HOW 10K SIMS WORKS
10K Sims plays every NFL, college football, NBA and college basketball game 10,000 times and reports what happened: a projected final score, fair lines on the hook, and a graded pick on the total and the side. Here is exactly how a projection gets built, step by step. One rule runs through all of it: the betting line never goes into the model. It's only used at the end, to grade the edge.
1. THE DATA: REAL SCORES, NOTHING IMPLIED
Every input is a final score from a completed regular-season game, pulled from ESPN: how many points each team scored and how many it allowed, this season and last season. No point spreads, no market totals, no "implied" team totals. A model built from the market can only agree with the market, so this one starts from the games themselves.
2. THE HYBRID: THIS SEASON + LAST SEASON
Each team gets four numbers, scoring and points allowed for this season (PRO) and last season (REG). The Hybrid blends them, and this season takes over as games are played:
k = 6 for the NFL and college football, 20 for the NBA, 12 for college basketball. An NFL team six games in is weighted 50/50; an NBA team hits 50/50 at 20 games.
3. RECENCY, PRO/REG AND "PUTTING A LINE THROUGH IT"
Form matters, so the last 5 games of this season are weighted 5-4-3-2-1, the most recent heaviest. Before weighting, any anomalous result in the last 3 (a blowout far outside the team's normal range) gets a line put through it and is replaced by the better of the 4th or 5th game back. Recency minus the Hybrid tells you whether a team is progressing or regressing.
4. REVERSE LOGIC: MEAN REVERSION
Hot streaks cool off and cold streaks warm up. Reverse Logic takes the four game-to-game changes across the last 5, adds them up, halves the sum, rounds up, and applies it to the most recent game in the opposite direction. A team that has climbed 12 points over its last 5 gets 6 taken back.
5. THE MATCHUP
Home field is withheld at neutral sites, and also when the home team is a big favorite (7+ in football, 8+ in basketball).
6. 10,000 GAMES
Then the game is actually played, 10,000 times. Football plays roughly 11 drives per team, each ending in a touchdown, a field goal or nothing. Basketball plays every possession as 0, 1, 2 or 3 points at the pace the projection implies. Forecastable Dispersion (FD) sets how wide each game can swing, based on how much these two teams' game totals have bounced around compared with a typical game. Margins are held to each sport's real-world range, and every sim also draws how far each team's true level might sit from its projection, wide early in the season and tighter as games come in.
7. FAIR LINES AND GRADES
The fair total and fair spread are the numbers on the hook (the .5) that split the 10,000 games closest to 50/50, the same way a book sets a line. Each pick is then graded against standard -110 juice, where you need 52.4% to break even:
Until both teams have played 3 games this season, nothing grades higher than LEAN.
8. THE SIGNALS
Alongside the numbers, each game carries TuttleMath signals: No-Show Quarters (football quarters where a team scores 0 or 3, compared with the NFL norm of about 44%), DFR Watch for Drastic Form Reversals off the last meeting, FD, and each team's progression/regression on offense and defense.
WHAT IT CAN'T SEE
The model knows scores, not news. Injuries, suspensions, trades and weather aren't in it, so check the injury report before you bet. Projections are information, not guarantees.
Simulators by sport: NFL · college football · NBA · college basketball