Staking Plans for Horse Racing: Level Stakes, Kelly and More, Tested by Simulation

Home › Betting Strategy › Staking Plans

By The Guv'nor · Last updated October 2026

The best staking plans for horse racing are the boring ones: level stakes of about 1% of your bank, or a quarter to half of the Kelly stake if you can honestly price your edge. I ran 20,000 simulated punters through 1,000 bets each. Martingale and Fibonacci broke almost every time, and 5% stakes went bust more often than not.

Why staking matters more than tipping

Look, everyone wants to talk about the horse. Hardly anyone talks about how much to put on it. A bettor with a genuine edge can still go skint with the wrong stakes, and no staking plan turns a losing bettor into a winner. It only decides how rough the ride is and if you're still standing at the end.

I'm essentially a statistics/maths geek. I was a Quantitative Researcher at a Tier 1 investment bank before all this, and in that world the size of a position matters as much as the position itself. Racing is no different, and it's the same maths that got me to the point where most of your favourite bookies won't take a bet from me. Get the price right and the stake wrong, and the losing runs will find you out.

So this page is the maths: exact losing-run odds, then eleven staking plans put through a simulation. If you want the basics of points first, start with what 1pt means in betting. The wider rulebook is in my horse racing betting strategy guide.

Level stakes: the honest baseline

Level stakes means the same stake on every bet, whatever the price and however much you fancy it. Pick a point size, say £2, and every bet is 1pt. Win, lose or beaten a nose, the next bet is still 1pt.

It's the baseline for a reason. Your profit or loss in points tells you straight away whether your betting is any good, because no bet counts more than another. Profit divided by total staked gives your return on investment (ROI), and that's the number that matters, not your win rate. Anyone can talk about a high win rate when they tip odds-on horses all day.

Worked example (made-up numbers): 100 bets at 1pt each, 22 winners at an average of 9/2. The winners return 22 x 5.5 = 121pts for 100pts staked. That's +21pts, or +21% ROI, from a 22% strike rate.

The weakness is that level stakes don't shrink when the bank does. If your point is too big a slice of the bank, a normal losing run can empty it before the edge has time to show.

Percentage of bank and the Kelly criterion: full, half and quarter

Percentage of bank

Here you stake a fixed percentage of whatever the bank is right now. At 2% of a £100 bank the first bet is £2. Lose a few and the bank drops to £80, so the next stake is £1.60. Win a few and the stakes grow with you.

The good bit: stakes shrink as you lose, so in theory you never hit zero. The catch: you claw back slowly, because you're betting smaller just when you need to recover. A bank that halves needs to double to get back.

The Kelly criterion for horse racing

Kelly is percentage-of-bank staking where the percentage comes from your edge. It was published by John Kelly in 1956 ("A New Interpretation of Information Rate", Bell System Technical Journal) and it sets the stake that grows your bank fastest over the long run. For a single bet:

Kelly stake = (b x p - q) / b, where b is the fractional odds (4/1 is 4), p is your honest chance of the horse winning and q = 1 - p.

Worked example: you rate a horse a 20% chance and you can back it at 17/4 (5.25 in decimal). Kelly says (4.25 x 0.20 - 0.80) / 4.25 = 0.0118, so 1.18% of the bank. On £100 that's £1.18. Half Kelly is about 59p and quarter Kelly about 29p. If your chance is below what the price implies, Kelly says don't bet at all.

Notice how small that is. A 5% edge on a 17/4 shot is a decent edge, and full Kelly still says barely more than 1% of the bank.

Why the pros use a fraction of Kelly

Kelly only works if p is right, and nobody knows p exactly. Bill Benter, who ran a computer-model betting operation on Hong Kong racing, wrote in his 1994 paper ("Computer Based Horse Race Handicapping and Wagering Systems: A Report") that betting the full Kelly amount is "unwise". He gave three reasons. If you overestimate your advantage by more than a factor of two, Kelly betting shrinks your capital, and he added that doing so "is easily done in practice". Regular withdrawals mean your real bank is smaller than it looks. And full Kelly is "a 'rough ride'", with downswings of more than 50% "a common occurrence". He recommended a fraction such as 1/2 or 1/3. More from his paper is in my piece on AI horse racing predictions.

