I've been collecting scratch-off prize data daily from official state lottery sites since 2011 (40+ states). For this analysis I took the 3,871 games where I could observe the complete lifecycle — launch to final day, weekly resolution — and traced every top-prize claim: 18,086 of them.

The question: when, over a game's life, do the top prizes actually get won?

Results:

• Median top prize is claimed just 30% of the way into a game's life
• 42.5% of top prizes are gone in the first quarter of a game's life
• 72.5% are gone by the halfway point
• Only 9.3% are won in the final quarter
• 23.7% are never won at all — the game ends with the jackpot still unclaimed

A few takeaways that surprised me:

The brand-new game everyone grabs has no measurable edge. I scored every game weekly on a value metric (remaining prizes vs. estimated remaining tickets). Essentially zero brand-new games ever rate strongly. The median game is mediocre at every age — but the small set of genuinely good bets is concentrated almost entirely in older games that still have their top prize left. Same jackpot, fewer tickets left to buy it with.

Most old games are traps though. Since top prizes go early, the majority of aged games have nothing left at the top. In an earlier analysis I found ~17% of games on sale have zero top prizes remaining. So "buy old" only works if you check remaining prizes first — every state lottery publishes this, most players never look.

The unclaimed-jackpot number is the one that bugs me. Nearly 1 in 4 advertised top prizes is never paid out, mostly because states end games early (a separate analysis found 60% of games end with a top prize still on the books). The jackpot on the front of the ticket is real, but there's a decent chance nobody will ever win it.

Methodology notes: games analyzed launched and ended between April 2022 and May 2026; restricted to games with ≤50 top prizes (real jackpots, not aggregate tiers); results stable across cutoffs of 10–100 and robust to weekly-vs-daily granularity. Happy to go deeper on any of this in comments.

Full writeup with tables and methodology: scratchsmarter.com/buy-old-not-new-scratch-off-study/