John-Z's Overlay · Trifecta ATM

Thu, Apr 30

Race 10

StMatthwsB200k

Post: 5:50 · 6f · Dirt · $200K purse · 44 runners

New — Ask Claude Opus about this race Why a horse rates the way it does, what the trifecta ticket means, where the public might be wrong.

Pace shape

44 runners

Hot pace

32 early types pressing for the lead — expect a fast pace, favors closers (P/S).

8 E — pure early 24 EP — early/presser 12 P — presser

Projected speed

Trifecta shape

Wide open

Lottery shape — low hit rates, larger payouts when they cash. Entertainment-only; the model is not confident in any structure.

3-horse box on posts 3

The model's most-likely-to-cash structure (informational — not a bet recommendation).

Cost

$1.00

1 combos

Hit prob

0.0%

simulated

Typical payout if hit

$0.00

mean $0.00

Expected ROI

-100.0%

net $-1.00

  • · Most-likely-to-cash structure: 3-horse box on posts [3] — hits 0.0% of simulated runs
  • · At $1 stake, a typical cash pays ≈ $0 (mean $0; trifecta payoffs are heavy-tailed). Estimated from the public pool, validated against archive payoffs.
  • · Wide-open shape — anyone could hit; longshots have room.

A Top picks

3, 3

B Contenders

3, 3, 9

C Value-edge longshots

Other ticket structures (3)
Ticket Cost Hit prob Typical payout ROI
4-horse box $1.00 0.0% $0.00 -100%
5-horse box $4.00 0.0% $0.00 -100%
6-horse box $7.00 0.0% $0.00 -100%

Hit prob is the model's per-race PL distribution. "Typical payout" is the median cash, weighted by the public pool's odds (not the model's own picks) and calibrated to actual historical trifecta payoffs — so it reflects what the tote really pays, not the model's disagreement. Walk-forward backtests show pari-mutuel takeout (~16-26%) is too steep to consistently clear at this data scale — these are tipsheet outputs, not bet recommendations.

Top-4 ITM concentration: 14% · Top 4 cover 0.42 of the 3.00 expected ITM finishers

Field

Click any horse name for the per-horse breakdown (last-10-races chart, SHAP feature contributions, full PP table).

PP Horse ML Pred Close Pred Fig Win ITM% Edge
3BPBE YOU4.0-14.1-193.44%11%-40%
3BPBE YOU4.0-14.1-193.44%11%-40%
3BPBE YOU4.0-14.1-193.44%11%-40%
3BPBE YOU4.0-14.4-193.33%10%-41%
9BEMAXIMUM BOURBON4.5-13.4-190.23%9%-38%
9BEMAXIMUM BOURBON4.5-13.4-190.23%9%-38%
9BEMAXIMUM BOURBON4.5-13.4-190.23%9%-38%
1ECONCRETE GLORY8.0-16.7-190.13%8%-20%
1ECONCRETE GLORY8.0-16.6-190.13%8%-20%
1ECONCRETE GLORY8.0-16.6-190.13%8%-20%
1ECONCRETE GLORY8.0-16.6-190.13%8%-20%
9BEMAXIMUM BOURBON4.5-13.5-190.53%8%-38%
8EPBOURBON BASH10-113-190.33%8%-15%
8EPBOURBON BASH10-113-190.33%8%-15%
8EPBOURBON BASH10-113-190.33%8%-15%
8EPBOURBON BASH10-113-189.92%7%-16%
7EPWENDELSSOHN12-116-190.62%7%-12%
7EPWENDELSSOHN12-116-190.52%7%-12%
7EPWENDELSSOHN12-116-190.52%7%-12%
7EPWENDELSSOHN12-116-190.52%7%-12%
6EPBACK EM UP6.0-17.0-188.72%7%-30%
6EPBACK EM UP6.0-16.9-188.62%7%-30%
6EPBACK EM UP6.0-16.9-188.62%7%-30%
6EPBACK EM UP6.0-16.9-188.62%7%-30%
2EPBUILT3.0-13.7-189.02%6%-57%
2EPBUILT3.0-14.0-188.82%6%-58%
2EPBUILT3.0-14.0-188.82%6%-58%
2EPBUILT3.0-14.0-188.82%6%-58%
10EPJACK'S PROMISE8.0-19.5-187.62%6%-23%
10EPJACK'S PROMISE8.0-19.5-187.62%6%-23%
10EPJACK'S PROMISE8.0-19.5-187.62%6%-23%
10EPJACK'S PROMISE8.0-19.5-187.62%6%-23%
5EPLOFTY ADJUDICATOR30-145-187.92%6%·-3%
5EPLOFTY ADJUDICATOR30-145-187.92%6%·-3%
5EPLOFTY ADJUDICATOR30-145-187.92%6%·-3%
5EPLOFTY ADJUDICATOR30-145-187.92%6%·-3%
4PBILLAL20-134-184.62%5%-7%
4PBILLAL20-135-184.72%5%-7%
4PBILLAL20-135-184.72%5%-7%
4PBILLAL20-135-184.72%5%-7%
11PAIR INVASION30-157-180.21%4%·-5%
11PAIR INVASION30-157-180.21%4%·-5%
11PAIR INVASION30-157-180.21%4%·-5%
11PAIR INVASION30-157-180.21%4%·-5%

Sort the table by clicking column headers. Edge = our predicted ITM% minus the public-implied ITM% (live odds when present, ML otherwise); green ▲ ≥ +5pp, red ▼ ≤ −5pp. See How to Use for the math.

Ask Claude Opus

Questions about the model's predictions for Race 10 · Thu, Apr 30 · Churchill Downs. Powered by Claude Opus 4.7.

Ask anything about the field, the trifecta ticket, or why the model rates a horse the way it does.

Limited to 30 questions per day per visitor. Answers are model-generated and may be incorrect — not betting advice.