Horse Algo V6

Entries monitor → · Updated 2026-09-12 09:01:01 · auto-refreshes every 5 min
Bet gate raw edge > 0 · turf/AW — the picks marked BET below grey line = all 580 published picks, gate ignored
◎ Closing-line value — does the model beat the market?
CLV = how often the pick actually won, minus what the final tote price said it would (de-vigged over the field). Closing odds come from DRF result charts, validated 55/55 to the cent against the official payoff.
Bet gate: +11.5 pts [-0.8, +23.7] over 54 bets — market priced them at 21.9%, they won 33.3%.
All bettable picks: +1.5 pts [-4.0, +7.0] over 241 bets — market priced them at 28.0%, they won 29.5%.
Read them together. Across all picks the model does not beat the close, and the market's price is the better probability estimate (Brier 0.192 vs 0.203). The gate is the subset where it might — on a sample whose interval only just clears zero, and whose ROI is front-loaded (+70.3% before Jul 3, +10.2% after).
Why show CLV at all: it resolves ~3× faster than ROI (≈546 bets vs ≈1,612 to settle the effect we see), because payoff variance doesn't swamp it. Coverage 50% of gate bets.
⚑ Off-the-turf races are in these numbers
22 of 109 gate bets (20%) and 86 of 580 published bets (15%) were scheduled on turf but ran on dirt. The entries feed never prints the new surface — across 234 such races not one snapshot ever said Dirt — so the model scored those horses on turf form and the surface gate passed them. 2 older bets predate surface tracking and are unclassified.
Gate with them removed: -6.5% over 87 bets (they themselves ran +35.5% over 22), versus +2.0% over 109 as shown above — so the headline is not propped up by them, but it is not a clean turf/AW record either.
They are also not dirt races: the scratches strip the field to the dirt-capable horses, so they come back chalkier than either surface (9.8% bombs ≥7-1 vs 16.0% native dirt, 21.8% native turf). They get their own Off-turf row in the Surface cut rather than being folded into either one. The Off-turf row in the Surface cut now has its own backtest baseline instead of being counted as Turf, so a losing run on “turf” can be traced to races that were never turf.
A lifetime dirt-speed feature was tried here and removed: it looked worth +16.9 ROI pts on off-turf races until a routine data re-pull reversed it to −0.5, and across all 4,332 backtest races it cost significantly more wins than it gained (McNemar p=0.004). The apparent gain was the comparison baseline being unusually bad on stale data, not the feature being good.
The switch is now partly visible before post. The feed hides the surface but not the scratches: off-turf races lose 34.8% of the field vs 14.3% for turf races that stay, and 89% of that is on the wire ≥30 min before the chart exists. Picks in races with ≥40% scratched carry an OFF-TURF RISK flag (~63% of flags do run on dirt). It is a warning, not a gate — still confirm the surface before betting a flagged race.
Flag scorecard, live: 78% precision (28 of 36 flagged turf bets did run on dirt), 45% recall (34 off-turf bets went unflagged) over 371 judged bets.
Live ROI
+2.0%
all picks -20.6%
Win Rate
30.3%
all picks 27.2%
Net P&L ($2 bet)
$+4
all picks $-239
Bets
109
of 580 published
Wins
33
all picks 158
Days w/ a Bet
65
of 105 tracked

