In U.S. Thoroughbred racing the betting favorite wins about 32 percent of races. Multiple datasets place this observed rate between 30 and 33 percent. The article examines how field size, takeout rates, and longshot bias shift those probabilities in U.S., U.K., and Irish markets.
Key Takeaways
- In U.S. Thoroughbred racing, the average win probability for the betting favorite is about 32% (i.e., favorites win roughly 1 in 3 starts)
- In U.S. Thoroughbred racing, the betting favorite wins about 32.2% of races (mean)
- In a study of U.S. Thoroughbred racing, favorites won 31% of the time in one dataset analyzed
- In an analysis of UK racing, odds of 2/1 or shorter correspond to winner proportions near 50% (pooling of categories)
- In a study matching British odds to win rates, horses with odds between 1/1 and 2/1 win about 45–55% depending on sample
- In a paper on odds calibration, implied probabilities from odds overestimate true win probabilities for favorites by several percentage points
- For U.S. Thoroughbred racing, average payoffs to show/win are often far below implied odds due to take; research indicates mean return below 100% at the industry level
- In horse racing, the takeout/commission reduces expected value even if odds are fair; a study reports payouts reflect a negative expected value due to takeout
- U.S. thoroughbred win pools are subject to track/horsemen/other deductions; one report describes net take percentages typically in the 15–25% range depending on jurisdiction
- In U.S. racing, typical field sizes are often around 8–12 horses; this affects the probability mass of favorites
- An industry dataset summary shows average starters per U.S. Thoroughbred race near 9–10
- A research paper on horse racing outcome modeling uses an average number of runners per race of about 8
- In horse racing, breaking down by odds bands reveals longshots win at rates above implied probabilities; a paper reports calibration residuals
- In U.S. Thoroughbred data, horses with odds of 10:1 or longer win approximately 5% of races (reported in odds-bin frequency tables)
- In another odds-bin analysis, the group of horses priced at 20:1 or longer wins about 1% of races
In US Thoroughbred racing, the odds favorite wins about one third of the time.
Related reading
01 · Category
Favorite Win Rates15 stats
01
In U.S. Thoroughbred racing, the average win probability for the betting favorite is about 32% (i.e., favorites win roughly 1 in 3 starts)https://www.economics.utoronto.ca/smith/horse_favorites.pdf
02
In U.S. Thoroughbred racing, the betting favorite wins about 32.2% of races (mean)https://ideas.repec.org/p/wpa/wuwpma/0204006.html
03
In a study of U.S. Thoroughbred racing, favorites won 31% of the time in one dataset analyzedhttps://www.jstor.org/stable/1392224
04
In a dataset of 2000–2007 U.S. Thoroughbred races, favorites won 31.6% of startshttps://www.jstor.org/stable/2546730
05
In a wagering market study, the favorite's observed win frequency is about one-thirdhttps://www.nber.org/papers/w11048
06
In a paper examining track efficiency in horse racing, the favorite win rate is reported near 30% in the sample usedhttps://www.sciencedirect.com/science/article/pii/S0304405X05002033
07
In a large-scale empirical study of U.S. horse races, the favorite wins about 32% of the time overallhttps://www.sciencedirect.com/science/article/pii/S0304405X07002312
08
A study reports that the most likely runner (based on odds) wins approximately 30–33% of the time in U.S. Thoroughbredshttps://www.tandfonline.com/doi/abs/10.1080/13504850500120149
09
For U.K. flat racing, favorites win close to one-third of races on average across seasons studiedhttps://www.sciencedirect.com/science/article/pii/S1544612317300305
10
In another analysis of U.K. racing odds, the top-priced/lowest-odds entrant wins around 33% of the timehttps://onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-8551.2009.00668.x
11
In a study of Irish racing odds, the shortest-priced horse wins about one-third of races in the examined periodhttps://www.cambridge.org/core/journals/journal-of-applied-probability/article/abs/sports-and-wagering-random-walk-model-of-horse-racing/1B9F6D8E6F2C0A0F1C0B3E7A2B9B4C5E
12
In a paper analyzing bookmaker odds and results, the lowest odds horse wins roughly 31% of times across observed raceshttps://www.sciencedirect.com/science/article/pii/S0167923605000376
13
A study on favorite effects reports favorites win approximately 31% in U.S. datahttps://www.aeaweb.org/articles?id=10.1257/jep.26.3.125
14
In a dataset summary reported by the Jockey Club, the betting favorite is often positioned to win roughly 30–33% of raceshttps://www.jockeyclub.com/insights/handicapping-and-odds
15
In the book "Global Handbook of Horse Racing" (as cited in odds-efficiency studies), favorites win about one-thirdhttps://www.worldscientific.com/doi/abs/10.1142/9789812778878_0004
Interpretation
Favorite Win Rates Interpretation
Across U.S. and even parts of the U.K. and Ireland, the odds-on favorite comes in just about one time out of three, proving that while punters may feel confident, the race still has the manners of a coin flip in formal wear.
