31.66% 命中率，為何仍會累積虧損和長連敗？How can a 31.66% hit rate still end in losses and a long losing run?31.66% 命中率，为何仍会累积亏损和长连败？
在 1,712 場歷史賽事中，每場以一單注平均分配至獨贏大熱門組合，542 場命中，命中率為 31.66%；但整段紀錄仍為 −260 注，最長連續 15 場未命中，最大損失 271 注。這不是投注策略或 Sigma 模型成績，而是用市場大熱作為基準：命中率不能單獨描述價格、波動和可承受風險。Across 1,712 historical races, a one-unit stake split equally across the win-favourite set hit in 542 races, a 31.66% hit rate. The full sequence still finished at −260 units, included a 15-race losing run and reached a maximum drawdown of 271 units. This is not a strategy or Sigma model result. It is a baseline showing that hit rate alone says little about price, volatility or survivable risk.在 1,712 场历史赛事中，每场以一个单位平均分配至独赢大热门组合，542 场命中，命中率为 31.66%；但整段纪录仍为 −260 单位，最长连续 15 场未命中，最大回撤 271 单位。这不是投注策略或 Sigma 模型成绩，而是一个基准：命中率不能单独描述价格、波动和可承受风险。
Start with the full baseline, not the last three races
This historical baseline covers 1,712 races from 10 July 2024 to 8 July 2026. One unit is allocated per race; when favourites are tied, that unit is split equally across the favourite set. Realised returns use the recorded win dividend and preserve the dead-heat adjustment in the source data.
The favourite set hit in 542 races but finished at −260 units, a simplified historical return of −15.19%. The sequence briefly reached +6.4 units before falling below zero. Selecting only that early stretch—or only the best months—would tell a completely different story.
Fifteen losses do not make race sixteen due
Even relatively independent outcomes naturally form clusters. Racing is more irregular than a fair coin because each race has different probabilities and prices. Previous losses do not create compensation in the next race. The estimate should change only if new information changes the chance in that next event.
The baseline contained a longest run of 15 races without a hit. A losing run can coexist with a reasonable long-term estimate. It neither proves that the next event must win nor, by itself, proves that a model has failed.
Evaluate a method on the complete sequence
Save the model version, probability, price timestamp and usage rule before the race, then evaluate the complete sample. Report at least sample size, cumulative return, maximum drawdown, longest losing run and whether the rule was fixed before seeing results—not only the hit rate.
Maximum drawdown in this baseline reached 271 units, far larger than the early peak of +6.4. Drawdown makes the worst historical fall from a previous high visible, but historical drawdown is not a ceiling on future loss.
Historical favourite-baseline summary
Races
Hits
Hit rate
Cumulative result
Longest losing run
Maximum drawdown
1,712
542
31.66%
−260 units
15 races
271 units
One unit per race, split equally across tied win favourites. Historical descriptive baseline, not a recommended strategy.
The favourite set won 31.66% of races and still finished −260 units
Historical baseline · not a strategy
One unit per race, split equally across tied win favourites, over 1,712 historical races. The line shows the complete sequence—not a selected month.
Races
1,712
10 Jul 2024–8 Jul 2026
Hit rate
31.66%
542 races
Longest losing run
15
consecutive races
Historical return
-15.19%
−260 units
Maximum drawdown
271
units
Peak+6.4 units
Finish−260 units
View the baseline summary table
The favourite set won 31.66% of races and still finished −260 units
Measure
Value
Meaning
Hit rate
31.66%
Favourite set won in 542 of 1,712 races
Longest losing run
15
Consecutive races without a favourite-set win
Cumulative result
−260 units
One unit allocated per race
Maximum drawdown
271 units
Largest historical fall from a prior peak
Set a firm risk boundary first: EV is a screening principle, not a profit guarantee
Any cap should be set before participation and must not compromise living costs, debt payments, savings or family responsibilities. Stop when the cap is reached. Increasing stakes or borrowing after losses does not improve the original probability; it only accelerates potential harm.
Model probabilities, EV and backtests can describe uncertainty, but none can guarantee income. Estimation error, market change, execution differences and extreme sequences can all make future results depart from history.
Responsible use and support
Sigma Quant provides analytics and education only and does not accept or process bets. Follow local law, be 18 or older, use a limit that cannot affect daily life and never chase losses.
If gambling is affecting finances, emotions, work or relationships, stop and seek support. Hong Kong Ping Wo Fund gambling counselling hotline: 1834 633.
