TrackSmart AI Weekly Scorecard: NYRA & Regional Tracks
Cumulative Performance Report — November 15, 2026 Update
TrackSmart AI | Prepared for Saratogabets | Week Ending November 15, 2026
1. Overview & Methodology
This scorecard tracks every race the TrackSmart AI engine has handicapped across the NYRA circuit and the regional tracks we cover, from December 26, 2025 through November 15, 2026. The dataset now spans fourteen venues: Belmont at the Big A (BAQ), Finger Lakes (FL), Aqueduct (AQU), Laurel Park (LRL), Santa Anita Park (SA), Penn National (PEN), Saratoga Harness (SARH), Woodbine (WO), Saratoga (SAR), Del Mar (DMR), Hollywood Casino at Charles Town Races (CT), Kentucky Downs (KD), Keeneland (KEE), and Yonkers Raceway (YON).
The headline metric is the Top 4 Strike Rate: the share of races in which the eventual winner was one of the AI's four ranked selections. Alongside it we track which rank produced the winner (1st through 4th pick) and the average win payout on those hits, which tells us whether the engine is finding winners at a price or just collecting favorites.
This week added 124 races run from November 9 through November 15 across Finger Lakes, Belmont at the Big A, Del Mar, Yonkers Raceway, and Saratoga Harness, bringing the cumulative total to 2,904 races.
2. This Week at a Glance
Volume Expansion
The tracking dataset grew to 2,904 total races with the addition of 124 races across five circuits — Finger Lakes, Belmont at the Big A, Del Mar, Yonkers Raceway, and Saratoga Harness — run between November 9 and November 15.
Strike Rate Resilience
Even with a large block of new races, the cumulative Top 4 Strike Rate held exactly where it was: 77.2% (2,243 for 2,904). The week was headlined by Yonkers Raceway on November 13, where the AI hit 90.0% (9-for-10), and Saratoga Harness on November 14 at 85.7% (12-for-14). The harness cards in particular were a strong week for the model.
Value Normalization
The average payout on AI hits eased slightly from ~$8.22 to ~$8.17 — a normal settling as the sample grows, not a sign the engine is drifting toward chalk. The week still produced several big overlays: Grizz At the Gate ($24.28), El Parrandero Q ($24.40), Admiral Sol ($23.66), Papa Vegas ($19.64), and Ms Grainger Danger ($18.40).
System Integrity & Scratch Protocol
The automated scratch adjustment protocol continued to earn its keep, moving lower-ranked alternates up the board when chalk scratched and turning unheralded runners such as Big Juanito and Inexorable into Top 4 winners. The model also held up through major Off-Turf and Tapeta surface shifts on both coasts, continuing to capture winners despite volatile conditions.
3. Daily Performance Breakdown
|
Date |
Track |
Top 4 Strike Rate |
Total Races |
1st Pick |
2nd Pick |
3rd Pick |
4th Pick |
Avg Payout (AI Hits) |
|
Prev. Totals |
MULTI |
77.2% |
2780 |
801 |
566 |
463 |
317 |
~$8.22 |
|
Nov 9 |
FL |
88.9% |
9 |
2 |
5 |
2 |
0 |
$4.94 |
|
Nov 10 |
FL |
77.8% |
9 |
2 |
1 |
1 |
3 |
$11.03 |
|
Nov 11 |
FL |
62.5% |
8 |
2 |
2 |
1 |
0 |
$5.02 |
|
Nov 12 |
BAQ |
75.0% |
8 |
1 |
1 |
1 |
3 |
$8.09 |
|
Nov 13 |
BAQ |
75.0% |
8 |
2 |
2 |
2 |
0 |
$6.36 |
|
Nov 13 |
DMR |
87.5% |
8 |
3 |
2 |
2 |
0 |
$5.51 |
|
Nov 13 |
YON |
90.0% |
10 |
2 |
3 |
3 |
1 |
$5.54 |
|
Nov 14 |
BAQ |
80.0% |
10 |
1 |
3 |
3 |
1 |
$9.73 |
|
Nov 14 |
DMR |
66.7% |
9 |
3 |
1 |
0 |
2 |
$7.63 |
|
Nov 14 |
SARH |
85.7% |
14 |
6 |
2 |
2 |
2 |
$6.92 |
|
Nov 15 |
BAQ |
75.0% |
8 |
2 |
2 |
2 |
0 |
$8.45 |
|
Nov 15 |
DMR |
66.7% |
9 |
1 |
1 |
1 |
3 |
$8.80 |
|
Nov 15 |
SARH |
78.6% |
14 |
4 |
5 |
1 |
1 |
$6.20 |
|
NEW TOTALS |
MULTI |
77.2% |
2904 |
832 |
596 |
484 |
333 |
~$8.17 |
Prior totals reflect cumulative data through November 8; daily lines from earlier reports are condensed into the Prev. Totals row. Late scratches shifted runners up in the rankings across all cards this week. The Nov 9 Finger Lakes rank totals include a dead heat between the 1st and 2nd AI choices, as do the Oct 24 Keeneland totals — which is why the rank columns sum to 2,245 hits for 2,243 successful races.