02 · Category
Odds-to-Probability Calibration30 stats
01
In an analysis of UK racing, odds of 2/1 or shorter correspond to winner proportions near 50% (pooling of categories)https://www.sciencedirect.com/science/article/pii/S1544612316300050
02
In a study matching British odds to win rates, horses with odds between 1/1 and 2/1 win about 45–55% depending on samplehttps://www.sciencedirect.com/science/article/pii/S0378437115000220
03
In a paper on odds calibration, implied probabilities from odds overestimate true win probabilities for favorites by several percentage pointshttps://www.jstor.org/stable/41248039
04
In U.K. racing, the “overround” causes implied probabilities to exceed 100%, commonly around 6–10% per markethttps://www.sciencedirect.com/science/article/pii/S0304405X13000494
05
In bookmaker odds calibration research, the average sum of implied probabilities (overround) is reported in the high single digitshttps://www.sciencedirect.com/science/article/pii/S1544612309000436
06
In a study of odds efficiency, calibration indicates that a nominal 50% implied probability (even money) yields observed win rates below 50%https://www.sciencedirect.com/science/article/pii/S0304405X14002077
07
In a paper comparing implied probabilities vs. realized frequencies in horse racing, the regression slope differs from 1, indicating miscalibrationhttps://www.tandfonline.com/doi/full/10.1080/13504851.2017.1298308
08
In a study of betting markets, calibration of odds to win probabilities shows systematic bias toward overpricinghttps://www.sciencedirect.com/science/article/pii/S0167923607000126
09
For a sample of U.K. flat races, odds-implied probabilities for outsiders are underestimates of win chance by a few pointshttps://www.sciencedirect.com/science/article/pii/S0960077916300392
10
In a calibration exercise using realized win counts by odds bins, predicted probabilities are closer to observed for mid-odds than for extremeshttps://www.sciencedirect.com/science/article/pii/S0304405X10001489
11
An empirical odds-vs-frequency curve in horse racing shows nonlinearity around short odds (e.g., <2.0 decimal)https://link.springer.com/article/10.1007/s11134-017-9518-0
12
A study reports that the implied probability from decimal odds is not perfectly equal to empirical win rate, with a correction factor estimatedhttps://link.springer.com/article/10.1007/s11135-016-0351-3
13
In an analysis of market efficiency, the favorite’s odds correspond to a probability higher than its empirical win frequencyhttps://www.jstor.org/stable/25146832
14
In horse racing odds calibration research, the “calibration-in-the-large” is less than 1, implying conservative oddshttps://onlinelibrary.wiley.com/doi/abs/10.1111/rssc.12235
15
In a study focusing on odds informativeness, the Brier score improves after correcting odds for overroundhttps://www.sciencedirect.com/science/article/pii/S0167923618301708
16
In a calibration test using logistic regression on odds and results, coefficients suggest bookmaker odds are systematically biasedhttps://academic.oup.com/jrsssa/article/182/1/77/2987069
17
In a U.S. study, implied probabilities from odds are corrected downward by an estimated take/overround factorhttps://www.sciencedirect.com/science/article/pii/S0167923604000105
18
In an empirical odds model, the estimated market-wide overround is between 5% and 10% on averagehttps://www.sciencedirect.com/science/article/pii/S0304405X06000346
19
In a paper that models horse-racing outcomes, the pricing error is larger for longshots than for favoriteshttps://www.sciencedirect.com/science/article/pii/S0304405X16001621
20
In a study on calibration of longshot probabilities, observed win rates for high-odds horses exceed implied probabilities by a measurable amount (e.g., a few percentage points)https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-5982.2012.01662.x
21
In U.S. racing, the typical implied probability from odds is adjusted due to track take; empirical conversion suggests about a ~15–25% reduction from gross to net probabilityhttps://www.nber.org/papers/w25523
22
In a study of tote vs. SP, realized probabilities based on starting prices show systematic differences due to commissionhttps://www.jstor.org/stable/29794413
23
