The backtest that already knew the answer
If a betting system tests beautifully, the first question worth asking is not how it works. It is what the test knew that the punter would not have.
Here is a thing that happens, and it happens to careful people. You want to know whether the way a horse ran last time tells you anything about how it will run today. Reasonable question. So you gather the written comments that describe each run — led and kept on, held up and never got into it, hampered two out — and you check whether those descriptions line up with the results.
They line up perfectly. Horses described as running on strongly win a great deal more often than horses described as weakening. The relationship is enormous and it holds everywhere you look.
It is also completely worthless, because the comment was written after the race. “Led, ran on well” is not a clue about the result. It is the result, in sentences. You have discovered that horses which won tend to be described as having won.
Why it is not as obvious as it sounds
Written out plainly like that, nobody would make the mistake. The trouble is that it almost never arrives written out plainly. It arrives as a column in a spreadsheet, sitting beside forty other columns, and the only thing separating the honest ones from the poisoned one is knowing the exact moment each was recorded.
A finishing position is safe if it belongs to a previous race and fatal if it belongs to this one. The same field, the same name, the same format. Add a starting price and you have another: the price at the off knows things the price that morning did not. Add an official rating and you have a third, because ratings get revised after the race that prompted the revision.
None of these announces itself. The test runs, the number comes back strong, and the strength is exactly what convinces you not to check.
The habit that stops it
We have one rule, and it has saved us more than any clever method has: every piece of information carries the time it became knowable, and nothing may be used to judge a race unless it was knowable before that race started.
Not “we were careful”. Not a review at the end. The check refuses the data outright and says why. If a comment is dated the same day as the race it is being asked about, it does not get used — there is no setting to turn that off, because there is no legitimate case for it.
It feels like bureaucracy right up until the morning it refuses something you were about to be pleased with.
What this means if you are reading somebody's results
You will see systems advertised on a backtest. Most of the time you cannot inspect the data, so the useful question is not “how good is the return” but “what did the test know, and when did it know it?”
A few things are worth asking. Was the selection fixed before the race or chosen from the finished card? Are the prices the ones you could actually have taken at the time? Was the rule written before the data was examined, or found by looking? And the quiet one: how many versions of the rule were tried before this one was shown to you?
None of that means a result is dishonest. Most of these mistakes are made in good faith by people who want the answer to be yes. That is precisely why the guard has to be mechanical rather than a matter of being careful, and why the most useful thing a test can do is refuse.
Check today’s races
See how Elite Pass reads today’s eligible races.
Related research
More notes
- Big fields and what they do to a favouriteMore runners means more ways to be beaten. That part is true. The awkward part is that the market worked it out before we did.
- What is the favourite-longshot bias in horse racing?Bettors have long taken too little of the favourite and too much of the outsider. Economists have studied it since 1949. What our 60,671-race study found.
- Are horse racing betting markets actually efficient?A beaten favourite is not a wrong price. Efficient means three different things, and confusing them is how racing arguments go wrong.
Elite Pass publishes research, not guaranteed outcomes. Findings are measured against the market and remain subject to replication and sample size. The evidence room and the public ledger carry the full record. All insights.