TrackSmart AI: Cumulative Performance Report - NYRA

Reporting Period: December 26, 2025 – February 21, 2026

Track: Aqueduct (AQU)

1. Executive Summary: February 21 Update

Volume Expansion: The tracking dataset has grown to 238 total races with the addition of 25 races run from February 19 to February 21.
Strike Rate Growth: The cumulative Top 4 Strike Rate increased slightly to 79%. This was heavily supported by back-to-back strong days on February 19 and 20, where the model successfully placed the winner in its Top 4 selections in 7 of 8 races (87.5%) on both cards.
Value Explosion: The Average Payout for AI hits saw a massive jump, rising from ~$7.61 to ~$8.16. This surge was driven by the engine catching chaotic longshots, specifically a $77.46 winner on February 19 and a $42.38 winner on February 21.
Underneath Strength: While the Top 2 choices still dominate, the 3rd and 4th selections proved highly lucrative this period, bringing in 11 combined wins over the three days and catching the biggest payouts of the period

(Note: The 77.8% for Feb 21 is rounded to 78% in daily summaries

AI Rank Wins % of Trend Note

Hits

Top Pick (#1)

38.3%

Steady Contender: Added 6 wins over the recent block. Picked up momentum on Feb 21 with 3 top-pick winners.

2nd Selection

23.4%

Slight Cooling: Added 4 wins on Feb 19 and 20, but was shut out entirely on the muddy Feb 21 card.

3rd Selection

25.0%

Overtaking 2nd: Passed the 2nd pick in overall volume by securing 7 wins this period. Caught the $42.38 winner, Dr. Merciless, on Feb 21.

4th Selection

13.3%

The Value Safety Net: Delivered 4 Alternate winners over the 3 days. Most notably, caught the $77.46 mudlark, Tiote, on Feb 19.

Metric Stat Description
Overall Accuracy	79.0%	(188/238 Races) The AI has successfully placed the winner in its Top 4 selections in 188 of the 238 recorded races.
Top 2 Accuracy	48.7%	(116/238 Races) The primary or secondary choice wins just under half of all races, maintaining long-term calibration.
Avg Payout	~$8.16	The average payout leaped upward this period, heavily supported by the 3rd and 4th selections consistently diagnosing chaos results and catching double-digit payouts.

Elite Pace Mapping & Mudlark Identification (Feb 19): The model showed extreme proficiency in pace mapping, correctly anticipating race flows and contested It successfully flagged Tiote as a dangerous underneath threat in Race 3 due to his proven status as a mudlark, triggering a massive $77.46 payout.
Scratch Protocol Vulnerabilities (Feb 20): In Race 3, the AI successfully predicted high early pace pressure, but its primary choice to capitalize off the pace (#2 Apalta) was The AI failed to elevate a secondary stalker/closer like the eventual winner, Come to Papa ($9.58). Implementing a secondary pace-meltdown beneficiary protocol when a primary off-the-pace selection scratches could resolve this.
Muddy Track Speed Bias Pace Meltdowns (Feb 21): The AI performed well on a muddy, sealed track, catching a $42.38 longshot. However, it incorrectly predicted pace meltdowns that would favor stalkers in Races 3 and 7. The muddy track ultimately favored early speed, allowing front-runners Commuted and Twirling Beauty to control the pace without succumbing to expected pressure. Reviewing early energy and meltdown parameters to ensure front-runners aren't penalized too harshly when multiple speed horses are entered on off-tracks is recommended.