Picture a decent set of selections and a fair long-run record. Six weeks in, the bank has had a rough patch, the next losing afternoon feels unbearable, and you stop. The selections may have been fine. The ride was simply too bumpy to stay on.
That is what poor staking looks like. Finding selections at prices that give you an edge decides whether you can win over time. Your staking plan decides whether you stay around long enough to find out.
By the end of this guide you will be able to do five things: set up a separate betting bank and a unit, work out how long a losing run your method can throw at you, choose a staking plan you can follow without second-guessing, size each stake by price and edge if your records justify it, and check the whole thing on paper before real money goes on.
In this guide
- The short answer: what is the best staking plan?
- What a staking plan can and cannot do
- Step 1: Set up a bank and a unit
- Step 2: Find out how long your losing runs can be
- Step 3: Start with level stakes or a small percentage
- Why the price you back at changes the ride
- Step 4: Stake by price and edge (the Jackson formula)
- Other plans, and which to treat with care
- Step 5: Test it on paper before you commit
- Common staking mistakes and how to avoid them
- Do this today: a short checklist
- Staking plan questions, answered
The short answer: what is the best staking plan?
There is no single best plan. For most backers I would start with level stakes at a small, fixed fraction of the bank, and only move to staking by price and edge once your own records give you an edge figure you trust. A sensible plan sizes each stake by the price and by how big an edge you think you have, rather than by mood or recent results.
Here is why price matters. In an earlier analysis of just over 1.4 million GB and Ireland runners (1 July 2013 to 31 March 2026, Betfair SP, after 2% commission), the swing in results per bet grew roughly seventeenfold from short prices to long prices at level stakes. Smaller stakes on bigger prices is one way to smooth that out. These figures describe that stretch of racing and will not repeat exactly, so treat them as a guide to the shape of the problem, not a forecast.
What a staking plan can and cannot do
Betting a selection method and living through it are different things. Many people never see the long-run result of their method because they stop part-way through a bad run, sure it is broken, when the maths said that kind of run was possible.
A method with a real edge and poor staking can still run out of money. A modest method with steady staking at least survives long enough for the results to tell you the truth. So the first job of a staking plan is not to maximise profit. It is to keep you in the game.
What breaks people is variance, meaning how widely your results swing around their average. A staking plan is how you control it.
What it cannot do
Staking does not create an edge. It only expresses one. If your bets are value, a good plan helps you stay alive while variance does its work. If they are not, no plan rescues them. It only decides how loudly you learn the lesson.
If you are not yet sure what makes a price value, read how to calculate your edge first. Everything after that is easier.
What it can do
- Stop one bad afternoon from ending your betting.
- Make your results comparable from bet to bet, so you can see what is working.
- Take the stake decision out of your hands when you are least reliable, which is straight after a loss.
- Let you judge a method on its merits, because the bank survives long enough to give it a fair run.
Step 1: Set up a bank and a unit
A betting bank is money set aside for betting and nothing else. It is not your account balance and it is not whatever is left at the end of the month. If you top it up every time it runs low, it is not a bank, because a loss never actually counts.
Then pick a unit. A unit is a fixed slice of the bank, and thinking in units keeps the link between bank and stake steady even when the pound amounts start to sting. As an example of the arithmetic (illustrative, not a recommendation of any particular size): a 1,000 unit bank with a 0.5% unit means a 5 unit stake.
If you bet at both bookmakers and an exchange, count the total across both as one bank. Splitting money over several accounts makes it easy to forget how much you actually have at risk.
For the longer version, including how to size the bank around your method, see our guides to stake sizing and bankroll management.
One bank, one plan, one review rhythm
Write your staking rule down before racing starts, in one sentence you would be happy to read back after a bad day. For example: “One unit on every selection, one unit is 0.5% of the bank, reviewed on the first of the month.” If the rule only exists in your head, the stake box will rewrite it when the pressure arrives.
Step 2: Find out how long your losing runs can be
Your stake must be small enough to survive a normal run of losers. So before you choose a stake, find out what a normal run of losers looks like for your method. The lower your win rate, the longer the dry spells, whatever the average return.
Here is how three simple level-stake signals looked over 1 July 2013 to 14 October 2026 in Great Britain and Ireland, at Betfair SP with 2% commission, with fallers and pulled-up horses counted as losers. VDW Rank 1 and Race Predictor top horse are Race Advisor’s own ratings. A/E is actual winners divided by the winners the prices expected, so 1.00 means the prices were about right. These are not tips, and they will not repeat exactly.
| Signal | Bets | Win rate | A/E | Longest losing run |
|---|---|---|---|---|
| Betfair SP favourite | 167,257 | 34.4% | 1.00 | 28 |
| VDW Rank 1 | 153,920 | 21.6% | 0.99 | 59 |
| Race Predictor top horse | 167,314 | 25.5% | 1.01 | 36 |
Nothing was broken in any of those runs. A method that wins about one time in five will, over a very long record, hit a long wait. Your own record will be shorter, so treat these as a fair warning of what can happen, not a ceiling.
