TrackSmart AI Weekly Scorecard

NYRA & Regional Tracks — Cumulative Performance Report

TrackSmart AI  |  Prepared for Saratogabets  |  Week Ending August 30, 2026

1. Overview & Methodology

This scorecard tracks the TrackSmart AI Top-4 selection model across NYRA and regional partner tracks on a rolling, cumulative basis. The season-to-date window now runs December 26, 2025 through August 30, 2026, and covers Belmont at the Big A (BAQ), Aqueduct (AQU), Saratoga (SAR), Finger Lakes (FL), Saratoga Harness (SARH), Laurel Park (LRL), Santa Anita Park (SA), Penn National (PEN), Woodbine (WO), Del Mar (DMR), Charles Town (CT), and Kentucky Downs (KD).

The core metric is the Top 4 Strike Rate — the share of races in which the winner appeared among the AI's top four selections — tracked cumulatively rather than card-by-card, so a volatile day at one track doesn't distort the read on the model as a whole. This week's update folds in 133 races run August 24–30 across six of those tracks, bringing the all-time dataset to 1,785 races.

The "Prev. Totals" row in the table below condenses everything reported through August 22 into a single line, so the week's 13 new cards can be read against the full-season baseline in one view.

2. This Week at a Glance

Volume Expansion

The tracking dataset grew to 1,785 total races with the addition of 133 races run across Finger Lakes, Saratoga, Charles Town, Del Mar, Kentucky Downs, and Saratoga Harness from August 24 through August 30.

Strike Rate Resilience

The cumulative Top 4 Strike Rate remained highly stable, settling at 77.2% (1,378 for 1,785) — essentially flat off last week's 77.5%. The model maintained exceptional hit rates on established dirt and harness tracks, posting perfect 100% cards on August 24 and August 25 (Finger Lakes) and a 92.9% strike rate (13 for 14) on August 29 (Saratoga Harness), while weathering real volatility at the chaotic Kentucky Downs turf meet (45.5% on August 29, 50.0% on August 30).

Value Normalization

The average payout for AI hits held strong at ~$7.95, up slightly from ~$7.87. Deep-value identification continued to drive large exotic payouts, capturing overlays such as Quibbler ($47.70), People Watching ($37.48), Mr. Einstein ($32.22), Niche ($30.00), Dotsy ($29.48), Dr. Kapur ($23.90), and mild harness upsets like Gideon ($10.32) and Blazing Home N ($9.32).

System Integrity & Scratch Protocol

The automated scratch-adjustment protocol was critical to preserving ROI this week. Post-scratch step-ups consistently moved winners into the Top 4 actionable slots, securing hits like Mr. Einstein ($32.22), Brickyard Belle ($11.86), Hymn ($11.40), and Englishman ($2.82) that would otherwise have fallen into the "Fringe" tier.

3. Daily Performance Breakdown

Every card run since the last update, with the pre-existing season total condensed into the "Prev. Totals" row for comparison.

Late scratches and altered race flows dynamically shifted runners up in the rankings across all cards this week, securing numerous top-tier hits and keeping the cumulative strike rate exceptionally stable.

4. Winner Distribution by Rank (Cumulative)

Reading the 1,378 season hits by which AI rank actually won shows where the model's edge is coming from — and it isn't only the top pick.

Read together: the top two picks alone cover almost two-thirds of all hits, but nearly a quarter of the model's wins come from the 3rd and 4th slots — which is exactly the tier the scratch-adjustment protocol is built to protect.

5. Cumulative Statistics Summary

6. This Week's Structural Findings

6.1 First-Time Starter (FTS) Blind Spot

A recurring pattern emerged where the model heavily penalized first-time starters for unproven or flat debut lines and inside draws. This reliance on historical velocity data caused the model to miss several double-digit winners — Sun Kissed Bliss ($21.98), Peppermint Twist ($19.06), and Club Reign ($18.72) among them. Implementing a tactical/intent bonus for FTS runners is advised.

6.2 Kentucky Downs / Turf Meet Volatility

The model struggled with Kentucky Downs' unique layout, leaning too hard on "Turf/Late Pace Priority" logic and aggressively fading tactical inside speed. Horses with prior course experience frequently outperformed their baseline speed figures, producing chaotic misses where the front-runners held on.

6.3 Pace Architecture & Bias Mastery

When the race-flow maps held true, the model thrived. It perfectly diagnosed front-running bias on the dirt — identifying "Alpha Controls" — which drove elite accuracy on chalk and logical top picks, evidenced by the dominant 84.6% hit rate at Charles Town.

6.4 Harness Track Bias Awareness

The model showed an elite read on harness dynamics at Saratoga Harness on August 29. By appropriately weighting "Engine" numbers and early speed, it continually favored horses making their own luck on the lead or taking a pocket trip, yielding a 92.9% hit rate on the card.

6.5 Scratch Protocol Efficacy

Adhering to the rule of dynamically moving horses up into the Top 4 upon scratches proved to be the bedrock of this week's stability. Mechanically sliding alternates into vacated spots salvaged highly-priced tickets on almost every card, preventing major value from being lost to the Fringe tier.

7. AI's Final Analysis — Building the Path Forward

Pulling all of the above together, five things are worth reinforcing or adjusting in the approach going forward:

• Add a tactical/intent bonus for first-time starters. The FTS blind spot cost real double-digit prices this week (Sun Kissed Bliss, Peppermint Twist, Club Reign); pure historical-velocity screening is under-crediting live first-out runners.
• Build a Kentucky Downs course-experience flag. Prior-course starters are outperforming their baseline figures there and the model isn't crediting it — this is the same shipper-vs-course-form gap the Kentucky Downs meet-prep study flags independently.
• Keep leaning into the pace-architecture model on conventional dirt configurations. Charles Town-style "Alpha Controls" identification is the model's clearest strength and the source of its most dominant hit rates.
• Preserve and reinforce the harness "Engine"/early-speed weighting. Saratoga Harness's 92.9% card shows this weighting is working as intended — no change needed, just continued monitoring as new harness data comes in.
• Hold the automated scratch-adjustment protocol as-is. It was the single biggest stabilizer of the cumulative strike rate this week and should stay a standing rule, not a manual override.

8. Data Notes & Caveats

• Reporting period: December 26, 2025 – August 30, 2026 (season to date); this week's window is August 24–30, 2026.
• Prior totals reflect cumulative data through August 22; the daily breakdown above covers only the 13 new cards run August 24–30 across six tracks (133 races).
• Track codes follow standard Equibase/NYRA convention: BAQ, AQU, SAR, FL, SARH, LRL, SA, PEN, WO, DMR, CT, KD.
• Top 4 Strike Rate, rank distribution, and payout figures are as computed by the TrackSmart AI Top-4 selection engine, inclusive of the automated scratch-adjustment protocol described in Section 2.