The most predictable thing in horse racing isn’t the horse.
It’s us.

My professional background, for anyone who hasn’t read my bio, has largely been in human performance.

Athletic performance. Health. Wellbeing. Behaviour.

For more than two decades, much of my work has involved trying to understand why people perform differently under different circumstances, physically, psychologically and environmentally.

Perhaps that’s one of the reasons I eventually fell so deeply into horse-racing data.

I’ve always been fascinated by patterns.

And horse racing might be one of the greatest pattern-recognition problems ever created.

But after thousands of races, systems and datasets, I’ve come to realise that there are actually two completely different pattern-recognition problems happening at the same time.

There are the patterns of the horses.

And there are the patterns of the humans betting on them.

One is extraordinarily difficult to predict.

The other can be surprisingly consistent.

The horse is only the beginning

Look at a racecard and everything appears beautifully measurable.

Official ratings. Weight. Draw. Distance. Going. Class. Recent form. Course record. Speed figures. Trainer. Jockey.

We naturally assume that if we gather enough of those variables and analyse them intelligently enough, the answer should eventually emerge.

But consider what happens when we put a human being on top of a half-tonne animal and ask them to compete against perhaps another dozen human-animal combinations travelling at 40mph.

Suddenly the number of possible interactions becomes enormous.

A horse misses the break.
Another becomes unsettled in the stalls.
The anticipated leader doesn’t lead.
Two jockeys contest the same position.
A horse gets boxed in.
Another races too keenly.
A jockey makes a tactical decision that nobody could have known beforehand.
The pace collapses.
The ground rides differently from its official description.
A horse simply doesn’t reproduce the performance its historical figures suggested it could.

None of that means the form was wrong.

It means the form was incomplete.

Because historical data describes what happened under previous circumstances.

A horse race asks what will happen under circumstances that have never existed before in precisely that combination.

That’s a very different problem.

Then consider what isn’t on most racecards

Here’s an example.

How many people who attended Goodwood or Royal Ascot during exceptionally hot weather looked beyond the traditional form and considered how individual horses had historically performed in heat?

What about recovery?
Hydration?
Travel?
Previous performance under comparable temperatures?
Whether particular bloodlines or individual horses appeared more or less tolerant of environmental stress?

We did.

And there’s an entire article coming on that research because what we found deserves proper examination rather than a throwaway paragraph here.

But that’s not really my point.

My point is this:

How many variables aren’t we measuring? And, even more importantly, how many variables can’t we measure?

That is what makes racing such a fascinating analytical problem.

Humans are complicated enough.

Put a human on a horse and introduce competition, tactics, weather, surfaces, physiology, market expectations and twelve other moving participants.

Complexity multiplies very quickly.

And yet we demand certainty

This is where the psychology becomes fascinating.

Despite all that uncertainty, humans desperately want a conclusion.

We don’t particularly enjoy:

“There isn’t enough evidence to know.”

We prefer:

“This is the one.”

And modern technology is becoming exceptionally good at giving us that feeling.

Ratings give us numbers.
Tipsters give us selections.
Algorithms give us probabilities.
AI gives us beautifully constructed explanations.

But confidence in an answer and accuracy of an answer are not the same thing.

And that’s where bookmakers have held an extraordinary structural advantage for generations.

The bookmaker doesn’t have to know the winner

The phrase “the bookies always win” obviously isn’t literally true.

Punters win races. Some win consistently. Professional betting operations exist. Markets get things wrong.

The important distinction is structural.

A bookmaker doesn’t need to predict every individual outcome correctly.

Its business exists across enormous numbers of events, with pricing, margin, risk management and repeated participation working over time.

The individual experiences racing differently.

Because we have emotions.

And emotions have patterns.

Winning and losing both change our decisions

A run of winners creates confidence.

Sometimes deservedly.

But confidence can grow faster than the evidence supporting it.

The next opportunity begins to look slightly better.

Risk can feel slightly smaller.

A race that might previously have been ignored suddenly becomes interesting.

Losing creates a different pressure.

Loss aversion is powerful. Humans dislike surrendering something they previously possessed.

So after a sequence of losses, the objective can subtly change.

Instead of asking:

“Is this next opportunity genuinely attractive?”

we can begin asking:

“Can this get me back?”

That’s an incredibly important distinction.

Because the next horse has no idea what happened in the previous race.

Your account balance doesn’t change its physiology.
Your last winner doesn’t improve its draw.
Your last loser doesn’t change today’s pace.

And yet previous outcomes can change our next decision.

That makes human behaviour almost another racing variable.

The invisible metric

This is the part I’ve become particularly fascinated by.

Imagine adding one more line to every racecard.

Alongside:

Going
Distance
Class
Draw
Pace
Form
Market

we add:

You.

What happened in your previous race?
Are you winning today?
Are you losing?
Has your confidence increased?
Are you trying to recover something?
Would you still make exactly the same decision if the previous five races had never happened?

No racing database normally contains that information.

Yet it can materially alter the decision being made.

In that sense, the bettor is part of the model.

And unlike almost every other variable in horse racing, it’s one of the few variables over which we have meaningful control.

This is where Elite Pass has become interesting

None of this has reduced my interest in finding predictive information.

It’s increased it.

But it has changed what we’re looking for.

When we started Elite Pass, the obvious ambition was the same question everybody asks:

Can we get better at identifying what happens next?

After analysing thousands upon thousands of races, we’ve become increasingly interested in a more nuanced question:

Can we identify the circumstances in which a race becomes more predictable, or more unstable, than the market might ordinarily suggest?

That distinction has become central to our research.

And recently, we’ve begun making progress.