How to estimate your own longest losing run
Paper trade your method until you have at least 100 results. Then use this spreadsheet formula, where the first number is how many bets you plan to place and the second is your win rate as a decimal:
=LOG(100)/-LOG(1-0.2)
That returns 20.63, so at a 20% win rate over 100 bets, expect a losing run of around 21 at some point. Change 100 to the number of bets you expect to place, and 0.2 to your own win rate. A simple rule from our stake sizing guide is to hold a bank of at least three times that expected run, counted in units.
Look at drawdown too, not just the run
A losing run counts consecutive losers. A drawdown is the fall from your highest running profit to the lowest point after it, and it can include winners that never quite got you back to the peak. Both matter. The post on drawdowns and bank size shows how to build the three spreadsheet columns that measure it.
Step 3: Start with level stakes or a small percentage
This is my decisive recommendation. Until you have a long, tracked record, stake the same one unit on every bet (level stakes), with the unit a small fraction of the bank. Level stakes are the honest default. They keep testing clean, they stop you chasing losses, and they show you what your selections really do.
A small percentage plan, where each stake is a fixed share of the current bank, is the other sensible start. As an example of the arithmetic, 0.5% of a 1,000 unit bank is 5 units. After ten losers in a row the bank is about 951, so the next stake is about 4.75. The stake shrinks when things go badly, so a run of losers cannot empty the bank in one go.
Many people use 0.5% to 1%. My own starting point is 0.5%, and I would only move up once I had watched my own method for a few hundred bets. Both numbers are conventions, not tested results.
Check the stake against your longest run
Here is a rule of thumb, which I have not tested against results. Decide how much of the bank you could live with losing in a bad patch. Then choose a unit so that your worst measured drawdown, plus a margin, stays inside that figure.
An illustrative example, not real results: suppose you could cope with a 25% dip and your worst measured drawdown is 40 units. Then one unit should be no more than 0.625% of the bank (25 divided by 40). Want a margin, say double the measured figure? Then one unit becomes about 0.3%. The margin matters because the worst drawdown in any sample is only the worst so far.
If a dip of that size would still feel like too much, you may not need a different plan. Halving the stake halves the money lost over the same patch.
There is a fuller comparison of the main options in our guide to the best horse racing staking plans.
Why the price you back at changes the ride
Level stakes mean the same one unit on each bet. They also hide how different bets feel at different prices.
In that earlier analysis, the GB and Ireland runners with a valid Betfair SP (BSP, the Betfair Starting Price) were grouped by price. Here is the typical spread of profit or loss on a one-unit bet (a bigger number means wilder results), after 2% commission on winners:
- BSP 1.01 to 2.0 (around 23,000 bets): 0.81
- BSP 2.01 to 3.0 (around 55,000 bets): 1.24
- BSP 3.01 to 5.0 (around 151,000 bets): 1.74
- BSP 5.01 to 10.0 (around 327,000 bets): 2.53
- BSP 10.01 to 20.0 (around 319,000 bets): 3.73
- BSP 20.01 to 50.0 (around 283,000 bets): 5.64
- BSP 50 and above (around 267,000 bets): 14.13
Same stake and same commission throughout. The bands hold different numbers of bets, but the spread per bet still grew about seventeen times from the shortest band to the longest. Long losing runs followed the same pattern. In that analysis the longest losing run was 33 bets at BSP 3.01 to 5.0, 58 at BSP 5.01 to 10.0, and 107 at BSP 10.01 to 20.0. Those are how things looked in that sample, not a limit.
Short prices feel very different. Using the later, fully checked figures, backing the Betfair favourite in 167,257 GB and Ireland races (1 July 2013 to 14 October 2026, level stakes, 2% commission, fallers and pulled-up horses counted as losers) won 34.4% of the time, and the longest losing run was 28. Those figures describe that period and will not repeat exactly, but they show why short-priced selections are easier to sit through.
Picking shorter-priced horses is one way to calm the ride. Your staking plan can do the rest, and that is where the next step comes in.
Step 4: Stake by price and edge (the Jackson formula)
Only move to this step once your tracked record gives you an edge figure you trust. Until then, stay on Step 3.
The formula I have used longest came from John Jackson at SmartSigger, who walked me through it over lunch in 2015. It is not the only option. It handles the variance problem without being as aggressive as full Kelly (a method that stakes the share of your bank said to grow it fastest).
