Research period August 2026 · James Kennedy, founder

Research notes

What price control did to our best looking ideas

I used to believe the draw was worth money. Not as a hunch, as a settled fact I would have argued about in a bar. It was the first thing we tested properly and it was the first thing to go.

What I believed, and why it was reasonable

Low draws at certain tracks over certain trips win more than high draws. That is not superstition. You can pull the data yourself in an afternoon and watch it come out of the numbers, clean and obvious, exactly as promised.

So the inference feels automatic. There is a physical advantage, the advantage shows up in the results, therefore there is money in knowing about it. I held that view for a long time and I do not think I was being stupid. Almost every racing idea I have ever been sold arrives in that shape, because the shape is genuinely persuasive.

The step that is missing is so small it is easy not to notice. Between the advantage existing and the advantage being worth money, there is a market.

The control that kills things

Here is the test that changed how I think about all of this, and it is embarrassingly simple.

Take your idea. Now compare horses that share your signal against horses that do not, holding the final price constant. Same odds, differing only on the thing you believe in. If your signal carries information the market has not already absorbed, the horses that have it should outperform the horses that do not at identical prices. If it does not, the advantage was real but the price already contained it.

That one control is responsible for almost every null result we have published. It is also why raw historical data is so dangerous. Raw data will happily show you that a factor is associated with winning. It will not tell you that the association is already sitting in the odds, because the odds are not in your spreadsheet unless you put them there deliberately and then refuse to look away.

We settle everything against Betfair starting prices for exactly this reason. It is not a formality. It is the only step that separates an observation from a claim.

Three ideas that did not survive it

The draw first, since I had the most invested in it. Taken on its own, on held-out races we had not looked at while building anything, it returned a clean null. Not weak. Not promising but unproven. Null. The draw effect is real and the price has it.

Then money. The idea that a horse being backed tells you something the number alone does not. We took horses that had shortened and compared them against horses that had drifted, at the same final price, and the advantage went away. What looked like information about a horse turned out to be information about where the price had ended up, which we already had, because we were holding it constant.

Then form, which is the one people find hardest to accept. Strong recent form genuinely predicts. It really does, and anyone telling you otherwise is being contrary for sport. The trouble is that by the time the stalls open, almost all of what it predicts is in the price. Everybody can read a form book. That is precisely the problem with using one to find value.

Two hundred and twenty seven, and none

The corpus underneath all this is 71,626 races and 633,474 runner records, with 1,180 races carrying a timestamped market receipt so we can see what the price was doing rather than only where it finished.

Against that, 227 separate racing systems and theories have been analysed. Zero have been validated.

I put that number on the website deliberately, because it is the one thing on there that nobody else in this industry would publish. Two hundred and twenty seven honest attempts, every one of them somebody's confident idea, and not one of them survived a proper price control. Some fell apart immediately. Several looked excellent for a while, which is far more dangerous, because an idea that dies quickly costs you a week and an idea that dies slowly costs you a year.

A validated claim would mean a real effect that survives holding the final price constant. We have not produced one.

Why this is not the defeat it sounds like

Because I know where the edge is not, and that turns out to be worth a great deal.

Every one of those 227 ideas is a road somebody is currently driving down. People are building products on the draw right now. People are selling market move alerts right now. There is a version of me who spent the next three years refining a draw model, getting better and better at extracting a signal the price had already extracted, and feeling productive the entire time.

Knowing where the edge is not is how you stop wasting years. It is not a consolation prize for failing to find an edge. It is the actual output, and it compounds, because each closed door means the remaining questions are better questions than the ones you started with.

It also does something to how you read everything else. Once you have watched five persuasive ideas evaporate under one control, you develop a useful reflex. Show me it at the same price. That question is free to ask and it disposes of most of what racing sells.

Where it leaves us looking

Somewhere less crowded. The question of which horse is best has had decades of serious effort and enormous amounts of money thrown at it, and the price you see is the aggregate of all that work. Across 60,671 races with one clear favourite, it won 35% of the time and did not win 65%, which is roughly what those prices were saying. The market is not being lazy. It is being very good at a question a lot of people have already answered.

So we ask a different one. Not which horse is best, but where a particular race stops being legible, and what about the way it will be run makes it so. Pace, position, shape, and the state a horse is actually in when the stalls open. Harder to price is not the same as unknown, and it is the only ground I can see that has not already been worked over by better resourced people than me.

I might be wrong about that too. If I am, we will publish it, in the same place and the same detail as everything above. That is the arrangement.

Related research

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.