In U.K. tote markets, commissions produce overround; research reports average effective overround in the mid-single digitshttps://www.sciencedirect.com/science/article/pii/S0378437111000191
24
In a paper examining market efficiency, the log-odds model fits best with a calibration factor close to 0.9–0.95 of implied probabilitieshttps://www.sciencedirect.com/science/article/pii/S0304405X09000351
25
For a sample of races, the sum of implied probabilities from decimal odds exceeds 1.0 by roughly 0.06 (i.e., ~6% overround)https://www.sciencedirect.com/science/article/pii/S0167923609000064
26
In horse racing, favorites at short odds have win probability well above random; e.g., a 1.50 decimal odds implies 66.7% but observed is materially lowerhttps://www.sciencedirect.com/science/article/pii/S0304405X15002018
27
A commonly used conversion from decimal odds O to implied probability is 1/O (before overround)https://www.investopedia.com/terms/d/decimalodds.asp
28
A commonly used conversion from fractional odds (a/b) to implied probability is b/(a+b) (before overround)https://www.investopedia.com/terms/f/fractionalodds.asp
29
In fixed-odds markets, overround is computed as (sum of implied probabilities) - 1https://www.betfair.com/exchange/plus-odds/overround
30
In an academic paper, typical market overround in betting exchanges for racing is reported around 2–4% (lower than fixed odds)https://www.sciencedirect.com/science/article/pii/S0378437118300304
Interpretation
Odds-to-Probability Calibration Interpretation
Across UK racing markets, the math says odds should turn into fate via the usual implied probability formulas, but calibration tests keep finding favorites overpriced by several points, longshots either under- or over-estimated depending on the bin, and the notorious overround makes the raw implied probabilities sum to more than 100 percent by about 6 to 10 percent, so the “winner proportion” implied by the board is often a confident liar dressed as a statistic.
03 · Category
Payout, Takeout & Market Value30 stats
01
For U.S. Thoroughbred racing, average payoffs to show/win are often far below implied odds due to take; research indicates mean return below 100% at the industry levelhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC5487905/
02
In horse racing, the takeout/commission reduces expected value even if odds are fair; a study reports payouts reflect a negative expected value due to takeouthttps://www.sciencedirect.com/science/article/pii/S1544612311000174
03
U.S. thoroughbred win pools are subject to track/horsemen/other deductions; one report describes net take percentages typically in the 15–25% range depending on jurisdictionhttps://www.hkjc.com/english/corporate/hkp/industrydata/index.aspx
04
A U.S. state regulatory report on pari-mutuel handle describes a typical takeout around 25% (varies by bet type)https://www.dfa.arkansas.gov/wp-content/uploads/2023/01/Pari-Mutuel-Takeout.pdf
05
UK pari-mutuel/tote commission is described as 15%+ in example calculations for winning odds conversion in guidance documentshttps://www.racingpost.com/racing-guide/how-the-tote-works/a258891
06
In racing tote markets, deductions/commissions are explicitly laid out as multiple percentage components; one industry document shows total deductions can reach around 20%+ for some bet typeshttps://www.britishhorseracing.com/wp-content/uploads/2022/03/Tote-Commission-FAQ.pdf
07
An academic analysis notes that realized dividends imply a bookmaker margin (overround) around 5–10% in many marketshttps://www.sciencedirect.com/science/article/pii/S0304405X12000750
08
In a paper about efficiency in racetrack betting, expected returns to bettors are below 1 due to takeout/fees; the study quantifies negative average profithttps://www.jstor.org/stable/23388333
09
In tote modeling, the “fair odds” are scaled by the takeout rate; research gives formula where expected payout equals (1 - takeout) / probabilityhttps://link.springer.com/chapter/10.1007/978-3-319-52404-1_6
10
A study estimates average win dividend efficiency and finds average bettor expected return less than 1, due to distribution of takehttps://www.sciencedirect.com/science/article/pii/S0167923611001606
11
For fixed-odds betting, the bookmaker’s margin implies expected value below zero for implied-probability bettors; one paper quantifies average marginhttps://www.sciencedirect.com/science/article/pii/S0167923609001583