Today's Picks — Sep 12, 2026

Fav
Tier
Bet when the pick is turf/AW and raw edge > 0. Skip when the model is on the favorite — those never clear the gate. On a turf bet, check the ⚑: a red ⚑ n/m SCR means ≥40% of the field is scratched and the race is likely coming off the turf — confirm the surface before you bet it. Fav/Tier narrow it further; the record on the right follows whatever you pick.
TrackRacePost#Horse MLWin % Edge Raw Fav Call
BTPR1 12:35 PM 3 Lightning Speed WATCH 1.6/1 36% +5.0% -3.1 On Fav PASS
BTPR3 1:35 PM 4 Going Streakin WATCH 2.0/1 31% 3.0% -5.2 On Fav PASS
BTPR5 2:35 PM 3 The Pretty One WATCH 1.6/1 36% +4.6% -7.0 On Fav PASS
BTPR7 3:35 PM 1 Little Juanita WATCH 3.5/1 23% 3.8% -0.6 Demote #2 PASS
BTPR8 4:05 PM 3 Moogie Son 3.5/1 23% +4.6% -1.9 #5 PASS
CBYR6 8:40 PM 3 Aero Star WATCH 1.6/1 28% 0.9% -5.9 Demote #2 PASS
CBYR8 9:40 PM 7 Auntie Vodka WATCH 1.6/1 39% +8.5% -3.1 On Fav PASS
CDR1 12:45 PM 1 Excel Calculator WATCH 4.0/1 16% 0.5% -2.5 Demote #2 PASS
CDR4 2:18 PM 10 Circle Tap WATCH 5.0/1 14% 1.4% -1.9 Demote #2 PASS
CDR6 3:21 PM 1 Usha WATCH 2.5/1 25% 2.6% -3.9 On Fav PASS
CTR8 10:32 PM 7 Starship Refund WATCH 5.0/1 18% +6.1% -0.0 Demote #2 PASS
CTR4 8:32 PM 7 Genuine Gomo WATCH 4.0/1 19% 3.0% -1.2 Demote #2 PASS
CTR5 9:02 PM 8 It's Sizzling Time WATCH 2.5/1 18% 0.4% -5.9 On Fav PASS
DELR4 2:11 PM 1 Cadet Corps WATCH 1.4/1 34% 0.2% -11.0 On Fav PASS
DELR5 2:42 PM 4 Ruse 6.0/1 14% +5.8% +0.3 #5 BET
DELR8 4:16 PM 9 Striking Sparks WATCH 4.0/1 16% 0.9% -3.7 On Fav PASS
DELR9 4:44 PM 9 Mopo 3.0/1 18% +4.5% -1.0 Demote #2 PASS
DELR10 5:14 PM 4 Miss Mo Town WATCH 3.5/1 20% 2.1% -6.4 On Fav PASS
FPR4 3:45 PM 7 Passing Passion WATCH 1.2/1 37% 1.3% -10.5 On Fav PASS
FPR9 5:50 PM 6 Hunts Promise WATCH 2.5/1 25% 2.0% -7.7 On Fav PASS
GPR8 4:04 PM 3 Hello Hello Hello 4.5/1 15% 0.8% -1.5 Demote #3 PASS
MTHR2 1:21 PM 1 She's a Bombshell WATCH 3.5/1 23% +6.1% -1.2 Demote #2 PASS
MTHR7 3:56 PM 7 Pivotal Moment 2.5/1 14% 0.6% -3.6 On Fav PASS
MTHR8 4:27 PM 6 Lucky Louisa 5.0/1 18% +5.2% +0.4 Demote #2 BET
WOR7 4:12 PM 10 Voyager 10.0/1 12% +4.4% +2.0 Demote #3 BET
WOR10 5:55 PM 8 See The Fire (GB) 3.5/1 19% 2.1% -3.5 Demote #2 PASS
WOR12 7:00 PM 4 Aryeh 3.0/1 16% 1.6% -3.3 Demote #3 PASS

Today's Results — Sep 12, 2026

Results not yet available — updates after races complete.

Yesterday's Results — Sep 11, 2026

TrackRaceHorseEdgeFavResult
No results for yesterday.

Tomorrow's Picks — Sep 13, 2026

Fav
Tier
Morning lines move — re-check the Raw column after scratches. Scratch-driven ⚑ off-turf flags only firm up on race day.
TrackRacePost#Horse MLWin % Edge Raw Fav Call
No picks yet — pipeline runs after BRIS PDF ingestion.

