Backtest Explanation
The NBA model results are built to answer a customer-facing question: if the Three Check Bets NBA process had been operating in order, using only information available before each game, and a customer risked $100 on every qualified play, what would the record, return, ROI, and closing-line value have been?
Walk-Forward Method
A walk-forward backtest moves through the schedule in order. The model is not allowed to use final-season information that would not have been known before the game. Each matchup is treated like a live betting decision: build the pregame feature set, compare the model number to the betting market, apply the Three Check rules, record the qualified plays, and grade those plays after the final score.
That structure is important because it is stricter than a normal end-of-season review. A normal review can accidentally know which teams were better than expected, which injuries mattered, or which market numbers moved the most. A walk-forward test asks whether the process could have identified value before the game was played.
The results page uses $100 per qualified play so the return is easy to understand. The same record can be scaled to smaller or larger bet sizes by treating one unit as one standard bet.
Three Check Process
The NBA model is not designed to force a play on every game. It first creates a fair number, then the governor asks whether the edge is strong enough to become a public card play. That is the same idea behind the Three Check Bets name: a play should survive more than one type of test before it reaches the customer.
The model must disagree with the sportsbook by enough to create a real edge against the decision-time number.
The play must still be available at a playable price, with filters that avoid stale numbers, fragile edges, or forced low-value bets.
The final card check uses matchup context such as efficiency profile, Four Factors, pace, and availability-sensitive team strength.
ATS Model
The ATS model is a market-residual spread process. In plain English, it is not only asking which team is better. It is asking whether the sportsbook spread is too high or too low after accounting for team strength, recent form, schedule context, rest, market price, and matchup indicators.
The current customer-facing ATS lane is the Premium Four Factor ATS product. The broader Balanced ATS Governor is kept as a research archive because it had more volume, but the tighter premium lane is the selected public product.
Totals Model
The NBA totals product prices the expected scoring environment for each game. It combines pace, offensive efficiency, defensive efficiency, shooting profile, free throw rate, turnover rate, rebounding context, and market information to decide whether the posted total is too high or too low.
CLV
Closing-line value, or CLV, compares the line the model bet to the final market line near tipoff. If the model bets Over 224.5 and the market closes 226.0, that is positive CLV because the card beat the closing number by 1.5 points. If the market closes 223.5, that is negative CLV.
CLV does not guarantee that a single bet wins. A great number can still lose and a bad number can still win. Over a larger sample, though, positive CLV is one of the cleanest signs that the process is finding prices the market later agrees were valuable.
Customer Framing
The goal is transparency. Customers should be able to see the tested rules, the bet count, the exact record, the financial return, and whether the market moved in the same direction as the model.
Important Note
The NBA backtest is evidence of how the model and Three Check governor performed under the stated rules. It is not a promise of future profit. Sports markets change, line availability changes, player availability matters, and future seasons will not unfold exactly like the tested window. The value of the page is that the method, rules, volume, record, ROI, and CLV are shown clearly before a customer relies on the product.