How True Strength makes and grades picks

01One engine: True Strength

Every pick, projection and graded result on Slam Wager comes from one engine, True Strength. Earlier pick systems have been retired, and nothing on this page describes them. True Strength covers football (NFL and college), basketball (NBA, WNBA and men's college), MLB, the NHL and eight soccer competitions: MLS, the Premier League, LaLiga, Serie A, the Bundesliga, Ligue 1, the Champions League and the Europa League.

Tennis, golf, UFC, F1 and other sports can have research pages, but True Strength does not rate them and makes no picks in them.

The method has three steps. Project the score for tonight's game. Turn that projection into a chance for every side, spread and total, using how far real results have landed from the engine's projections. Then compare those chances with the price the sportsbooks are offering, and post a pick only when the price is clearly wrong in our favor.

02Rating every team against everyone it played

True Strength solves a whole league at once. Each team gets an offense rating (what it scores against an average defense) and a defense rating (what it allows to an average offense), plus one league-wide home edge. Every final score becomes one equation, and all of them are solved together, so scoring 30 on a top defense counts for more than scoring 30 on the worst one. Ratings are in the sport's own units: points, goals or runs a game.

Each season starts from last season's rating, but only as much of it as history says carries over, measured by how well one season's ratings predicted the next. That carry-over is worth a fixed number of games and fades as real results arrive.

Leagues are not treated the same where testing showed they differ. NBA games are weighted toward recent weeks (a 60-day half-life). College football caps blowouts (a margin past 24 points counts 30%), limits the home edge, and rates divisions and power conferences on their own levels. Men's college basketball solves on Division I games only and gives every non-Division I opponent one pooled rating. The NFL, NBA, WNBA and NHL pull each season's home edge toward last season's.

03Who actually plays

Ratings are solved with each game's real lineup in them, so a team is not marked down for a game its stars missed, and its opponent is not credited for beating a short-handed roster. Before each game, tonight's injury report and confirmed lineups are applied the same way.

Absences are priced player by player where the data supports it: each missing player's own value against the value of the player who replaces him. NFL starters are priced by position group and quarterbacks by their own games; NBA, WNBA and college basketball players by minutes and box-score production; NHL goalies by goals saved; MLB lineups hitter by hitter against tonight's starting pitcher; soccer players by minutes, goals, assists and shots.

A listed player counts by his chance to play, not as simply in or out. When a key player's status is uncertain and his playing or sitting would swing the projection by a set amount (1.5 points in football and basketball, 0.3 runs, 0.25 goals), the pick is posted as contingent: the card says what it assumes and what the pick becomes otherwise.

04Tonight's conditions and matchups

Tonight is never an average night. In the NFL, the forecast moves the total: 3.8 points off in 15+ mph wind and 2.9 off at 32°F or colder, sized from every regular-season game from 2015 to 2025. Back-to-backs are priced league-wide in basketball and hockey. A team's own record in unusual conditions (short rest, a long trip, real weather) counts only once at least four games have shown it, and is pulled heavily toward zero.

Player values also depend on tonight's opponent. A starter's worth is multiplied by a matchup factor sized from history, such as an NFL cornerback against the passing offense he faces or an NHL top-pair defenseman against a strong top line. With no measured matchup signal, the factor is exactly 1 and the reason claims nothing.

Style matchups are shown on game pages for context, but only one style term has passed testing and moves a projection: college football pace, each offense's plays and seconds per snap against what the other defense allows. On the 2025 holdout it cut the average total miss. No NFL style term beat the team ratings, so none is used.

05From a projected score to a probability

A projected score is not a bet. Real games scatter around any projection, and True Strength measures that scatter instead of assuming it. Through each season it re-solves the league at up to 12 cut dates using only earlier games, projects the next week's games, and records how far the results landed. That measured error sets the spread of outcomes for the margin and the total.

From the projected margin and total and that spread, every side gets a chance: to win, to cover each spread, to go over or under each total. On a whole-number line the chance of a push is counted separately, because a push returns the stake. The NHL puck line uses a fit on earlier seasons' results, because one-goal games and empty-netters decide it.