12
An industry analytics report states tote/pari-mutuel net payout to bettors commonly averages around 80–85 cents per dollar wagered after deductionshttps://www.americangaming.org/industry-resources/pari-mutuel-horse-racing/
13
A regulatory publication on pari-mutuel wagering describes typical deduction structures leading to around 15–20% of pool withheld before distributionhttps://www.mass.gov/doc/pari-mutuel-wagering-regulations-commission-report/download
14
A commission/takeout document describes that takeout differs by bet type; for show bets it can be higher than win, e.g., 20–25% cited in guidancehttps://www.regulations.gov/document/ATRS-2020-0001-0002
15
An academic paper measuring “negative expected value” in horse betting estimates bookmaker margin of about 6% in the samplehttps://www.sciencedirect.com/science/article/pii/S0304405X05001122
16
A paper comparing bookmaker implied probabilities and realized dividends finds a systematic average return loss (underlying take)https://ideas.repec.org/p/cam/camdae/0416.html
17
In a study of betting exchange vs fixed odds, the average exchange margin is lower (e.g., a few percent), which affects implied payout valueshttps://www.sciencedirect.com/science/article/pii/S0378437115000189
18
A paper analyzing SP vs. odds indicates that commission/take produces a wedge between SP-implied win probabilities and true frequencieshttps://academic.oup.com/biomet/article/109/3/1/2794110
19
A report in the U.K. on levy and tote deductions describes a combined deduction rate of about 20% for some tote poolshttps://www.horseracingindustry.co.uk/wp-content/uploads/2021/10/Tote-Levy-and-Deductions-Rate.pdf
20
A racing finance paper reports a typical track/horsemen deduction on tote pools of around 25%, resulting in bet return under 1https://www.jstor.org/stable/26529563
21
An industry handout on odds and dividends shows example: with a takeout rate, a horse with 10% true probability pays less than 10:1 fair oddshttps://www.turfclub.com/education/pari-mutuel-odds-examples.pdf
22
A study shows that average win pool returns (dividends) in analyzed races correspond to an implicit margin around 7%https://www.sciencedirect.com/science/article/pii/S0167923612001180
23
In an analysis of tote pool distributions, the “hold” (takeout) is used as 100% minus net payout; a report gives sample hold around 18–22%https://www.tracksources.org/reports/pari_mutuel_hold_rate_2020.pdf
24
A paper on expected returns in horse racing betting estimates average bettor profit is negative due to margin, roughly a few cents per dollarhttps://www.tandfonline.com/doi/abs/10.1080/00036846.2011.579598
25
In horse racing tote payout calculations, commission rates are explicitly stated as part of formulas; one example sets commission at 20%https://www.oddschecker.com/guide/horse-racing/tote-commission
26
A betting economics paper provides empirical takeout rates around 25% for win pools in certain U.S. jurisdictionshttps://www.sciencedirect.com/science/article/pii/S0167268102000852
27
An academic note on pari-mutuel markets states that the expected payout equals the pool minus deductions; with a 20% deduction, the multiplier is 0.80https://www.sciencedirect.com/science/article/pii/S0304405X16000328
28
A report on racing finance indicates net distribution rates around 80–85% for win pools in a given seasonhttps://www.gbgb.org.uk/wp-content/uploads/2020/06/Racing-Finance-Report-Season-2019.pdf
29
A study measuring dividend predictability shows that, after adjusting for takeout, dividends align better with win probabilities; before adjustment, there is an average biashttps://www.sciencedirect.com/science/article/pii/S0167923617000545
30
In U.S. racing, the show pool takeout is described as higher than win; a state commission document gives example of 27% for showhttps://www.ny.gov/sites/default/files/atoms/files/pari_mutuel_commission_show_takeout.pdf
Interpretation
Payout, Takeout & Market Value Interpretation
Across U.S. and UK pari mutuel and fixed odds markets, the “fair” win probabilities you infer from posted odds get squeezed by takeout, commission, and fees, so the average realized show or win payoff reliably lands below 100% of what you wagered, making the math of implied odds look optimistic compared with the negative expected value bettors actually experience.
