SaratogaBets Weekly Score Card

TrackSmart AI Cumulative Performance Report — NYRA & Regional Tracks

TrackSmart AI | Prepared for Saratogabets | Week Ending September 13, 2026

1. Overview & Methodology

This report covers the reporting window of December 26, 2025 through September 13, 2026, tracking TrackSmart AI's pre-race Top-4 selections across a wide footprint of NYRA and regional circuits: Belmont at the Big A, Finger Lakes, Aqueduct, Laurel Park, Santa Anita Park, Penn National, Saratoga Harness, Woodbine, Saratoga, Del Mar, Hollywood Casino at Charles Town Races, and Kentucky Downs.

The core measure throughout this report is the Top-4 Strike Rate — the share of races in which the model's first, second, third, or fourth-ranked selection crossed the wire first. It is a measure of the model's ability to consistently frame the realistic win contenders in a field, not simply pick the single most likely winner. This week's update adds 89 newly completed races run September 7 through September 13 at Finger Lakes, Saratoga, Kentucky Downs, and Saratoga Harness, growing the cumulative dataset to 1,983 total races.

2. This Week at a Glance

Volume Expansion

The tracking dataset grew to 1,983 total races with the addition of 89 races run across Finger Lakes, Saratoga, Kentucky Downs, and Saratoga Harness from September 7 through September 13, continuing to broaden the sample base underneath the model's cumulative accuracy figures.

Strike Rate Resilience

The cumulative Top-4 Strike Rate bumped slightly higher, remaining exceptionally stable at 77.3% (1,533-for-1,983). The model delivered an absolute masterclass at Saratoga Harness, stringing together back-to-back phenomenal days of 92.3% and 91.7% accuracy, while holding a highly consistent 87.5% mark across the week at Finger Lakes.

Value Normalization

The average payout for AI hits remained heavily sustained at roughly $7.95. Deep value identification and “TS Closer” pace mapping continued to root out lucrative overlays, capturing massive mid-priced winners and longshots such as Sylvester ($22.78), Haatem ($21.84), Ada Clare ($21.44), Mission Improbable ($18.36), and Laigina ($18.54).

System Integrity & Scratch Protocol

The automated scratch-adjustment protocol proved flawless once again this week. Mechanically moving “Next-Man-Up” alternates into active slots seamlessly navigated late race-flow shifts, keeping top-tier hits like Eric From Miami ($8.24) and Palpable ($4.84) in actionable structures despite last-minute changes to the field.

3. Daily Performance Breakdown

Note: Prior totals reflect cumulative data through September 6 and have been condensed into the Prev. Totals row above. Late scratches dynamically shifted runners up in rankings across all cards this week, securing numerous top-tier hits and elevating the overall strike rate.

4. Winner Distribution by Rank (Cumulative)

The rank distribution continues to skew heavily toward the model's top two selections, which together account for 64.0% of all hits (981 of 1,533), while the third and fourth slots remain an essential safety net — together responsible for more than a third of all winning tickets and repeatedly the difference-maker in chaotic, scratch-affected, or longshot-heavy fields.

5. Cumulative Statistics Summary

6. This Week's Structural Findings

Saratoga Harness & “Golden Chair” Geometry

TrackSmart AI mapped the half-mile track perfectly over the weekend, resulting in incredible 92.3% and 91.7% hit rates. The model demonstrated that raw speed figures can be slightly mitigated by premium Post 1 advantages, and recognizing inner draw/rail retention heavily drove its success in isolating key value plays, such as the $14.06 winner One More Tale.

Kentucky Downs & Shipper Superiority

Navigating massive, chaotic fields at a uniquely undulating track requires nuance, and the model held a firm 71.4% accuracy rate on both days it raced there this week. The “International Shipper” protocol flawlessly recognized class drops for European imports such as Take Charge Star and Haatem, while the “TS Closer” metric effectively predicted late-rally sweeps in Stakes and Allowance ranks.

Pace Architecture & Adjusting to Scratches

The model's capability to map pace meltdowns (“cook the leaders”) versus “Alpha Controls” (uncontested speed) was remarkably accurate all week. However, the model showed minor rigidity when dominant front-runners scratched late — when an Alpha Pacesetter scratches, secondary speed horses must be dynamically upgraded to reflect their newly inherited tactical advantages.