EP Rank is starting to show us something

We’ve developed our own within-race scoring metric, EP Rank.

Importantly, EP Rank isn’t designed to be another colourful number telling people which horse to back.

That would be easy.

We’re interested in something harder.

We want to understand separation.

How strongly does one runner separate from another when multiple pieces of information are evaluated together?

What happens when that separation is large?

What happens when the leading horses compress together?

Do particular race types behave differently?

Are there configurations where the apparent hierarchy repeatedly proves stable?

And, perhaps more interestingly, are there configurations where the hierarchy repeatedly breaks down?

Early work is beginning to isolate races that appear to behave differently.

That’s exciting.

But it isn’t a victory lap.

It is the beginning of another test.

Because the next question is whether additional independent variables can improve that discrimination further.

Pace.
Race structure.
Course.
Distance.
Conditions.
Market behaviour.
Environmental factors.
Potential behavioural instability.
And other variables we’re currently testing.

If those layers repeatedly improve the probability of identifying particular race outcomes on genuinely unseen races, then we have something worth discussing.

If they don’t, they join the growing pile of racing ideas that looked compelling but failed validation.

That’s the deal.

The closer we get, the less I want to bet

This has been one of the strangest consequences of the entire Elite Pass project.

You might assume that analysing more racing data makes somebody want to bet more.

For me, the opposite has happened.

The research has probably been the greatest betting deterrent imaginable.

Not because I believe racing analysis doesn’t work.

Because I’ve developed considerably more respect for how high the bar should be before we claim that something does.

I’ve watched attractive theories disappear when tested against enough races.

I’ve seen variables that appear predictive until price is introduced.

I’ve seen extraordinary subsets that look almost impossible to explain, until they stop working.

And I’ve seen how easy it is to find a story after an outcome has occurred.

The more you understand uncertainty, the less appealing marginal decisions become.

That isn’t pessimism.

It’s discipline.

Losing is not evidence that the next race owes you something

This might be the single most useful lesson racing research has taught me.

Loss isn’t an anomaly that needs correcting.

Loss is mathematically inevitable in any probabilistic system.

Even if we eventually identify a subset of races with a significantly elevated probability of a particular outcome, some of them will lose.

They have to.

That’s what probability means.

The psychological danger begins when an inevitable statistical event becomes an emotional one.

Then behaviour changes.
Stake changes.
Selection criteria change.
Patience changes.

The research hasn’t failed.

The human has changed the experiment. And suddenly the most predictable variable in the entire system might be us.

Perhaps the real edge has two sides

This is where I think the Elite Pass journey is becoming considerably more interesting.

We’re trying to understand two things simultaneously.

Can we get better at measuring uncertainty in the race?

And:

Can we get better at recognising uncertainty in ourselves?

EP Rank and the research surrounding it address the first question.

Discipline addresses the second.

Neither requires pretending that racing can be made certain.

Quite the opposite.

A genuine edge, if one exists, should survive because we respected uncertainty rather than ignored it.

Perhaps eventually we identify a small subset of races where multiple independent measurements align and the probability of a particular outcome becomes materially stronger.

That’s what we’re trying to discover.

But equally valuable may be identifying the enormous number of races where they don’t.

Because sometimes the smartest racing decision isn’t finding the winner.

It’s recognising that you don’t have enough information to justify trying.

So why am I more excited about Elite Pass than ever?

Because we’re getting closer to asking the right questions.

Not:

“Which horse wins?”

But:

“Which races deserve our attention?”

Not:

“Which system produced the biggest historical profit?”

But:

“Which relationships survive when exposed to races they have never seen?”

Not simply:

“Which horse ranks first?”

But:

“How meaningful is the separation between first, second and the rest of this particular field?”

And ultimately:

“Can we identify instability before the race rather than explain it afterwards?”

That last question is where much of our current work is heading.

I don’t know where it ends.

That’s precisely why it’s research.

One final experiment

Next time you’re looking at a race, write down your decision before it starts.

Then add one question:

“Would I make exactly the same decision if I knew nothing about the results of my previous five bets?”

If the answer is no, you’ve just discovered a variable that isn’t in the Racing Post, Timeform, the betting market or EP Rank.

It’s sitting on your side of the screen.

And perhaps that’s the beautiful contradiction at the centre of racing.

We are using increasingly sophisticated technology to understand one of the most complicated competitive environments imaginable.

We’re measuring horses.
Markets.
Geometry.
Performance.
Conditions.
Behaviour.

We’re beginning to isolate races where our own scoring suggests the structure may be unusually stable, or unusually vulnerable.

And we’ll continue adding variables and trying to destroy our own conclusions before we believe them.

But throughout all of that sophistication, one of the most valuable pieces of information may remain remarkably simple:

Understand the race.
Understand the probability.
And understand yourself.

Because the horse doesn’t know what happened in the last race.

The market doesn’t care what you need to win back.

And probability has no memory.

The smartest edge may ultimately be knowing when the numbers deserve your attention, and having the discipline to do nothing when they don’t.

The Elite Pass research continues

EP Rank is now helping us investigate whether measurable separation between runners can identify races with meaningfully different characteristics. The next stage is testing whether additional independent variables strengthen that signal, or destroy it.

Both outcomes matter.

We’ll publish what survives.

We’ll publish what fails.

And there’s another piece coming soon on something most racegoers probably didn’t consider during those scorching days at Goodwood and Royal Ascot:

Does extreme heat change equine performance, and can the effect actually be measured?

That one became considerably more interesting than we expected.

Elite Pass Racing.
We don’t need every race to be predictable. We’re trying to discover whether some races are more predictable than others.

James Kennedy
Founder, Elite Pass