The maths backs him up. In the 17/4 example, staking twice the Kelly amount (about 2.35% of bank) leaves you with almost no long-run growth, and anything bigger shrinks the typical bank despite your edge. Crazy right? A winning bettor can go backwards purely by staking too big.

Plans that blow up: Martingale and Fibonacci

These are loss-chasing plans dressed up as systems. Most sequences end in a small profit, then one doesn't and takes everything.

Martingale

Double the stake after every loser and go back to the start after a winner. At 4/1 a winner recovers everything. The problem is the ladder: £1, £2, £4, £8, £16, £32, £64. Seven losers in a row cost £127, and the eighth bet needs £128. At a 20% strike rate, any given sequence of seven bets has a 21% chance (0.8 to the power 7) of being all losers. On a £100 bank you can't even cover the seventh step after six straight losers.

Fibonacci

Stakes follow 1, 1, 2, 3, 5, 8, 13, 21, 34, 55 and so on. In the common version you go up one step after a loser and back two after a winner. It climbs more slowly than Martingale, but nine losers in a row still cost 88 units and the tenth bet needs 55 more. At a 20% strike rate you go up a step four times as often as you come down two, so the stakes drift upwards over time.

Neither plan changes your edge by a penny. They just bunch your losses into one big hit. If I am honest, chasing is the fastest way to ruin a good bet.

Staking plans for horse racing, simulated: losing runs and drawdowns

Everything below is my own maths: the losing-run table is calculated exactly and the staking tables are a simulation I ran. None of it is anyone's real betting record.

How long losing runs get

These are exact probabilities, not a simulation, assuming each bet is independent with the same chance of winning. "Typical longest run" is the average longest run of losers you'd see in that many bets.

Read the 20% row again. If you back horses around 4/1 and win one bet in five, you should expect a run of about 26 losers somewhere in your next 1,000 bets. Over 500 bets, the chance of at least one run of 15 or more is 98%. That's not bad luck. That's just what a 20% strike rate looks like.

Eleven staking plans, 20,000 simulated punters

How I set it up: every simulated punter starts with £100 and has 1,000 win bets. Each bet wins 20% of the time at 17/4 (5.25), so the punter has a genuine edge of 5p in the pound. "Bust" means the bank fell to £10 or less at any point, or couldn't cover the next stake; bust punters stop betting. Drawdown is the biggest fall from a high point to a later low. N = 20,000 punters per plan, same random seed for every plan; a rerun with a different seed moved the headline figures by about 1 percentage point or less.

• Size beats system. Level £1 went bust 4.8% of the time and level £5 went bust 59.9% of the time. The only difference is the stake.
• Bigger stakes win more, when they survive. Level £2 has the best median bank at £190, but a one-in-four chance of going bust along the way. That's the trade you're making.
• Fractional Kelly is the smoothest ride. Quarter Kelly had a median worst drawdown of just 16% and nobody went bust. Full Kelly's median drawdown was 53%, which is exactly the "rough ride" Benter described.
• Overestimating your edge is lethal. The punter who thought his 20% shots were 25% shots staked 7.35% a bet and went bust 94.2% of the time, despite having a real edge.
• Martingale and Fibonacci died young. Half the Martingale punters couldn't cover the next stake by their 13th bet. Their "in profit" figure means the bank was above £100 when the ladder broke.

I also ran the same plans with no edge at all (20% at 4/1). Level £1 finished with a median bank of £100, right where it started, and 15.1% went bust; Kelly simply says don't bet. And for a punter losing 10p in the pound (20% at 7/2), level £1 went bust 67.9% of the time. No staking plan fixes a losing bet. Finding the edge is a separate job, and it's covered in value betting in horse racing.