2YO FTS Volatility & Pedigree Corrections

The recurring pattern of juvenile maiden volatility continued this week, particularly where the model penalized slow morning workouts. Elite sire pedigree — Bolt d'Oro and Into Mischief among the standouts — combined with strong barn intent heavily outperformed the model's baseline morning-drill metrics, pointing to a clear correction still needed for juvenile sprint chaos.

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

Four structural adjustments stand out from this week's data as the clearest path to tightening the model further:

• Build a dynamic upgrade rule for secondary speed horses that triggers automatically when a confirmed Alpha Pacesetter scratches, rather than relying on the model's default pace-shape read.
• Strengthen baseline ratings for elite sire pedigrees (Bolt d'Oro, Into Mischief-type lines) so a slow morning workout no longer offsets a strong pedigree-and-barn-intent signal in 2YO FTS sprints.
• Continue leaning on the “International Shipper” and “TS Closer” protocols at chaotic, undulating layouts like Kentucky Downs, where they are consistently identifying class-drop and late-rally value that raw figures miss.
• Preserve the current “Next-Man-Up” scratch-adjustment logic as-is at half-mile harness ovals — it is a primary driver behind the standout accuracy at Saratoga Harness and should not be altered without strong cause.

8. Data Notes & Caveats

• Reporting period covers December 26, 2025 through September 13, 2026, cumulative across all tracks listed in Section 1.
• This week's added sample (September 7–13) totals 89 races across Finger Lakes, Saratoga, Kentucky Downs, and Saratoga Harness only; other tracks in the season-long footprint did not card new races in this window.
• “Top 4 Strike Rate” reflects whether the eventual winner appeared anywhere in the model's first four ranked selections, not solely the top pick.
• Average payout figures reflect the win payout on the AI's hit selection within each race, not a blended return across the full Top-4 set.
• Prior-period figures (“Prev. Totals”) are drawn directly from the last cumulative report and are not independently recalculated here.

TrackSmart AI: Cumulative Performance Report

Reporting Period: December 18, 2025 – February 7, 2026

Tracks: Gulfstream Park (GP), Aqueduct (AQU), Fair Grounds (FG), Tampa Bay Downs (TAM)

1.    Executive Summary: February 7 Update

• Volume Expansion: The dataset has expanded to 559 total races with the addition of the Saturday card at Tampa Bay Downs (11 races).
• Performance Resilience: The cumulative Strike Rate rebounded to 4% (up from 76.1%). The model delivered an exceptional daily strike rate of 90.9% (10-for-11) at Tampa Bay, significantly outperforming the historical baseline.
• Metric Adjustment: The Average Payout held steady at ~$8.41 (virtually unchanged). Tampa Bay provided solid value ($8.28 avg on hits), including two double-digit payouts ($11.60 and $18.20), balancing the lower-priced
• Selection Shift: The "Top Pick" remained quiet with only 2 wins. However, the model’s secondary depth was the star of the day; the 2nd Selection captured 4 wins, including key strikes in the Stakes races, effectively driving the day's high accuracy.

2.       Daily Performance Breakdown

The following table reflects the current cycle. All data prior to the Feb 5 update remains consolidated in "Prev. Totals."

3.       Winner Distribution by Rank (Cumulative)

4.       Cumulative Statistics Summary

5.       Daily Observations: February 7 Insights

Tampa Bay Downs (Stakes Dominance & Value)

• Stakes Dominance: The AI performed exceptionally well in the feature races, identifying the winner within the Top 2 selections for all four stakes (Minaret, Suncoast, Pelican, and Sam F. Davis).
• Vulnerable Favorite Survival: In Race 1, the AI flagged the eventual winner (#9) as a "Vulnerable Favorite" due to surface concerns but kept him ranked 4th due to This defensive ranking successfully saved the strike rate despite the negative trip projection.
• Value Captures: The model successfully identified significant value overlays that the public underestimated, including Neshume ($18.20) in Race 8 (ranked 3rd) and Concrete Glory ($14.40) in Race 10 (ranked 2nd).
• Missed Winner Analysis (Race 9):
• The Miss: The AI missed Kokomotion ($23.80), a First Time Starter (FTS).
• The Cause: The model penalized the FTS status heavily because the trainer's stats were weaker compared to elite barns (like Chad Brown) in the same The horse was marked as a "Non-Contender" despite a "sharp workout" note.
• Adjustment: In future maiden turf routes, the "sharp workout" variable will be weighted more heavily against trainer percentages to better identify live longshot debut runners.

