Building a profitable horse racing betting system sounds easy enough.
Find something that appears to work, test it against previous results, add a few filters to improve the profit and then start betting the selections.
Unfortunately, this is also how some very impressive-looking betting systems are created that subsequently fall apart as soon as they’re asked to predict races that haven’t happened yet.
The problem isn’t necessarily the original idea. It’s often what happens during the testing process.
You discover that your system loses money at Newmarket, so you remove Newmarket. It doesn’t perform particularly well on soft ground, so that goes too. Four-year-olds aren’t profitable, so you exclude those. Results are poor above 8/1, so you add another rule.
Before long, you’ve created a system with fantastic historical results.
The trouble is that you may not have discovered a profitable betting angle at all.
You may simply have created a perfect description of what happened in the past.
There’s a big difference.
Start With an Idea You Can Explain
A good betting system should begin with a logical idea that you can explain in one sentence.
For example:
‘Horses with a top-two speed figure who are drawn low in five-furlong handicaps.’
That’s understandable.
There’s also a logical reason why the angle might work. You’re combining demonstrated ability with a potentially favourable draw in a type of race where track position can matter.
Your starting idea could involve trainers, jockeys, speed ratings, recent form, course records, distance, going, pace or countless other factors.
But before testing anything, ask yourself:
Why should this work?
If you can’t answer that question, be careful.
A system that exists purely because you’ve discovered a profitable combination of historical numbers may have found nothing more than coincidence.
No amount of filtering will turn a bad idea into a good one.
Keep the Number of Rules Low
This is probably one of the most important lessons in system building.
Every additional rule reduces your sample size.
Imagine starting with 8,000 historical qualifiers.
You add a distance rule and you’re down to 4,500.
Add a going requirement and you’re down to 2,000.
Then age, draw, trainer strike rate, days since last run, previous finishing position and an odds restriction.
Suddenly your magnificent system is based on 83 bets.
It might show a 35% return on investment and look fantastic.
But what have you actually proved?
Not very much.
Generally, I’d much rather see three or four logical rules producing several thousand historical selections than ten rules producing eighty.
Large samples give you much more confidence that what you’re seeing might represent a genuine pattern rather than random variation.
Simple systems also have another advantage.
You understand why they’re selecting horses.
That’s surprisingly important.
Beware of the ‘Perfect’ Back-Test
If you’ve spent an hour tweaking a system and eventually turned a £200 historical loss into a £4,000 profit, don’t immediately celebrate.
Ask yourself how you achieved it.
Did you discover something genuinely useful?
Or did you keep changing the rules until the historical results looked attractive?
This is known as overfitting.
It’s one of the biggest dangers when creating betting systems because modern software allows us to analyse enormous amounts of racing information very quickly.
The Inform Racing System Builder, for example, allows users to test combinations of form, ratings, trainers, jockeys, race conditions and many other factors against historical results. The results can then be broken down by different categories to see where a system has performed particularly well or badly.
That’s extremely useful.
But it also means discipline is required.
If you examine enough breakdowns, you will almost certainly find something profitable somewhere.
The important question is whether there’s a sensible racing reason for it.
If your system performs poorly over six furlongs but brilliantly over seven, perhaps there’s a logical explanation.
If it loses money when horses are drawn in stalls 7 and 11 but makes a profit from every other stall, that’s probably not a reason to exclude stalls 7 and 11.
Don’t confuse coincidence with information.
Test Across Different Seasons
A system shouldn’t just work during one particularly favourable year.
Test it across at least three seasons if sufficient data is available.
Better still, look at each season separately.
Suppose your results are:
2023: +46 points
2024: +39 points
2025: +51 points
That’s interesting.
Now compare that with:
2023: -12 points
2024: -8 points
2025: +156 points
Both systems show an overall profit.
But they tell completely different stories.
The second system needs investigating.
