Scratch Watch

Methodology — the un-opaque part

Québec edition: same method, different sources. Loto-Québec publishes a prize-claims report (about weekly) listing every prize tier's total and claimed counts — better coverage than OLG's ~$100+ feed, so the pool estimate is pooled over all tiers — and each game's rules PDF carries the full prize structure plus the ticket count it's based on. Self-check: each parsed prize table is summed against Loto-Québec's stated "theoretical payoff rate" and must match within 1 pt. Free-ticket prizes are valued the way Loto-Québec values them — face value × the game's own payoff rate, i.e. the expected cash return of re-rolling (a free $3 ticket in a 59.5% game is worth $1.78) — which reproduces their stated rates exactly. Free tickets don't appear in the claims report, so they're assumed to deplete with the pool. National (ILC) games show only $1,000+ tiers in Québec's report. The Ontario notes below describe the shared method.

Everything comes from two public OLG sources, cross-checked against each other:

• Player's Guide Fact Sheets (one PDF per game, linked from each game's page on olg.ca): the full printed prize structure — every tier's prize, winner count and odds, total tickets printed, payout %, claim deadline.
• The INSTANT Unclaimed Prizes feed (gateway.www.olg.ca/feeds/instant-unclaimed, updated daily): for each game's larger prizes (~$200+), how many are still unclaimed right now.

From those two, per game:

• Estimated tickets remaining: for the tracked prize tier with the most printed winners (the most statistical mass), unclaimed ÷ total. Mid-tier prizes are claimed quickly, so this ratio tracks how much of the print run is still on shelves.
• Adjusted payout: value still in the pool ÷ cost of the remaining tickets. Tracked tiers use their live unclaimed counts; small untracked tiers are assumed to deplete in proportion to tickets sold.
• Top prize odds now: estimated remaining tickets ÷ unclaimed top prizes.
• +EV chance: if tickets sell in random order, every prize's expected remaining count shrinks in step with the pool — so on average a game's adjusted payout never changes as it sells down. It only crosses 100% through luck: big prizes happening not to get hit. We simulate that 20,000 times per game. Tiers with more than 200 unclaimed prizes deplete smoothly (law of large numbers) and scale with the pool; every prize in smaller tiers gets a random position in the remaining tickets. The column shows the share of runs in which the leftover pool is worth more than its price at some point before shrinking to 12% of the print run, and the pool size where that typically happens.
• Freshest-of merge for national games: for ELITE, EXTREME, VAULT, ULTRA and 30X the unclaimed counts are Canada-wide and published separately by OLG (daily), BCLC (daily JSON) and Loto-Québec (weekly). They lag each other by a day or two in either direction, so per tier this site takes the lowest count among sources whose totals match, and shows all three.
• Older print runs and national games: olg.ca's listing only links the newest edition of each game (and, oddly, usually that edition's page links the previous run's fact sheet). The unclaimed feed, however, still reports the older run — which is typically still on some shelves — and national games like 30X and $1,000,000 ULTRA that the listing never shows. This site keeps every fact sheet it has ever seen, indexed by the sheet's own game number, so those games are analyzed too; they're tagged older run / national in the table.
• How OLG retires games (and why 12%): per Ontario's Auditor General (2022 report on OLG), about 88% of all printed INSTANT tickets are sold, so the average game leaves shelves with ~12% of its print run unsold — that's the simulation's stopping point. OLG's stated rule is that games with top prizes of $100,000 or more are removed once all top prizes are claimed; games with smaller top prizes may keep selling for a while with the unclaimed-prize report as disclosure. Mechanically, OLG assigns a game a "pack activation end date", after which retailers can deactivate and return unsold tickets for credit (retailer terminal training guide) — but nothing stops a retailer selling an already-open pack, which is what the AG flagged. OLG's fix — single-ticket barcodes that stop validating/selling once all top prizes are won — was funded in May 2024 and is due by June 2027 (AG 2024 follow-up); until then retirement is a manual recall via retailer notices. That end date isn't public; this site detects retirements when a game drops off olg.ca's listing and logs its pool % and top prizes left at that moment (data/catalog-log.csv) to learn the real pull levels over time.

Self-checks: every parsed prize table is reconciled against OLG's own published payout % and prize counts (all games currently reconcile to within rounding). Prizes with no fixed cash value (wheel spins, live-event wins, "for life" annuities) are valued from OLG's published wheel odds tables where available, otherwise implied from OLG's own payout % — and flagged as estimates in the game detail.

Honest caveats: claimed ≠ sold (winners sit on tickets, sometimes for months); the remaining-fraction proxy is noisy for games with few tracked prizes; distribution isn't uniform across stores; OLG can reorder or pull tickets at any time ("quantity of tickets ordered and/or distributed may vary").

Reality check: most games keep 18–38¢ of each dollar wagered, and even a "+EV" flag is paper math, not a strategy: near a game's end of life, "unclaimed" increasingly means sold-but-unscratched or lost tickets (inflating apparent value exactly when it looks best); OLG recalls unsold stock and retires games before the pool can be bought down; and the excess value sits in a handful of tickets scattered across ~10,000 retailers, so the variance is unbankable. This is a curiosity dashboard.