TrackSmart AI: Aqueduct 5-Day Performance Report

Reporting Period: February 11 – February 15, 2026

Track: Aqueduct Racetrack (AQU)

Algorithm Parameters: Pure Data-Driven (Blind to Morning Line, Odds, and Public Support)

1.     Executive Summary

Over the five-day racing week at Aqueduct, the AI successfully handicapped 44 races relying exclusively on fundamental speed, pace, and class data. By operating completely blind to the morning lines and public betting pools, the model bypassed crowd bias and successfully isolated value overlays based purely on biological and trip metrics.

The algorithm secured an exceptional 79.5% Top 4 Strike Rate across the week, catching 35 winners. Even without recognizing odds, the model naturally guided selections toward robust payouts, culminating in a strong $10.90 average payout on Sunday and a blended weekly average of ~$7.57 per hit.

2.        Daily Performance Breakdown

3.        Winner Distribution by AI Rank

The distribution of winners reinforces the model's accuracy at the top of the ticket. The primary and secondary selections accounted for over 65% of all successful hits this week.

4.        Track Dynamics & Algorithmic Insights

• Track Bias Navigation: The track surface played fast all week, with a distinct rail/inside speed advantage noted earlier in the week that gradually shifted to a fairer profile by the Because the AI relies solely on data, it successfully adjusted to these micro-changes in the track profile without being swayed by the public over-betting inside horses once the bias dissipated.
• Pace Meltdowns Genuine Speed: The model excelled at differentiating between vulnerable early speed and resilient front-runners. On Sunday, it correctly anticipated a pace collapse, pulling deep closers like Tahila ($19.72) and Bourbon N Lace ($17.78) into the top selections based purely on pace-shape metrics, completely ignoring their high odds.
• Area for Refinement: The primary blind spot this week occurred in routes where the AI misidentified the controlling speed due to inside gate pressure (e.g., assigning a "loose on the lead" projection to the wrong horse). Applying heavier weighting to tactical, off-the-pace ability in these specific route scenarios will tighten up the few missed

TrackSmart AI Weekly Performance Report

Cumulative NYRA & Regional Tracks Analysis — Volume Expansion, Strike Rate Resilience & Scratch Protocol Validation

TrackSmart AI | Prepared for Saratogabets | Week Ending September 20, 2026

1. Overview & Methodology

This report covers the reporting window of December 26, 2025 through September 20, 2026, tracking TrackSmart AI's handicapping performance across a broad footprint of NYRA and regional tracks: Belmont at the Big A, Finger Lakes, Aqueduct, Laurel Park, Santa Anita Park, Penn National, Saratoga Harness, Woodbine, Saratoga, Del Mar, Hollywood Casino at Charles Town Races, and Kentucky Downs.

The core measure throughout is the Top 4 Strike Rate — the percentage of races in which the eventual winner appeared among the AI's top four ranked contenders. This threshold reflects how the model is actually used in practice: as a shortlist for structuring exotic wagers (exactas, trifectas, superfectas) rather than a single win pick, so a hit is counted whenever the winner surfaces anywhere in that shortlist.

This week's update adds 94 races run between September 14 and September 20 across Finger Lakes, Saratoga Harness, and Belmont Park, bringing the cumulative dataset to 2,077 total races.

2. This Week at a Glance

Volume Expansion

The tracking dataset has grown to 2,077 total races with the addition of 94 races run across Finger Lakes, Saratoga Harness, and Belmont Park from September 14 through September 20.

Strike Rate Resilience

The cumulative Top 4 Strike Rate bumped slightly again, remaining exceptionally stable at 77.5% (1,610 for 2,077). The AI model delivered another masterclass week, including a flawless 100% (9-for-9) day at Finger Lakes on September 14 and continuing its dominant run at Saratoga Harness with 92.9% and 84.6% accuracy marks.

Value Normalization

The Average Payout for AI hits remained heavily sustained at roughly $7.92. Deep value identification and scratch protocol mapping continued to root out lucrative overlays, capturing massive mid-priced winners and longshots such as Beau Cheval ($23.68), Raw Courage ($23.58), Mayweather Hanover ($23.10), Amazon ($15.84), and Life's Dynamic ($14.84).