Live Performance — Since 2026-05-30 all published picks · gate NOT applied

This section is the full published universe: every turf/AW full-coverage pick, before the bet gate. Dirt is watch-only under the surface gate (the model loses on dirt in every backtest fold), so dirt picks are tracked but excluded here.
It is deliberately not the headline number — apply the raw-edge gate and the same window is +2.0% over 109 bets vs -20.6% over 580 here. Use this to watch the model; use the headline to judge the strategy.

Bet Picks
639
Scored / Pending
580 / 20
Win Rate
27.2%
ROI ($2 bet)
-20.6%
Net P&L
$-239

Stored picks from each evening's prediction run, graded against official BQ results. Tracking starts 2026-05-30 — the first day of picks generated by the current feature set. Pending = picks not yet matched to a result (race not in BQ or scoring skipped). V6 publishes a race's top-scored horse when its honest edge is positive — the calibrated, race-normalized model win prob minus the de-vigged (vig-free) market prob — and tiers it high/mid/actionable by edge size. Backtest benchmarks shown for comparison. Pending = picks not yet matched to a result (race not in BQ or scoring skipped).

DatePicksScoredPendingWinsWin%ROI
Backtest Confidence TiersFrozen · pre-2026-05-15confidence tiers were retired from the backtest 2026-05-15 — these figures are a frozen historical reference and do not track the current model; click to expand

⚠️ Overall backtest ROI: -17.2% (32.3% WR over n=6056 races). The tier ROIs below are small, selective subsets (Actionable ≈ 134 of 6056) — high-variance and NOT the live strategy’s expected return. Live betting is full-coverage, turf/AW only (dirt = watch). Treat these as diagnostic, not a forecast.

TierEdge RawRacesWin RateROI
HIGH > 0.10 86 37.2% +34.4%
Mid > 0.08 105 35.2% +24.1%
Actionable > 0.06 134 35.1% +25.5%
No bet ≤ 0.06 1,702 31.8% -14.1%

edge_raw = model softmax - ML implied prob. Pick is always the #1 model-ranked horse; edge gates the bet.
AGREE + edge > 0 = the model picks the favorite but is more confident than the public. See the row below for its current backtest vs live numbers (both bettable-gated).

By Favorite Stance x Edge

StanceFilterRacesWin RateROI
AGREE + HIGH edge > 0.08 — — —
AGREE + any edge > 0 — — —
DISAGREE + HIGH edge > 0.08 — — —
DISAGREE + any edge > 0.06 — — —
All disagree — — — —

AGREE = model's top pick matches the ML favorite. DISAGREE = model picks a different horse. Edge thresholds use raw softmax edge. The model is excellent at confirming undervalued favorites; it struggles to find profitable alternatives when fading the public's choice.

Edge > 0.06 by Track Tier

Track TierRacesWin RateROI
Fair — — —
Mid — — —
Top — — —

At edge > 0.06, fair-track picks dominate both volume and ROI. Top-tier tracks (OP, KEE, CD, SA, AQU, GP) show negative ROI — those markets are too efficient for the edge signal to work. The best betting strategy may be: edge > 0.06 at fair + mid tracks only.

Cross-Cut Pivot

Live (post-2026-05-30) picks pivoted by your chosen Row × Column. Data: Bet = bettable turf/AW (the live strategy — matches the headline ROI); Watch = dirt (gated out, tracked but not bet); All = both combined. Cells show the selected metric over non-scratched picks; n appears underneath. Color follows ROI (green = +EV, red = −EV) regardless of metric chosen, so you can read the visual gradient as profitability while inspecting any aggregate. Cells below the min-n threshold are hidden — sample size kills inference fast in 2D cuts.

Live Diagnostics live vs backtest — spot where the divergence is

Each panel slices live (post-2026-05-30) picks by a dimension. Live column shows non-scratched bets; BT is the matching backtest cut. Red ROI ≥ 20 pts below backtest = candidate to investigate first. Cells with n<5 are too thin to draw conclusions but useful for spotting tracks/buckets we've barely sampled.

Results History — All Graded Picks

Date Track Race Horse Edge ML CLV Fav Fin Result
No graded results yet.