Perhaps there was one enormous-priced winner in 2025. Maybe racing conditions were unusual. Perhaps a particular trainer completely changed their approach.
Consistency doesn’t mean every month or year must make a profit. Losing periods are inevitable.
What you’re looking for is evidence that the underlying idea keeps showing signs of working.
Find Out Where the Profit Actually Came From
This is another area where headline profit figures can be misleading.
Suppose a system shows:
1,200 bets
+180 points profit
15% ROI
That sounds excellent.
But then you discover that two winners at 80/1 and 100/1 contributed 180 points between them.
Without those two horses, the remaining 1,198 bets broke even.
Does that mean the system is useless?
Not necessarily.
Big-priced winners are part of racing and shouldn’t automatically be removed simply because they’re inconvenient to your analysis.
But you need to know how dependent the results are on them.
A system producing steady profits across several odds ranges is very different from one whose entire historical success depends on a couple of exceptional results.
Always look underneath the headline figure.
Check Both Strike Rate and Profit
Profit naturally attracts our attention, but strike rate tells you something different.
A system with a 30% strike rate and modest profit might be much easier to follow than one showing spectacular profits from a 6% strike rate.
Why?
Losing runs.
A low strike-rate system can easily produce twenty, thirty or more consecutive losers even when it’s working perfectly well.
That creates psychological pressure.
People reduce their stakes, stop betting the system or, worst of all, abandon it just before the winners arrive.
When testing a system, therefore, don’t just ask:
‘How much did it make?’
Also ask:
How often does it win?
What’s the longest historical losing run?
How frequently do losing months occur?
Could I genuinely continue betting this through the bad periods?
A profitable system you can’t psychologically follow isn’t much use to you.
Compare SP and Betfair SP
Price matters enormously when assessing historical systems.
Look at results at both Starting Price and Betfair Starting Price where possible.
A system that loses at SP but makes a consistent profit at BSP may still be perfectly usable.
You simply need to know that before staking real money.
Likewise, if historical profits depend on obtaining prices that would have been extremely difficult to get in practice, your back-test isn’t giving you a realistic picture.
This is an important distinction between theoretical and practical profitability.
The question isn’t just:
‘Would this system have made money?’
It’s:
‘Could I realistically have obtained the prices required to make that money?’
That’s a much better question.
Use Result Breakdowns Carefully
Breaking system results into categories can uncover extremely useful information.
Perhaps a trainer system performs particularly well between two and three miles.
Maybe top-rated horses perform much better in lower-class handicaps than they do in Group races.
Perhaps a particular angle works extremely well with three-year-olds but shows nothing among older horses.
These discoveries can improve a system because they may reveal something genuine about the underlying racing angle.
Inform Racing’s System Builder provides result breakdowns specifically for this purpose, allowing users to examine where an idea has historically performed well and where it hasn’t.
But again, there is a danger.
Don’t automatically remove every losing category.
If you do that repeatedly, you’re back to describing history rather than predicting the future.
Every exclusion should ideally have a logical explanation.
Don’t Be Frightened of a Small Edge
Another common mistake is rejecting systems because they don’t produce spectacular historical profits.
People naturally want to discover:
500 bets
200 winners
+400 points
80% ROI
If you find that, be suspicious before you become excited.
A system producing a modest but consistent advantage across thousands of bets can be considerably more interesting.
For example, imagine an approach that historically produced a 6% return on investment across 5,000 selections and remained reasonably consistent across different seasons.
It doesn’t look particularly exciting.
But it may be far more credible than an angle showing a 45% return from 73 bets.
Good betting doesn’t need to be spectacular.
It needs to be repeatable.
Consider Micro Systems Too
Not every worthwhile system needs thousands of bets.
There is also a place for what I call micro systems.
These concentrate on very specific racing situations where there is a logical reason for an advantage to exist.
Perhaps a particular trainer has an established pattern when sending staying handicappers to certain courses.
Maybe another trainer repeatedly targets one meeting or performs particularly well with horses returning quickly after their previous race.