Staking plans for place betting and outsiders

Your strike rate decides how big a bank you need. Place bets win more often at shorter prices, so the losing runs are shorter. Outsiders win less often, so the runs are longer. Same simulation, level stakes, every bet type set to the same 5p-in-the-pound edge:

So a sensible staking plan for place betting can run on a smaller bank than one for win bets, and an outsiders-only approach wants 200 points or so. Two things to remember with each-way: 1pt each-way is two bets, so it costs 2pts, and the place half softens the losing runs without removing them. The full each-way maths is in each-way betting explained.

Multiples are a different beast again. A Lucky 15 is 15 bets on four horses, and the stake and the variance both add up quickly; I've worked through it in Lucky 15 explained.

The plan I use: 1pt = £5 on a £100 bank

I bet on every horse I tip, and I tell every member the same thing: if you don't follow strict staking sizes you will likely lose. Divide your bank into points. The example I give is 1pt = £5 on a £100 bank, and in the newsletter I'll try to suggest a points stake for each bet. A 'Banker' in my newsletter is one of my shorter-priced bets; the outsiders come from a system that picks out horses that have won or gone close off a higher weight or mark than they're running off today.

What you can't do is wake up, fancy one a bit more because you have a feeling and stick your whole balance on it. No chasing, no going all in. If I'm unsure about a bet, I skip it. Some days there are no tips at all, a stinky day for the ponies, and that's part of staking discipline too.

Now the honest bit. A £100 bank at £5 a point is 20 points, and the tables above say 20 points is a thin cushion for win bets. I like outsiders and each-way value, because even when they place we can make more profit than an odds-on shot winning. But outsiders mean lower strike rates and longer losing runs than someone backing favourites. So if you want a smoother ride, use a bigger bank. Keep the £5 point and build the bank towards £250 to £500 (50 to 100 points), or keep the £100 and make a point £1 or £2.

I judge myself on profit month to month, not win rate. Understanding that losers are a thing is the key to being a winner longer term. Size your points for the run you'll actually get, and trust the process.

Betting bankroll management: my short version

1. Keep a separate betting bank and decide its size before you start, not after a bad week.
2. Split it into points: 100 points if you mainly bet win singles around 4/1, more if you like outsiders.
3. Stake level points, or a quarter to half of Kelly if you can price your own chances and you've kept records long enough to trust them.
4. Never chase. Martingale-style doubling is the quickest route to an empty bank.
5. Re-size your points occasionally (every month or every 100 bets, say), not after every result.
6. Keep records in points and judge yourself on ROI over hundreds of bets, not on last Saturday.

Following a tipster? Same rules, same losing runs. I've put the case for and against in are horse racing tipsters worth it, and if you're thinking about doing this for a living, read can you make a living betting on horses first. My own selections and suggested stakes go out through my tips service.

Staking plan questions

What is the best horse racing staking plan?

For most punters, level stakes of about 1% of the bank. In my simulation of 1,000 bets at a 20% strike rate, level stakes at 1% went bust 4.8% of the time against 59.9% at 5%. If you can genuinely price your chances, a quarter to half of the Kelly stake gave the smoothest ride of all.

What does level stakes mean in betting?

The same stake on every bet, usually 1 point, whatever the price or how confident you are. It keeps your profit in points and your ROI honest.

Should I use the Kelly criterion for horse racing?

Only a fraction of it, and only if you have a tested way of estimating each horse's chance. Full Kelly assumes your probabilities are right. Overestimate your edge by more than a factor of two and it shrinks your bank, which is why Benter recommended betting 1/2 or 1/3 of the Kelly amount.

What is a good staking plan for place betting?

Level stakes still work, and because place bets land more often the losing runs are shorter. In my simulation at a 40% strike rate, a 50-point bank went bust 4.0% of the time over 1,000 bets, against 25.6% for win bets at 20%.

How big should my betting bank be?

Big enough to survive the longest losing run you're likely to see. At a 20% strike rate that's about 26 losers in a row over 1,000 bets, so 100 points is a sensible minimum for win betting, and more for outsiders.

More in this guide

• Horse racing betting strategy: 10 rules from an ex-quant
• What does 1pt mean in betting? Points, stakes and betting banks
• Value betting in horse racing
• Horse racing betting systems that work? I tested the popular ones
• Can you make a living betting on horses?
• How to bet on horse racing