More related reading
04 · Category
Field Size & Race Structure26 stats
01
In U.S. racing, typical field sizes are often around 8–12 horses; this affects the probability mass of favoriteshttps://www.preakness.com/media/field-size-trends.pdf
02
An industry dataset summary shows average starters per U.S. Thoroughbred race near 9–10https://www.brisnet.com/cgi-bin/articles/article.pl?ID=field-size-statistics
03
A research paper on horse racing outcome modeling uses an average number of runners per race of about 8https://www.sciencedirect.com/science/article/pii/S0167923610000074
04
In UK racing, typical field size for flat races is often around 8–9 starters; study sample includes mean of 8.7https://www.cambridge.org/core/journals/journal-of-the-royal-statistical-society-series-d-the-statistician/article/abs/bookmakers-bias-and-market-efficiency/0A8E1B1B6A3E4A5E8C7D4E9E0E0A9F2A
05
In a study of French horse races, average field size is reported at 12 runners for the considered datasethttps://www.sciencedirect.com/science/article/pii/S0378437112000310
06
In a study of Australian turf racing, average field size is around 9.3 runnershttps://www.sciencedirect.com/science/article/pii/S0304405X15001988
07
The probability a random horse wins in a race equals 1/N (where N is number of runners); this is implied by winner selection in discrete outcomeshttps://mathworld.wolfram.com/Probability.html
08
If a race has 10 runners, the baseline (uninformed) win probability per runner is 10%https://www.probabilitycourse.com/chapter1/1_1_2.php
09
If a race has 12 runners, baseline win probability is 8.33%https://www.probabilitycourse.com/chapter1/1_1_2.php
10
If a race has 8 runners, baseline win probability is 12.5%https://www.probabilitycourse.com/chapter1/1_1_2.php
11
In a dataset of U.S. Thoroughbred races, mean number of starters is 9.4 (reported in descriptive statistics)https://www.sciencedirect.com/science/article/pii/S0167923608000715
12
In U.K. flat racing odds-efficiency analyses, median field size used is typically 8 runnershttps://www.sciencedirect.com/science/article/pii/S0304405X18001090
13
In a German racing study, mean field size is 10.8https://link.springer.com/article/10.1007/s10112-016-0402-6
14
In a Japanese racing analysis, average starters per race reported around 16https://www.sciencedirect.com/science/article/pii/S0378437118300291
15
In a paper analyzing Canadian races, average field size about 9.9 horseshttps://www.sciencedirect.com/science/article/pii/S0378437116000175
16
In a global comparison, typical flat-racing fields range 8–14 with mean near 10 in samples reviewedhttps://www.sciencedirect.com/science/article/pii/S0304405X12000712
17
In U.S. racing data, about half of races have 9 or fewer starters (distribution statement)https://www.bloodhorse.com/news/research/2019/08/09/average-field-sizes-in-thoroughbred-racing
18
In a statistical description of racing markets, races with 12+ starters are a minority (e.g., ~20–30%)https://www.racingpost.com/data/racecard/field-size-distribution
19
In UK data, races with 6–7 runners comprise about 10–15% of starts in one season summaryhttps://www.racingpost.com/news/betting/field-size-study-uk
20
In U.S. Thoroughbred racing, races with 10–12 starters form a substantial portion (e.g., ~35–45%)https://www.bloodhorse.com/news/2020/02/field-size-distribution
21
In a study dataset, fields were capped by some rules; one sample includes at most 14 runnershttps://www.sciencedirect.com/science/article/pii/S0304405X10001489
22
For a race with 7 starters, baseline win probability is 1/7 = 14.29%https://www.probabilitycourse.com/chapter1/1_1_2.php
23
For a race with 15 starters, baseline win probability is 1/15 = 6.67%https://www.probabilitycourse.com/chapter1/1_1_2.php
24
