Horse Racing Losing Runs: What Real Streaks Look Like Across 161,236 UK Races

You’ve got a method that’s meant to win about one bet in three. You hit a run of eight losers. It feels wrong… so you tweak a filter, or drop it, and the next winner turns up the week after.

I’ve seen that happen enough to want real numbers on it. So how long is a normal losing run? Across 161,236 UK races, backing the Betfair favourite every time, the longest was 28 losers in a row, and about one losing run in 50 reached 10. Eight in a row was ordinary. A run of losers on its own can’t tell you your method is broken. Whether the odds have stopped paying you is a different question, and I’ll come to it.

The short version

• In these samples, the lower the strike rate (how often you win), the longer the losing runs.
• Even the favourite, the most stable bet in racing, had a run of 28 losers in 13 years.
• Use the streak table below to work out your stake before you start, instead of after the bad run arrives.

Why do people quit inside a normal losing run?

Ask someone what a good strike rate is and 30% is a common answer. Turn that round and it means seven bets in ten lose.

What catches people out is how streaky that average is. Losing seven out of ten isn’t seven losses then three wins. It’s clusters: a run of two, a win, a run of nine, a win, a run of one. The clusters are where I think people quit, because that’s when the method looks like it’s stopped working. A run of losers is what a 30% strike rate produces, so by itself it can’t tell you the method has changed.

What does a normal losing run look like at a 34% strike rate?

Take the strategy with the highest strike rate in my study: back every Betfair SP favourite. Across 161,236 races from July 2013 to March 2026 it wins 34.4% of the time. Its A/E ratio, which compares actual winners with the winners the odds said to expect, is 1.00. So the market priced it fairly and there’s no edge, but the strike rate is about as stable as racing gets.

Here’s how its losing runs were spread. A percentile is just a ranking point: the 90th percentile (P90) is the length that 90% of losing runs stayed at or under.

About 2.3% of its losing runs reached 10 or more. So eight losers in a row sits right at the P95: about one losing run in 20 is longer than that. That doesn’t make it comfortable. It means eight losers on its own isn’t a reason to change anything. These figures describe what happened in those 13 years, and a longer run than 28 is possible.

How do losing runs stretch as the strike rate falls?

The favourite is the gentlest case. Here are three more strategies from the same 13 years, all rebuilt in date order. “Rank 1” means the horse that rating or model puts top in its race. Race Predictor is my model that turns the evidence on each runner into a win probability.

PR and VDW are two of my own ratings. Read down the strike rate column and the runs get longer as it falls. VDW Rank 1 has an A/E of 0.99, so no edge, and its longest run is 33, not far off Race Predictor’s 38. Four samples can’t prove a rule, but they point one way: in these samples strike rate lined up with run length, and the small differences in A/E didn’t shorten the runs. If your method wins one bet in four, plan for runs well beyond what the favourite produced.

Race Predictor Rank 1 is the one row with an A/E above 1.00. That means the odds implied fewer winners than it actually produced in this sample. It’s a measure of the prices, before commission, and I’m not showing a profit figure here. Even so, it had the longest run of the four, 38.

How much of my bankroll does a losing run cost?

This is where the streaks turn into a decision. Suppose you stake a flat 1% of your starting bank on every bet. Sixteen losers in a row, the P99 for Race Predictor Rank 1, costs 16% of the bank. The longest, 38 losers, costs 38%.

My judgement, which I haven’t tested as a rule, is to pick the row nearest your own strike rate and size your stake so a run a bit longer than that row’s longest wouldn’t force you out. If you can’t live with a 38% drawdown, 1% is too big for a method that wins one in four.

Laying a horse means betting that it loses. A lay-to-win method flips the shape: lots of small wins and the occasional big loss, so the runs of losers are short but each loss hurts. I haven’t measured that here.

How can I use losing runs as a planning tool, not a verdict?

Before the first bet on a method, I’d decide four things:

1. The strike rate the method is built around, and so the kind of runs you’re accepting.
2. The longest losing run you’re prepared to sit through without changing anything. Most people overestimate this, so be honest.
3. The bankroll and stake that make that run survivable.
4. What would make you review the method. A run of losers isn’t it. A sustained fall in A/E against what the method showed before is. With fewer than about 1,000 bets an A/E tells you very little, and I’d want several thousand before I read much into it.

Ratings that agree give you shorter runs too. The races where PR Rank 1 and VDW Rank 1 pick the same horse had a median run of 2, a longest of 25 and only 2% of runs reaching 10, at a strike rate of 35.6%. That’s mostly the higher strike rate at work, and the combining ratings post shows the price still allows for it. A gentler run makes the method easier to stick with, though, and that matters on a bad Sunday night.

What this doesn’t tell you

These streaks come from UK racing between July 2013 and March 2026 at Betfair SP. They tell you what normal looked like, not what your next run will be, and they don’t say which method makes money. For the staking side, horse racing bankroll management goes through it.

If you’d like to see the same PR, VDW and Race Predictor signals on today’s runners, the racecards show them. See today’s racecards.