How We Rank NY Scratch-Off Tickets

The Seven Core Factors

• Value — Expected dollar return per ticket relative to price.
• Edge — Payout rate differential vs. game-price category baseline.
• Density — Remaining prizes per unsold ticket (top + mid tier).
• Playability — Mid-tier prize frequency ($100–$10,000 band).
• Momentum — Rate of prize depletion in the last 14 days.
• Confidence — Sample size & data-freshness penalty.
• Upside — Top-prize probability weighted by jackpot size.

Data Pipeline

1. Check the official NY Lottery game list during scheduled refresh cycles.
2. Pull retailer & sales data from NY Open Data.
3. Apply Bayesian smoothing on remaining-prize counts (Beta prior, empirical posterior).
4. Recompute all 33 Smart Score factors per game; blend via walk-forward-validated weights.
5. Publish ranked JSON to S3 + rebuild this site in under 90 seconds.

Validation & Backtesting

Factor weights are tuned through 6-month walk-forward cross-validation against historical NY Lottery outcomes. An online-learning layer fine-tunes weights within a 7-day rolling window with a 2σ safety gate that blocks unstable updates. The full pipeline (source code) is open for inspection.

See live prediction accuracy → Daily pairwise accuracy, Kendall τ, and rolling 30-day averages, refreshed every cycle.

Realized Outcome Validation

We grade our own math against reality, not against our own future scores. Using 88 consecutive daily snapshots, we measured each game's realized window payout — the actual prize value claimed divided by the actual dollars spent on tickets over each 14–28 day window. Games our model ranked in the top decile by payout rate went on to return roughly 70 cents per dollar over the following 28 days versus roughly 57 cents for bottom-decile games (Spearman rank correlation +0.57 across ~3,500 game-windows).

The same study measured price-tier lifecycle decay now built into the Smart Score: $1–$3 games lose roughly 0.3–0.75 percentage points of payout per 1% of additional sell-through once past 60% sold, while $10+ games historically hold or slightly improve late in their run. Top prizes are claimed most often in the 60–85% sold window (~14% of game-days vs ~3% early-life).

What the Score Does Not Predict

Smart Score measures current ticket value — not future luck. Scratch-off outcomes are independent random events. A Smart Score of 85 does not guarantee a win; it signals that the ticket's remaining prize pool offers mathematically favorable value versus alternatives.