A micro system might only generate twenty or thirty selections a year.
That’s perfectly acceptable.
The important distinction is between a genuinely logical niche and a tiny sample created by repeatedly adding filters until the results look good.
One has reasoning behind it.
The other has hindsight.
Now Make It Face Reality
Once you’re happy with the historical results, stop changing the rules.
This part is important.
Write the system down exactly as it stands.
Then paper-trade it forward for a month.
Every time a future horse qualifies, record it.
Don’t add a new rule because the first three lose.
Don’t remove a race because you suddenly don’t fancy the selection.
Don’t conveniently forget one because the price looks too short.
Let the system operate exactly as it was designed.
This is the first time your system is being asked a question it hasn’t already seen the answer to.
That’s why forward testing is so valuable.
Historical testing asks:
‘What would have happened?’
Forward testing asks:
‘What happens next?’
Those are very different tests.
Don’t Panic if the First Month Loses
A month of paper trading doesn’t prove that a system works.
And a losing month doesn’t necessarily prove that it doesn’t.
You need to compare the forward results with what the historical data told you to expect.
If your historical strike rate is 12%, a run of fifteen losers shouldn’t come as a surprise.
If the historical system regularly experienced losing months, encountering one during forward testing isn’t evidence of failure.
This is another reason why understanding the characteristics of your system is so important.
You need to know what’s normal.
Then Start Small
If the logic makes sense, the historical evidence is convincing and the system behaves reasonably during forward testing, you can consider using real stakes.
But there’s absolutely no reason to start aggressively.
Use small, sensible stakes.
You’re still gathering information.
After another hundred bets, compare the live results with your historical expectations.
Is the strike rate broadly similar?
Are the odds similar?
Are the losing runs within the expected range?
Are the types of horses being selected what you expected?
You’re not looking for identical results.
You’re looking for evidence that the original pattern still exists.
Keep Testing Your System
One of the advantages of using a tool such as the Inform Racing System Builder is that a saved system can be run against newer results as the database grows, as well as against upcoming declarations to identify current qualifiers.
That means testing doesn’t have to stop when you start betting.
Continue reviewing the evidence.
But resist the temptation to continually interfere with the rules.
A system that has three losing weeks doesn’t necessarily need fixing.
Sometimes nothing is broken.
Sometimes you’ve simply experienced three losing weeks.
Knowing the difference is one of the hardest parts of systematic betting.
A Simple Checklist Before You Bet a System
Before putting real money behind a betting system, you should be able to answer these questions confidently.
Can I explain why the system should work?
Are the rules simple and logical?
Is the historical sample large enough to mean something?
Has it performed reasonably across different seasons?
Is the profit spread across plenty of selections rather than dependent on a handful of extraordinary winners?
Have I checked strike rate and losing runs as well as profit?
Would the required prices have realistically been obtainable?
Have I tested the exact rules on future races without changing them?
And perhaps most importantly:
Would I still be comfortable following this system after ten consecutive losers?
If the answer to that last question is no, you probably haven’t finished your research.
Final Thoughts
Building a betting system isn’t difficult.
Building one you can trust is much harder.
Start with an idea you can explain in a sentence. Keep the number of rules low. Test it across plenty of historical races and several different seasons. Examine where the profits came from. Check SP and BSP. Understand the strike rate and potential losing runs.
Then stop tinkering and make it face races that haven’t happened yet.
Paper-trade it forward.
If it survives that stage, start cautiously and continue recording the results.
The Inform Racing System Builder can make the research considerably quicker by allowing ideas to be tested against historical racing data and then used to find qualifying runners in future declarations. But software can’t decide whether an idea makes sense.
That’s still your job.
And perhaps that’s the most important lesson in system building.
Don’t build a system that perfectly explains yesterday. Build one simple enough, logical enough and robust enough to have a chance of working tomorrow.
this is a test comment