Stake = Bankroll / (Risk Level × (Decimal Odds – 1) / Edge)
- Bankroll is the current size of your betting bank.
- Decimal odds is the price you are taking, so 5/2 is 3.5.
- Edge is your estimated advantage on that bet as a decimal. A 10% edge is 0.10.
- Risk Level is a personal dial. A lower number means bigger stakes, a higher number means smaller ones.
As a rule of thumb, not something I have tested against results, 10 is a cautious place to start if you are new to variable staking. Adjust it once you have seen what your stakes look like across a card.
The shape is simple. Stake shrinks as the price rises, so a win at a big price and a win at a short price add roughly the same to your bank. The bank then climbs in smaller, steadier steps, and each individual bet feels closer to the same size.
A worked example: how the stake changes by price
Take a 1,000 unit bank, a Risk Level of 10, and a 10% edge on each bet. This is an illustration of the arithmetic, not a claim that you will find that edge.
- Decimal odds 2.0 (evens): stake = 1000 / (10 × 1.0 / 0.10) = 10 units
- Decimal odds 3.5 (5/2): stake = 1000 / (10 × 2.5 / 0.10) = 4 units
- Decimal odds 6.0 (5/1): stake = 1000 / (10 × 5.0 / 0.10) = 2 units
- Decimal odds 11.0 (10/1): stake = 1000 / (10 × 10.0 / 0.10) = 1 unit
If the 10/1 shot wins, you collect 10 units of profit. If the 5/1 shot wins, 10 units. If the 5/2 shot wins, 10 units. If the evens shot wins, 10 units. The wins line up. The losses are far smaller where the price is big: 1 unit when the 10/1 shot misses, against 10 units when the evens shot misses.
Over a season, the same selections may give a smoother line of results this way. Staking does not change which horses win. It changes how bumpy the journey is.
How the Risk Level dial works
Keep the same 1,000 unit bank, the same 5/2 price (decimal 3.5) and the same 10% edge, and change only the Risk Level. Again, this is illustrative arithmetic, not real results.
| Risk Level | Stake at decimal 3.5, 10% edge |
|---|---|
| 5 | 8 units |
| 10 | 4 units |
| 20 | 2 units |
Double the Risk Level and the stake halves. That makes it easy to tune: if a card of stakes looks too big to sit through, raise the number.
The input that matters most is the edge
Bankroll, price and risk level are all easy to see or choose. Edge has to come from your own analysis. Treat any edge estimate as unproven until it has held up across a full record, losing runs included, before it is allowed to raise a stake. To build that figure, see how to calculate your edge. Without an edge figure you trust, the formula becomes a neat way of staking in proportion to price. That is still a reasonable idea, but it is not the full method.
Overstate your edge and the formula stakes too much. In the example above, a claimed 30% edge on a 5/2 shot turns a 4 unit stake into 12. Halve the edge you claim and the stake halves too, so a cautious estimate gives a gentler plan. The formula has no instinct for honesty, so you have to supply it.
Add a ceiling
One extra safeguard, which I have not tested against results: set a maximum stake for any single bet, such as 1% of the bank, and let the formula never exceed it. That protects you on the day an over-optimistic edge estimate meets a short price.
Other plans, and which to treat with care
You will meet plenty of other staking plans. A useful way to sort them is by one question: does the plan raise your stake after a loss, or only when your records justify it?
Kelly and fractional Kelly
Full Kelly sets the stake from your estimated win chance and the price. It takes your chance estimate as gospel, so if your estimate is too high, it simply makes the stake bigger. If you want to see it worked through, our Kelly criterion guide shows how it behaves on a measured edge and why a fraction of Kelly is the usual compromise. Treat Kelly as a later step, after you have thousands of tracked bets and an edge you can defend.
The square root plan
The square root staking plan (Base Bet plus Square Root) is not a loss-recovery plan. You keep a fixed base bet, and when you are in profit over your chosen period, you add the square root of that profit to it. The idea is that as profit grows, the extra staked becomes a smaller share of it. It is simple to run and worth a look if you want a gentle way of increasing stakes.
Loss-recovery and reverse systems
Plans such as the retirement staking plan try to win back earlier losses by raising stakes after a losing bet. The related idea of reverse Kelly is a deliberate step-up after losses. In Race Advisor’s test of a plan that staked up after losses, the bank was gone partway through the record, while flat stakes on the same bets survived. My advice is to keep your stakes tied to price and edge, not to the last result.
None of this means a clever plan cannot help. It means any plan that depends on the next bet making up for the last one is relying on something racing does not owe you. Each race starts fresh.
