Find Live Horses First. Rank Them Second.

Most amateurs handicap by ranking. They look at every horse, weigh figures and form and trip notes, and produce a 1-to-N order for the whole field.

That's the wrong question.

The right question, the one sharp handicappers ask first, is much narrower:

Which of these horses can actually run a competitive race today?

The answer is almost never "all of them." In a 10-horse field, the answer is usually four or five. Sometimes three. The first job of a handicapper isn't to rank horses. It's to find the contenders. Ranking comes after.

Why this matters

If your contender pool is wrong, everything downstream is noise.

Picture a 10-horse race where the four live horses are #2, #5, #7, and #9. You build a contender set of #2, #5, #6, #8. You picked four horses. Three of them can't win today. It doesn't matter how cleverly you rank them. The winner is sitting outside your screen.

We see this in benchmarks of older models. The top-pick win rate looks fine. But the false-contender leakage, the rate at which non-contenders end up in premium slots, quietly eats the ROI. The model is technically picking horses. It just isn't picking from the right pool.

In our current Brohamer-based system, 76.6% of winners survive the contender screen at an average pool size of 5.3 horses. Said another way: more than three out of four races, we're picking from roughly a third of the field. That isn't a side effect. It's the architecture.

What a contender actually is

A contender is a horse that's live to finish in the competitive cluster. Not necessarily the winner. Not necessarily an overlay. Live.

For diagnostic purposes, we label horses against the competitive anchor of each race:

A: winner, or within a length of the competitive anchor
B: 1 to 2 lengths back
C: 2 to 4 lengths
D: 4 to 8 lengths
X: 8+ lengths

The contender screen exists to find A's and B's, accept C's as backup coverage, and avoid promoting X's. That's it. It's not trying to identify the winner. It's trying to keep the competitive cluster in scope.

This sounds like a subtle distinction. It isn't. "Find horses that can run a competitive race" is a different problem from "rank all horses by win probability." Conflating them is the source of most amateur model rot.

Four sub-questions

We break the contender screen into four stages.

1. Eligibility

Is this horse even eligible to be treated as a serious contender today?

This is the first hard gate. Inputs are simple and explainable:

Comparable speed evidence
Comparable pace evidence
Class belonging
Distance and surface plausibility
Structural fragility flags

This stage is allowed to reject horses. No comparable race fast enough. Pace profile that can't survive today's shape. Clearly below class for this race family. Severe distance or surface mismatch with no supporting evidence. Out.

Most amateur models won't reject anything outright. Every horse gets a number. That's the trap.

2. Competitive ability

Of the eligible horses, who has enough raw ability to land in the live cluster?

This is where speed and pace belong, but not as one blunt average. The right question isn't "who has the strongest recent figure?" It's "who has shown repeatable, race-relevant ability for today's context?"

Comparable beats generic. Repeatable beats peak. Race-relevant beats impressive-looking.

A horse with one big number and four ordinary ones is not the same as a horse with five consistent solid numbers, even if the peak figure is higher. Peak-chasing is one of the most expensive mistakes in handicapping.

3. Fragility

What makes this horse likely to collapse even if its raw ability looks credible?

This is the missing layer in most amateur models. Fragility isn't the opposite of speed. It's a separate dimension. A fast horse can be fragile. A slow horse can be reliably consistent. They're independent variables.

Examples:

Peak-dependent figures (one big number, can't repeat)
Need-the-lead without tactical control
Pressure-sensitive pace
Wide-trip vulnerable closer in a big field
Route form masquerading as sprint ability

This stage gets its own post. It's that important.

4. Contender tiering

What kind of contender is this horse?

The output isn't one scalar. It's a tier:

Prime contender: should be at the heart of any ticket
Live contender: solid second-tier, worth covering
Underneath contender: vertical-bet coverage only, not a win bet
False contender: fast enough on paper, too fragile or mismatched to bet

Downstream layers consume tiers, not raw scores. That gives the rest of the pipeline permission to refine among primes, use underneaths differently, and stop promoting false contenders into premium slots.

Field-size matters

Contender selection should not work the same way in a 6-horse field as in a 12-horse field.

Large and deep fields require a stricter policy:

Trip fragility matters more
Generic speed counts less unless it's survivable under pressure
Pace and field navigation matter more
Tactical dead-ends should be demoted earlier

This is one of the clearest lessons from our diagnostics. The leak rate in big fields was the loudest signal in our backtests, and the fix wasn't a better ranker. It was a stricter screen.

What doesn't define a contender

These factors can influence later ranking. They should not do the primary job at this stage:

Generic trainer or jockey edges
Betting market rank
Price or value logic
Explanation-only pattern flags
Small additive bonuses from many weak factors

If a horse isn't a real contender on ability, survivability, and class belonging, no trainer angle or market signal should rescue it into a premium slot. Those layers exist to refine among legitimate contenders, not to manufacture them.

The order is the point

Identify live horses. Then rank them. Then decide whether to bet.

Three stages. Three different questions. Most amateur handicapping collapses them into one, a single "who do I like?" ranking, and that's where the EV leaks out.

We do it the other way. Screen first. Tier second. Bet third. The screen is doing 76% of the work. The tier sets the structure. The bet decision is its own discipline, because picking the right horse and making the right bet are also not the same problem.

That's the post after next.

Trifectly screens, tiers, and prices every contender in every race we cover. See today's cards →

Previous: Rebuilding Brohamer from the Book · Next: Fragility Is Not the Opposite of Speed