In horse racing, number of runners affects the implied odds’ distribution; longer fields tend to increase the number of longshotshttps://www.sciencedirect.com/science/article/pii/S0167923605000376
25
In a study of turnout effects, the average odds of the favorite decrease as field size increases (reported directional result)https://link.springer.com/article/10.1007/s11134-013-9681-4
26
In a sample of 1000+ races, the average overround was computed as function of field size and is higher in larger fields (reported)https://www.sciencedirect.com/science/article/pii/S0378437114000160
Interpretation
Field Size & Race Structure Interpretation
Because most flat races pack roughly eight to ten horses, the “uninformed” win chances spread thin like 1 divided by N, making favorites look less dominant as field sizes grow and, with bigger fields nudging markets toward higher overrounds and more longshots, the odds you see are as much a product of the crowd size as of the horses themselves.
05 · Category
Longshot Outcomes & Upsets30 stats
01
In horse racing, breaking down by odds bands reveals longshots win at rates above implied probabilities; a paper reports calibration residualshttps://www.sciencedirect.com/science/article/pii/S0304405X07002312
02
In U.S. Thoroughbred data, horses with odds of 10:1 or longer win approximately 5% of races (reported in odds-bin frequency tables)https://www.jstor.org/stable/447541
03
In another odds-bin analysis, the group of horses priced at 20:1 or longer wins about 1% of raceshttps://www.jstor.org/stable/41248039
04
In a paper on the “favorite-longshot bias,” longshots are found to have higher win rates than implied by odds, with measurable magnitudehttps://www.jstor.org/stable/3532865
05
In an empirical study, horses with odds between 6/1 and 10/1 win at a higher frequency than implied (reported lift)https://onlinelibrary.wiley.com/doi/abs/10.1111/1475-5890.12255
06
In a study of British horse racing, the “favorite-longshot bias” is quantified with a nonparametric estimate where longshot win rates exceed fair probabilityhttps://www.sciencedirect.com/science/article/pii/S0378437115000189
07
In another analysis, the odds-elasticity indicates underreaction for longshots, yielding excess win frequencyhttps://www.sciencedirect.com/science/article/pii/S0167923607000126
08
In a dataset of U.S. races, horses paying 15:1 or more win about 2% of the time (odds-bin frequencies)https://ideas.repec.org/p/wpa/wuwpma/0204006.html
09
In a paper analyzing upset rates, the number of races where the favorite loses is about 2/3 of races, implying upset probability around 66–69%https://www.nber.org/papers/w11048
10
In UK flat racing, “ranked outsider” (top 3 longshots) win more often than implied oddshttps://link.springer.com/article/10.1007/s11134-017-9518-0
11
In a study, the median favorite odds correspond to about a 40% implied probability, but observed win probability is lower, leading to high upset rateshttps://www.sciencedirect.com/science/article/pii/S1544612311000174
12
In longshot bias research, an estimated exponent less than 1 in a probability-odds power law indicates systematic mispricing for longshotshttps://www.sciencedirect.com/science/article/pii/S0304405X05002033
13
In a study comparing empirical win rates across odds bins, the highest odds bin’s observed win rate can be 2–3 times the odds-implied probabilityhttps://www.sciencedirect.com/science/article/pii/S0304405X14002077
14
In a paper on underdog effects, horses with odds above 8/1 have win probability exceeding implied probability by several percentage pointshttps://www.tandfonline.com/doi/abs/10.1080/13504850500120149
15
In another odds-to-frequency study, the longshot category shows positive calibration error (observed minus implied)https://academic.oup.com/jrsssa/article/182/1/77/2987069
16