Step 5: Test it on paper before you commit
A sensible test is to paper-track both approaches at the same time on the same bets, level stakes in one column and price-and-edge stakes in the other, and compare how each bank moves. That shows you the ride, not whether either plan has an edge, and a small sample is suggestive only. It will still tell you which plan you could live with.
How to run the test
- Pick one method and one bank size. Write both down.
- Record every selection, the price you could actually get, and the result. Do not leave out the ones you would rather forget.
- Work out the level stake and the price-and-edge stake for each bet, side by side.
- Keep a running total for each column, and note the lowest point each reaches.
- After a few hundred bets, ask which line you could have stayed with through its worst patch.
The Race Advisor bet tracker keeps this record for you, and a spreadsheet works too. Judge the result only once you have a decent sample, and do not change the rules halfway through.
A note on bank size
One idea, which I have not tested properly: variable staking might let you work with a smaller bank than level stakes, because the long-priced bets are staked small. As a rough rule of thumb only, many people like a bank of a few hundred typical stakes, with more if the Risk Level is set aggressively. Treat that as a starting point to adjust, not a promise.
If you want to see how this plays out on live prices, open today’s race cards and work out the stake for a few horses on paper first.
Common staking mistakes and how to avoid them
Feeding the formula a bad edge
The quickest way to break a price-and-edge plan is to give it a number you do not believe. Fix: use the cautious end of your estimate, and apply a ceiling.
Layering a progression on top
Some people add a recovery system, or an override to stake more on strong fancies. Each addition brings back the swings the formula was meant to remove. A deep losing spell plus a progression can turn a recoverable bad patch into a bank-ending one. Fix: pick one plan and leave it alone. Our guide to drawdowns and bank size explains the cost.
Using staking where there is no edge
Staking cannot create value. If the A/E (actual winners against expected winners, where 1.00 means results matched the odds) is below 1.00, edge-based staking just delivers the losses more efficiently. Fix: test the selections at level stakes first.
Topping up the bank
Topping up after a loss hides the real drawdown and lets a too-big stake carry on. Fix: if the bank is too small for your stake, cut the stake, not feed the bank.
Betting more races to get it back
A hunt for the bet that repairs the day is where marginal selections slip in. Fix: decide your daily list before racing and stick to it.
Changing the rules mid-run
If you change your stake rule during a losing spell, you cannot tell what caused the result. Fix: change rules only at your planned review date.
Do this today: a short checklist
- Set aside a betting bank that is separate from your everyday money, and write down its size.
- Choose a unit as a small slice of that bank, such as 0.5%.
- Write your one-sentence staking rule and put it where you can see it.
- Find your method’s longest losing run and worst drawdown from your own records, or paper trade until you have 100 results.
- Check that your longest dry spell, at your unit size, leaves the bank in a state you could live with.
- Stake level for now. Move to price-and-edge staking only when your records support an edge you trust.
- Log every bet: price taken, stake, result.
- Set a review date a month ahead, and do not change the rules before it.
Responsible betting note: only stake money you can afford to lose. No staking plan guarantees a profit, and every method, however well tested, can lose.
Staking plan questions, answered
What is the best staking plan for horse racing?
There is no single best plan. For most backers, level stakes at a small fixed fraction of the bank is the best start. Move to staking by price and edge only when your own tracked records support an edge you trust.
How much of my bank should I stake on each bet?
Many people use 0.5% to 1% of the bank. My own starting point is 0.5%. Both are conventions, not tested results. Check the stake against your method’s longest losing run and worst drawdown, and cut it if the dip would be too much to live with.
Is level staking better than variable staking?
Level staking is the simpler and cleaner default, and it keeps testing honest. Variable staking by price and edge can smooth the ride, because long-priced bets get smaller stakes, but it depends on an edge figure you can trust.
Should I increase my stake after a loss?
No. Raising stakes to recover losses brings back the swings a sound plan is meant to remove. Tie stakes to price and edge, not to the last result.
What is the Jackson staking formula?
It sets the stake as Bankroll divided by (Risk Level times (Decimal Odds minus 1) divided by Edge). Stake shrinks as the price rises, so wins at short and long prices add roughly the same amount to the bank.
What Risk Level should I use in the Jackson formula?
As a rule of thumb, not something I have tested against results, 10 is a cautious place to start. A higher number means smaller stakes and a lower number means bigger ones.
Can a staking plan turn a losing method into a winning one?
No. Staking expresses an edge. It cannot create one. A plan can make a method with a real edge easier to stay with, and it can slow the losses on one without.
How do I know how big a losing run to expect?
Paper trade your method for at least 100 results, then use the spreadsheet formula =LOG(bets)/-LOG(1-win rate). At a 20% win rate over 100 bets it gives about 21. Real runs can be longer, so leave room.