System Integrity & Scratch Protocol

The automated scratch adjustment protocol proved remarkably lucrative this week. Mechanically moving “Next-Man-Up” alternates into active slots flawlessly navigated late scratches and changing race dynamics, keeping top-tier hits like Beau Cheval and Mayweather Hanover in actionable structures.

3. Daily Performance Breakdown

Note: Prior totals reflect cumulative data up to September 13. Daily breakdowns from previous reports have been condensed into the Prev. Totals row. Late scratches dynamically shifted runners up in rankings across all cards this week, securing numerous top-tier hits and elevating the strike rate.

4. Winner Distribution by Rank (Cumulative)

Taken together, the rank distribution shows a model whose edge is not concentrated solely in top-pick accuracy: the second, third, and fourth selections combined still account for nearly two-thirds of all hits, underscoring the value of carrying the full Top 4 shortlist into exotic wager structures rather than leaning on the top pick alone.

5. Cumulative Statistics Summary

6. This Week's Structural Findings

Flawless Sweep & Scratch Protocol Validation

TrackSmart AI achieved a perfect 100% (9-for-9) Top 4 strike rate at Finger Lakes on September 14. Additionally, the manual rule to slide the 5th-ranked contender into the Top 4 following a late scratch proved exceptionally lucrative all week. This simple structural adjustment directly resulted in catching massive payout winners like Mayweather Hanover ($23.10) and Beau Cheval ($23.68).

Surface Shifts & Off-the-Turf Adjustments

The model encountered specific “chaos” misses at Belmont Park (Sept 20) when turf routes were shifted to the All-Weather (Tapeta) track due to weather. The AI's heavy reliance on turf-sire angles and late-pace dynamics was somewhat nullified by the synthetic track, signaling that a dynamic recalculation of Sire/Pedigree anchors must be triggered automatically for off-the-turf events.

Saratoga Harness Half-Mile Geometry

The heavy reliance on half-mile track geometry (rail protection and blast-off protocols) was consistently validated throughout the weekend. Inside draws and early speed dominated the program, perfectly aligning with TrackSmart's “Golden Chair” and “Engine Seal” projections to yield absurd hit rates (92.9% on Sept 14 and 84.6% on Sept 19).

Pace Dynamics on Turf vs. Dirt

The AI demonstrated an elite read on turf route tracking setups at Belmont Park, successfully identifying “Pace Beneficiary” trips, easily mapping the race flow behind hot speeds (e.g., Life's Dynamic, Love Cervere). On standard dirt ovals, front-end speed and pocket-sitting horses thoroughly dominated, with the AI successfully snagging mid-priced winners in “Meltdown/Tough Speed” scenarios where they dictated the terms.

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

• Automate the off-the-turf pivot: build a trigger that automatically recalculates Sire/Pedigree anchors whenever a card is moved off turf onto an All-Weather (Tapeta) or synthetic surface, rather than relying on the standard turf-sire weighting that produced this week's Belmont misses.
• Codify the scratch slide rule permanently: the 5th-ranked-contender promotion on a late scratch has now proven its value across multiple weeks and tracks — lock it in as a standing protocol rather than a manual adjustment.
• Lean further into half-mile harness geometry: continue prioritizing “Golden Chair” and “Engine Seal” inside-draw/early-speed signals at Saratoga Harness, where the model is producing hit rates well above its overall average.
• Preserve the Top 4 structure for exotics: with second-through-fourth selections combining for nearly two-thirds of hits, continue building exacta/trifecta/superfecta structures around the full shortlist rather than narrowing to the top pick.
• Watch dirt “Meltdown/Tough Speed” scenarios for continued overlay value: front-end and pocket-sitting dirt winners continue to deliver mid-priced value that the model is correctly anticipating — keep this angle weighted as a primary value source heading into the fall meets.

8. Data Notes & Caveats

• Reporting period is cumulative from December 26, 2025 through September 20, 2026; the Prev. Totals row condenses all prior weekly reports through September 13.
• Tracks covered this week: Finger Lakes (FL), Saratoga Harness (SARH), and Belmont at the Big A (BAQ); the full cumulative dataset spans twelve tracks.
• Average Payout figures reflect AI hits only (winner within the Top 4) and are not a full wagering ROI calculation.
• Late scratches and the “Next-Man-Up” alternate-promotion protocol were active across all cards this week and are reflected in the results above.