Modern race analysis · Built 2026

Two picks.
One coach.
Your record.

A top pick and best-value selection on every US thoroughbred race. A weekly coach who reads your record. And a research workspace for the questions that don't fit on a card.

Start 30-day free trial → See a sample brief

✓ $29/mo after. Cancel any time. Full refund within 7 days of being charged.

Sam · weekly brief
"You are not chasing losses. The leak is on the favorites. One rule for this week…"
Tested in the open On 5 years of races our model never saw during training
+6.0pts
More accurate than the public favorite
+21%
Theoretical return on best-value picks
5/5
Test years where the model held up
See the full results →
A note from the founder
I spent three years getting beaten by the favorites and not knowing why. Every analysis tool told me which horse to pick; none told me what was actually wrong with the way I was picking. So I built Derbee — a model that tests itself in the open and a coach that reads your own record. The product I wish I had had.
Founder · Derbee

One subscription, three real tools.

Picks for what to follow today. Coaching for what your record is teaching you. Research for the questions that don't fit on a card.

The picks

A top pick and a best-value selection on every US thoroughbred race, every day.

  • Top pick from a ranking model — tuned to identify the actual race winner
  • Best-value pick — finds horses the market has underrated
  • First-and-second pick when the model thinks two horses will finish 1-2
  • Live odds, race status, and a plain-English brief on every race
CD R3 · Maiden Special Weight
Top pick: #4 Storm Atlas · 34% · 5-2
Best value: #7 Quill Cat · +8% · 9-1
Two-horse pick 4/7 — model sees $26 fair return on $12.

Sam, your coach

Every Monday, Sam reads every selection you logged and tells you what worked, what didn't, and one rule for the week.

  • Honest analysis from your own pick history
  • Spots the patterns you can't see — chasing losses, picking the wrong favorites, misreading a track
  • Sets weekly goals you can act on next race day
  • Answers follow-up questions and drills into specific races on request
Week of May 18 · Coach brief
"Two weeks in, your accuracy on low-odds favorites is 28% — they win 32% of the time. The gap is on the favorites, not the long shots. One rule for this week…"

Research

When a question doesn't fit on a race card. Ask anything — about a horse, a trainer, a track, your own picks — and get charts and tables back, not just text.

  • Ask anything about a race, a horse, a trainer, a track
  • Charts and tables appear inline as part of the answer
  • Keep separate conversations going for each topic you're tracking
  • Drills deeper as you follow up
How has #4 Storm Atlas performed on a wet track this season?

Storm Atlas has 4 starts on a wet surface this season — 0 wins, 1 second, 3 unplaced. Average finish position 5.2 vs 2.8 on dry.

Storm Atlas — finish position by surface (2026)
Dry
2.8 avg
Wet
5.2 avg

Bottom line: Storm Atlas is materially worse in the wet. Today's track is listed sloppy.

What Sam actually says.

A real-shape weekly brief built from a real subscriber's pick history.

Coaching brief — week of May 18

You're not chasing losses. The leak is on the favorites.

Pick accuracy this week
28%
on 23 selections · model expected 36%
Your accuracy by odds range — the 5-2 to 3-1 range is the gap
8-5 to 2-1
42%
5-2 to 3-1
28% / 32%
7-2 to 6-1
21%
8-1 and up
11%
Solid bars: your accuracy. Hollow line: what the model expects. The gap on the second row is the leak.

Two weeks of your picks is enough to see one clear pattern: on horses with low odds (between 5-2 and 3-1), you're picking the winner 28% of the time. Horses at those odds win 32% of the time. You're picking the right favorites less often than the field does — but only in that one odds range.

You're not chasing losses. Your accuracy on horses with longer odds is right where the model expects. The gap is on the favorites.

One rule for this week: when your top pick is a low-odds favorite (around 5-2), check the best-value column on the same race. If they disagree, pause. Three races last week where that rule would have flipped a wrong call to a correct one.

How it works.

Three steps. No setup, no spreadsheets.

1

Sign up

30 seconds. Pick your home tracks.

2

Browse picks every race day

Top pick + best value on every race. Live odds, first-and-second picks, plain-English briefs.

3

Log your picks — Sam coaches

Track your own selections. Every Monday, Sam writes you an honest brief.

What Derbee won't do.

This space is full of products that promise winners and quietly retire when their predictions go cold. We're not one of them. Four promises we make to every subscriber.

  • Promise winners. Anyone in racing analysis who does is misleading you.
  • Hide our misses. When a model stops working on fresh data, we shut it down. Four were retired in April. The kill list is public.
  • Be a brokerage. We're an analytical research tool. Not a tip service, not a guarantee, not a place to put down money.
  • Sell your data. Your pick history is yours. We use it only to coach you.

Four ways in. One product.

From a single race card to year-round coaching — same picks, same honesty, your pace.

Free

Free
Create account
  • Log your bets, no limits
  • Yesterday's full briefs, every track
  • Triple Crown + Breeders' Cup previews

No card required.

Race Card

$9 one track, one day

11 races · first post 1:00 PM ET
  • Full Today briefs for one card
  • Live status, pick tracking
  • Same model, same honesty

Sales final. Auto-refund if the card is cancelled.

Most popular

Monthly

$29 /month
Subscribe monthly
  • Every track, every day
  • Sam — your weekly coach
  • Research — ask anything
  • You — full bet analytics
  • Cancel any time

7-day refund window. Cancel any time from Settings.

Annual

$290 /year — save $58
Subscribe annual
  • Everything in Monthly
  • Two months free
  • 14-day refund window

14-day refund window. Cancel any time from Settings.

Honest answers.

If you've read this far, you probably have a few of these.

Which tracks does Derbee cover?

Every US thoroughbred track running on any given day — roughly 25-30 race cards, including Aqueduct, Belmont, Churchill, Santa Anita, Gulfstream, Saratoga, and the regional circuits.

What is Derbee, exactly?

Derbee is an analytical research and educational tool for horse racing enthusiasts. We use machine-learning models to identify the top selection and best-value pick on every race, and we coach you on patterns in your own analysis history. We are an information service, not a brokerage.

How accurate are the picks?

We tested our models on 5 full years of races they had never seen during training. The top-pick model identified the winner 6 percentage points more often than the public favorite. The best-value model returned +21% on the picks it flagged. Full results on the proof page.

If I am new to racing, is the coach still useful?

If you log 10+ picks a week, yes. Below that, Sam will still write you a brief, but the patterns Sam spots need a meaningful sample of your own selections to be specific to you.

Can I try Derbee before subscribing?

Yes — $9 buys you one race card. That's a real day of the product.

What's your refund policy?

Race Card: sales final, but we auto-refund if a card is cancelled. Monthly: 7-day full refund window. Annual: 14-day full refund window. Cancel any time from Settings.

What happens if a card is cancelled (weather, etc.)?

Automatic refund the next morning. No support ticket needed.

Do you have any free content?

Yesterday's full briefs for every track — read what we said, see how it played out. Plus full coverage of every Triple Crown and Breeders' Cup race.

Try Derbee for 30 days, free.

Picks, a coach who knows your record, and a research workspace — all for one fair price.

Start 30-day free trial →
5 years of public test results 30-day free trial, full refund Cancel any time