Rebuilding Brohamer from the Book

Tom Brohamer's Modern Pace Handicapping gets name-dropped in every serious handicapping conversation. Feet-per-second. ESP. Turn-time. Energy distribution. The vocabulary is everywhere.

Almost nobody who cites the book has actually rebuilt it.

We did. Chapter by chapter. No porting forward from older models, no shortcuts, no "inspired by." Each addition gets benchmarked before we trust it. Here's what we learned.

Why a clean-room rebuild

The tempting move when you already have a working handicapper is to bolt new ideas onto it. Layer a feature, A/B it, move on. That's how models accumulate cruft: features that sound like what the book says, but drift further from the source with every revision.

So we started over. The rule is simple. Build from the book, not from the last model. Keep source traceability explicit. Benchmark after every meaningful change. Separate ranking logic from wagering judgment.

That last rule matters more than it sounds. Most public handicapping models conflate two different jobs: identifying which horses can run, and deciding whether to bet. We'll come back to that.

The spine

Brohamer's framework isn't a single formula. It's a pipeline. Here's the live path, stage by stage.

Feet-per-second (Chapter II). The foundation. Convert raw fractional times into velocity. Sounds trivial. It isn't. Beaten-lengths adjustments, run-up distances, and timing-point conventions all change the number. Get this wrong and every downstream layer compounds the error.

ESP, for Early, Sustained, Presser (Chapter III). Running-style classification, position-first. The book is emphatic. A horse's position in the early fraction is the primary signal, not its raw early speed figure. We had to be strict about this. "False early" horses, the ones with fast early speeds but no usable position, used to carry forward like real early types. They no longer do. That single change moved benchmarks.

Turn-time (Chapter IV). Middle-fraction strength, but par-relative. Generic middle-fraction numbers are noise. The comparison that matters is against the par for today's distance, surface, and class. Switching from generic to par-relative was a real lift.

Pacelines (Chapter V). The Sartin layer: comparable-line selection. Which past performance do you actually use? Same surface, same distance bucket, same class context, repeatable. Sprint-converted-route lines lose tie-breaks to true sprint lines. The last usable comparable line is harder to displace than the freshest impressive-looking one.

Decision model (Chapter VII). How the comparable lines combine into a contender screen. Negative class drops, severity ordering, false-contender flags.

Track profile and energy distribution (Chapters VIII and IX). What kind of energy pattern wins at this track at this distance? Front-loaded? Even? Late?

Practical worksheet (Chapter XIV). The bet itself. One horse, two, or pass.

What moved the needle

Here's where we owe you honesty.

Some chapters delivered real benchmark gains. Some became more literal to the book without improving results.

What helped:

Position-first ESP. False early horses no longer treated as legitimate early types.
Par-relative turn-time. Replaced generic middle-fraction strength.
Sartin paceline discipline. The last usable same-surface, same-bucket line is hard to displace. Sprint-converted-route lines demoted in tie-breaks.
Chapter V worksheet outputs. Explicit four-horse and two-horse screens, not just a single ranked column.
Chapter XIV one-horse vs. multi-horse split. Real decision branching in the practical layer.

What didn't move the needle, despite being more faithful to the book:

Track profile bucket tuning (Chapter VIII). Shorter profile windows, exact-distance-first construction, profile-led selectors. All more literal. None of it shifted benchmark rank performance.
%E target construction (Chapter IX). Recent-winner samples only, trimmed acceptable ranges, contextual targets. Same story.
Decision-model worksheet refinements (Chapter VII). Cleaner race selection, segmentation tweaks, profile-fit ordering. Closer to the book. Didn't help the rank line.

The honest read: these chapters are now much closer to the source than they were. But fidelity-to-the-book and benchmark performance are not the same metric. We chased one for a while assuming it would deliver the other. It didn't.

What the current model actually does

On our 2025 benchmark, the live model lands at:

21.4% top-pick win rate
38.2% winner in top 2
65.8% winner in top 4
−19.6% flat-bet ROI on the top pick

That last number deserves its own post, and it's getting one. Picking winners 21% of the time isn't the goal. The goal is identifying mispricings, and a model that picks favorites well will reliably lose money against the takeout. We'll come back to that.

What we got wrong about Brohamer

We assumed the gap between book and benchmark was in the late chapters. That we just hadn't been literal enough about energy distribution and track profile.

That was wrong. The late chapters are now close to the book. The remaining gap is in judgment: when Brohamer removes an obvious horse despite a strong figure, when he keeps a marginal horse alive for multi-horse play, when price changes the action without changing the preferred horse.

Those decisions don't live in any formula. They live in his example races. The next round of work is example-level translation, not more chapter enablement.

Why we publish this

Most handicapping content is sales copy in disguise. A magic formula. A proprietary edge. A "we know something you don't" pitch.

We're not interested in that. The interesting work is hard, slow, frequently disappointing, and only legible when you publish your benchmarks. Every chapter we shipped, we measured. Every change that didn't help, we said so.

If you want a sharper handicapping product, that's the kind of work behind it. If you want a magic formula, you should be skeptical of anyone selling one. Including us, if we ever start.

Trifectly is built on a chapter-by-chapter implementation of Tom Brohamer's Modern Pace Handicapping, layered with a contender screen and a value decision model. See it in action →

Next in the series: Find Live Horses First. Rank Them Second.