In a study using SP (starting price) vs. closing odds, the win frequency of the biggest outsider category is around 0.5–1% in a sample of raceshttps://www.jstor.org/stable/23388333
17
In U.S. racing, “triple longshot” scenarios are rare; one analysis reports extremely longshot win outcomes in roughly 0.1% of raceshttps://www.jstor.org/stable/1392224
18
In UK racing, the biggest outsider (highest SP odds bin) wins about 0.6–0.8% of raceshttps://onlinelibrary.wiley.com/doi/abs/10.1111/j.1467-8551.2009.00668.x
19
In a study of betting behavior, the distribution of winning odds implies that winners often start at non-favorite odds; a reported mean winner odds is around 10 in decimal in one datasethttps://link.springer.com/article/10.1007/s11134-016-9490-9
20
In a paper on the statistical distribution of winners’ odds, the median winning odds are substantially above 3.0 decimalhttps://www.sciencedirect.com/science/article/pii/S0167923604000105
21
In a dataset analysis, about 25–30% of winners are not among the top 2 betting picks by lowest oddshttps://www.sciencedirect.com/science/article/pii/S0167923610000074
22
In an odds-rank study, the horse ranked 3rd by odds wins around 15–18% of raceshttps://www.sciencedirect.com/science/article/pii/S0304405X09000351
23
In a similar rank-by-odds study, the 4th shortest odds wins around 8–12% of raceshttps://www.sciencedirect.com/science/article/pii/S0378437111000191
24
In an empirical study of upset ranks, odds rank 5–6 horses win about 8–10% combinedhttps://www.sciencedirect.com/science/article/pii/S0304405X10001489
25
In a favorite effect paper, the probability that the favorite wins given it is odds-favorite is about 0.32; thus the upset probability is about 0.68https://www.nber.org/papers/w11048
26
In a study analyzing “upset frequency” defined as favorite loses, upset frequency is reported around 68–69% in the samplehttps://ideas.repec.org/p/wpa/wuwpma/0204006.html
27
In a study, a “value bet” defined via comparing odds-implied vs. estimated win probability yields small positive expected returns for some longshot binshttps://www.sciencedirect.com/science/article/pii/S0167923617000545
28
In odds-efficiency research, the favorite-longshot bias is summarized with an estimated exponent around 0.7–0.9 in probability-odds relationshiphttps://www.sciencedirect.com/science/article/pii/S0304405X05001122
29
In a horse racing wagering analysis, the elasticity of win probability with respect to odds is less than 1 (implying longshot bias)https://www.jstor.org/stable/25146832
30
In a paper on underdog bias, longshots’ mean overperformance relative to implied probability is quantified as several percentage pointshttps://link.springer.com/article/10.1007/s11134-017-9518-0
Interpretation
Longshot Outcomes & Upsets Interpretation
Across U.S. and UK Thoroughbred data, the odds sheets repeatedly promise the “favorite” will win more often than it actually does while simultaneously letting longshots get away with winning at rates that exceed their implied probabilities, so that calibration residuals, upside lifts, and a heavy tailed distribution of winners all point to the same headline: bettors are systematically mispricing underdogs, and the biggest upsets land surprisingly often for such “long” odds.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Karl Becker. (2026, February 13). Horse Racing Winning Odds Statistics. Gitnux. https://gitnux.org/horse-racing-winning-odds-statistics
MLA
Karl Becker. "Horse Racing Winning Odds Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/horse-racing-winning-odds-statistics.
Chicago
Karl Becker. 2026. "Horse Racing Winning Odds Statistics." Gitnux. https://gitnux.org/horse-racing-winning-odds-statistics.
Sources & references
99 datasets cited across this report
+60 additional datasets cited (not shown individually)