting Price baseline implied they would. 1.000 is par, so this is APPROXIMATE PARITY: no meaningful aggregate over- or under-performance relative to price. BINDING LABEL: A/E is DESCRIPTIVE and is NOT a registered endpoint. It supports no edge claim in either direction, and parity must not be quoted as evidence against 'top-rated horses underperform' — it is the absence of a directional effect, not the reverse of one.from a recorded research file
Technical record
SourceFORENSIC_AUDIT_1.0 correction A3, de-vigged baseline, 30,088 races; frozen in PROGRAMME_B_ENDPOINT_RESOLUTION_1.0 as the static rating-rank hypothesis, NULL beyond BSP
As at2026-08-29
StatusGOVERNED_ARTEFACT
Forensic audit 1.0 (forensic audit 2026-08-29), the rating archive corpus; ruling programme b endpoint resolution 1.0, self-hash verified abacf1db…, frozen conclusion NULL beyond BSP2 figures · as at 2026-08-29
30,088Races in the ratings-vs-price studyRaces carrying a unique top-rated horse, a complete Betfair SP book, a field of at least four and a book overround between 0.85 and 1.60. Drawn from an archive of 725,869 unique runner-races spanning 2002-2026; because a complete book is required, BSP COVERAGE IN THAT ARCHIVE begins in August 2020 and the analysable window starts there. That is a statement about this archive's coverage, not about when Betfair SP itself began. This figure must never be described as covering 2002-2026. Split chronologically, trained before 2023 and held out from 2023; that boundary is also a coverage boundary, so the two periods are not exchangeable.from a recorded research file
298,444Runners in the ratings-vs-price studyThe runner panel within those 30,088 races, across 1,900 race-days. Each runner is scored against a de-vigged Betfair Starting Price baseline - the market's implied probability with the margin removed - rather than against chance.from a recorded research file
Place accumulator verdict (benchmark subset, n = 531,115)1 figure · as at 2026-08-21
531,115Runners benchmarked against the marketThe benchmark subset of the same completed study: runner-race records carrying a genuine Betfair exchange place price, each scored against the market's own implied probability rather than against nothing. The study was well-powered (minimum detectable effects 0.18-0.43pp) and returned a null: no structural predictor beat the price. That null is published, not hidden.from a recorded research file
Technical record
SourcePLACE_ACCUMULATOR_VERDICT (benchmark subset, n = 531,115)
As at2026-08-21
StatusGOVERNED_ARTEFACT
GASA backtest report, section 1 (data and power), 2026-07-053 figures · as at 2026-07-05
9Race-days in the market-movement ladder setDistinct race-days in the ladder set. A single short window with no true walk-forward, which is why single significant readings in this lane are treated as noise until replicated.from a recorded research file
335Races in the market-movement ladder setThe only races in the market-movement corpus that carry a full pre-off price ladder, and therefore the ceiling on every steam hypothesis. The wider 160,989-row corpus cannot test a steam rule because it has no ladder.from a recorded research file
2,962Runner-rows in the market-movement ladder setRunner rows in the clean, leakage-firewalled ladder build behind the market-movement case. Not the corpus: the corpus is 160,989 rows, of which only these carry the ladder the test needs.from a recorded research file
Technical record
SourceGASA backtest report, section 1 (data and power), 2026-07-05
As at2026-07-05
StatusGOVERNED_ARTEFACT
GASA backtest report, finding F5, 2026-07-051 figure · as at 2026-07-05
61Market overtakes behind the overtaker comparisonCases where a horse overtook the favourite in the market late. The sample the published −12.9 percentage-point comparison is computed on, not the corpus.from a recorded research file
GASA backtest report, structural arm, 2026-07-051 figure · as at 2026-07-05
4Structure-aligned chances behind the 0 of 4 figureClean structure-aligned overtake opportunities in the graded set. The source states this as approximately four within an eleven-race graded set and calls the arm anecdotal; it is not powered evidence and may not be described as settled.from a recorded research file
Supplied by the founder2 figures · as at 2026-08-22
227Systems, rules & racing theories examinedRacing systems, theories and hypotheses examined by the founder over the life of the programme, including work predating and sitting outside this repository. Held on operator authority alone.confirmed by the founder
HundredsRacing systems, rules & hypotheses examinedThe founder's own account of the body of work, stated non-exactly and deliberately. 'Examined' includes ideas that failed outright, proved market-priced, became conditional, or remain unresolved. It does NOT mean proven false. No exact figure is published because no register supporting one exists.confirmed by the founder
Technical record
SourceOPERATOR_SUPPLIED
As at2026-08-22
StatusOPERATOR_CONFIRMED
Promotion protocol + docket + repository sweep1 figure · as at 2026-08-21
0Systems that have cleared our validation barTested ideas that have reached a validated state, the only state that would grant decision authority. Zero, all-time. Confirmed five independent ways: no promotion git tags exist; all promotion-docket entries sit at the entry stage with no preregistration filed; the global decision-authority constant reads no validated edge claimed; the hypothesis registers carry no validated row; and research capital remains GBP 0.00.recomputed from source
Supplied by the founder, from outside the research repository1 figure · as at 2026-08-22
1Figures held but not publishedCounts the operator holds on their own authority, for which no register in this repository can evidence an exact value. They are used internally and deliberately not published as exact figures.withheld pending validation