4. Winner Distribution by Rank (Cumulative)
|
AI Rank |
Wins |
% of Hits |
Trend Note |
|
Top Pick |
832 |
37.1% |
Pace Flow Precision: Consistent mapping of race shapes continued to yield dominant top-pick success. Identifying extreme speed advantages and executing off-turf adaptations allowed the AI to sweep logical cards at Saratoga Harness and Del Mar. |
|
2nd Selection |
596 |
26.6% |
Class Elevation & Adaptability: A highly reliable safety net, capitalizing on tactical setups and capturing deep value finds this week like Uncle Jim ($16.44) and Papa Vegas ($19.64). |
|
3rd Selection |
484 |
21.6% |
Alternate Protection: Benefited heavily from "Next-Man-Up" logic overrides and scratch adjustments. Thrived during chaotic conditions, flagging deep-priced winners like Ms Grainger Danger ($18.40). |
|
4th Selection |
333 |
14.8% |
Rebounding Safety Net: Elevated longshots in chaotic and depleted fields and benefited from scratch-bump rules, yielding hits like Grizz At the Gate ($24.28) and El Parrandero Q ($24.40). |
The distribution keeps its healthy shape: the top pick carries the biggest share of hits, and each rank below it contributes less, in order. What matters for bettors is where the money comes from. The top two picks deliver the bulk of the hit count, but the 3rd and 4th selections are where this week's biggest prices came in — Papa Vegas ($19.64), Ms Grainger Danger ($18.40), Grizz At the Gate ($24.28) and El Parrandero Q ($24.40) all came from the 2nd through 4th selections. That is the case for playing the full Top 4 rather than keying only the top choice.
5. Cumulative Statistics Summary
|
Metric |
Stat |
Description |
|
Overall Accuracy |
77.2% |
(2243/2904 Races) The AI has placed the winner in its Top 4 selections in 2,243 of the 2,904 recorded races, holding steady through a large field expansion and volatile weather conditions. |
|
Top 2 Accuracy |
49.2% |
(1428/2904 Races) The primary or secondary choice remains highly reliable, winning in nearly half of all modeled races. |
|
Avg Payout |
~$8.17 |
The average payout remained resilient, supported by the automated scratch logic pulling longshots into the winning frame despite heavy chalk results at smaller venues. |
6. This Week's Structural Findings
The Scratch Cascade Keeps Paying
Sliding alternates up the board after late scratches was highly effective again. The automated cascade caught backup horses such as Big Juanito ($5.20), Inexorable ($6.50), and Bourbon Story ($4.20) as higher-confidence selections once the chalk came out. Updating the board for every scratch remains mandatory — this is now one of the most reliable parts of the system.
Turf Handicaps Surviving the Surface Switch
Weather forced major surface changes at Belmont at the Big A (Muddy/Tapeta) and Del Mar (Good Dirt). The AI's turf-based handicaps still found winners inside the Top 4 through these transitions, including a 3-for-3 run on Tapeta transfers on November 13. The gap is that the model has no Tapeta-specific form adjustment, which left blind spots such as Grace and Grit on November 15.
Reading Pace Meltdowns Well, Dominant Speed Less So
The engine showed good foresight on chaotic, contested paces, correctly identifying the horses that would benefit from destructive early speed duels (e.g., Battle School, Stagger Lee). The weakness was on the other side: in projected "Dominant Speed" races where the frontrunner faced hidden pressure from the inside, the model sometimes misread the shape and left mid-pack closers undervalued.
Too Hard on Bottom Claimers and Layoffs
Form-cycle and layoff rules were a little too rigid this week. Horses dropping to bottom-level claiming company (e.g., Tekila, Nov 10) and horses returning from freshenings under elite trainers (e.g., Burhan for Chad Brown; Let Seamus Go for Mark Glatt) won despite heavy AI penalties. Softening those penalties for sharp local barns and large class drops should close this gap.
Post Geometry Handled — But One Data Error Slipped Through
At the half-mile harness ovals, Saratoga Harness and Yonkers, the AI handled outside-post penalties well by correctly judging gate speed and the ability to tuck into a pocket trip. Separately, a data-ingestion error wrongly flagged a live favorite as scratched (Mele Kalikimaka, Nov 15). Checking scratch status against official feeds before finalizing the pre-race model is critical.
7. AI's Final Analysis — Building the Path Forward
Five things worth changing or protecting, based on this week's findings:
- Keep the scratch cascade locked in. It produced Top 4 winners again this week and is a big reason the payout average stays above $8. No changes — just strict, every-race enforcement.
- Add a Tapeta form modifier. The turf-to-Tapeta transfers mostly worked, but without any adjustment built for the synthetic surface. A dedicated adjustment for synthetic switches would turn misses like Grace and Grit into contenders.
- Pressure-test "Dominant Speed" projections. Before labeling a race as a lone-speed setup, check for inside horses with enough early foot to press. When that pressure exists, give mid-pack closers more weight.
- Ease the claiming-drop and layoff penalties. Scale the penalty down for large drops into bottom-level claimers and for freshened horses from elite or sharp local barns — the Chad Brown and Mark Glatt winners are the template.
- Validate scratches against the official feed. One false scratch (Mele Kalikimaka) removed a live favorite from the board. A final cross-check against official scratch data before publishing should prevent a repeat.
8. Data Notes & Caveats
- Reporting period runs December 26, 2025 through November 15, 2026; this update adds races from November 9–15, 2026.
- Daily lines prior to November 9 are condensed into the "Prev. Totals" row and are not broken out here.
- Two dead heats between the 1st and 2nd AI choices (Nov 9 Finger Lakes; Oct 24 Keeneland) mean rank-level win counts total 2,245 against 2,243 successful races.
- Average payout figures are approximate cumulative values (~$8.22 prior, ~$8.17 current) as reported in the source data.
- Rankings reflect the board after late-scratch adjustments; scratch-bumped alternates are credited at their adjusted rank.
- The dataset mixes thoroughbred and harness circuits (Saratoga Harness, Yonkers), so cross-track comparisons should keep the race type in mind.

