no archive data yet
Profit per $1 = for every dollar we bet on a kind of bet, what it gives back on average. It’s the fair way to compare bet types — a 60¢ bet and a 95¢ bet win at different rates, so win-percentage alone misleads.
Beating the closing price = did we buy cheaper than the last price before kickoff (in cents). Short stretches can be luck; this measures accuracy against the market.
Odds accuracy check = when we say 80%, does it happen 8 times out of 10? Closer to 0 is better; 0.25 means we’re no better than a coin flip.
| fit | value |
|---|---|
| fitted | Nov 3 |
| odds trim | subtract 0.04 above 0.90 (the earlier model overestimated) |
| games measured | 52,841 |
| fitted from file | dfe1bf9f4ee7fb3c… |
Practice money only. Bets only before kickoff. Once bought, we hold to the final score.
CERTAIN = the line sits under the safety number (90+ out of 100 hit it) → $30. STRONG = comfortably inside our projection, hits about 75 to 88 out of 100 → $20. We never bet anything we rate under 70%.
We keep lowering the bet line until only an injury could ruin it. Max $75 on one player, $150 on one game, never the same market twice in 5 days. Thin lines (under 20 traded) are off limits. Every number is rebuilt from actual box scores, never from summary pages.
| part | newest file | how fresh |
|---|---|---|
| the numbers behind the model | playerweeks_2026.csv | 0h ago |
| the injury feed | live_status-latest.json | 2h ago |
| saved prices for the price charts | prices-Nov 9.jsonl | 1h ago |
| the newest betting board | certainty-2026 wk9-LVSF.json | 2h ago |
| the newest grading sheet | settle-Nov 1.json | 1d ago |
| the newest projections | projections-2026 wk9-TBCHI.json | 3h ago |
| when | what happens |
|---|---|
| 4:30 every morning | Download the latest box scores, rebuild every player’s numbers, run the model, check the data for holes, rebuild this website and publish it |
| 6:00 every evening | Save a snapshot of every Kalshi price — that’s how the price charts and “beating the closing price” numbers get built |
| 9:30 Thu + Sun | Score every game starting in the next 3 days and place the bets that pass every test |
| 10:30 am & 2:30 pm Sun · 5:30 & 7:30 pm Thu+Mon | Re-check right before kickoff: did injuries move anything? did the price move? place what newly qualifies, refresh this site |
| Hourly during games | Every held ticket shows live progress on its game page; an alert fires only when a bet player is ruled out |
| 1:00 Tuesday | Grade every ticket against the final scores and post the week’s report |
No buying after kickoff — stale prices are traps. No betting on players who aren’t active. The CERTAIN badge only exists under the safety number. Thin lines are off limits. $75 a player, $150 a game, no repeat buys of the same market within 5 days. If the weekly data fails its quality check, the system stops and shouts instead of betting on bad numbers.