Where past results showed the model was too sure of itself, win chances are pulled toward 50%: the NHL keeps 70% of a chance's distance from a coin flip, the NFL and college football 85%. Each market's share is tuned further from how often its graded picks actually won. The same projection and spread drive the 20,000-game Monte Carlo simulation on each game's page.

06Price against model: when a pick is made

Each side's chance is compared with the best price on the board: the best line across the live sportsbooks at the most-quoted main number. Expected value (EV) is the chance of winning times the profit, minus the chance of losing times the stake. The game's best side becomes the pick when its EV is at least 2%. A market can be held to a higher bar.

Guardrails come first. Moneylines longer than +500 are never posted, MLB moneylines longer than +250 are never posted, a price of +300 or longer needs at least a 30% market-blended chance, and nothing at −200 or chalkier is posted. A spread or total is posted only on the side the projection itself covers, so a big price never buys a side the engine expects to lose. Soccer has a draw, so there is no two-way moneyline: soccer picks are spreads and totals.

One pick per game at most, and no parlays or teasers. When nothing clears the bar, the game is a pass. A pass is a result, not a missing pick.

07Stars and units

Which side posts is True Strength's own call. How big it is comes from a more cautious number: the engine's chance blended one third with two thirds of the market's no-vig chance for the same bet. That blended EV sets the stars, and each star is one unit.

MLB and NFL picks all post at one star and one unit. Testing showed that sizing them by edge added exposure without adding results, so the stake stays flat while the side and the decision to bet are unchanged.

A unit is your own fixed amount, set from an entertainment budget. The stars are not a confidence score or a win probability.

08When picks post and when they change

The engine runs every 30 minutes for today's and tomorrow's games (Eastern dates). A game is decided once a sportsbook prices it and while it is still at least 10 minutes from the start. An NBA spread that backs a team on the second night of a back-to-back waits until about six hours before tip, when the game-day injury report is due.

Open picks are re-checked on every run. A player newly out or back, a starting pitcher change, a spread move of a point, a total move of a point and a half, or a moneyline move worth three points of win chance makes the game get decided again, as does credible late news. The earlier call is kept in the pick's history, and when late news leaves the same pick at a worse price, the earlier price stands.

Nothing changes within 10 minutes of the start or after it. The call that stood at the start is the one that is graded, and the projection shown after the game is the one from before it, not a number rebuilt from the final score.

09Graded at the posted price

Picks are graded from the final score at the price and line they were posted at. A win pays that American price on the units risked, a loss costs the units, and a push returns the stake. A posted pick that later breaks a price rule is taken down before the start; if its game has already started it is voided, never graded.

The members' record keeps every graded pick, losses included: overall, by sport and by market, split into favorites and underdogs and overs and unders. The headline counts each pick at a flat 1 unit so a few big stakes cannot flatter or hide the results, with the result at Slam's 1–3 unit sizing shown beside it. Passed games are also settled with their final score, so the projections are checked on every game, not only the ones picked.

10Closing line value

After a game starts, each pick is compared with the closing price for the same bet: Pinnacle's last pregame price with the margin removed or, when that is not available, the close at the book the pick was posted from (DraftKings when that book has none on record). If the line moved, the close is carried to the posted line using the measured spread of outcomes.

Closing line value (CLV) is the close's fair chance minus the chance implied by the posted price. Positive means the pick was taken at a better price than the market finished at. CLV appears in the members' record table, by sport and market.

CLV is a diagnostic, not proof. Over hundreds of picks, beating the close is a strong sign of a real edge; over a short record it still swings with luck, and a pick can beat the close and lose.

11How the engine learns, and what it refuses

Each sport's adjustments are saved as dials the engine reads at projection time. Once a day, a sport's scoring level and home edge move a quarter of the way toward what the finished games say, capped per night, and each market's share of the model's edge moves toward how often its graded picks actually won. A dial with no games behind it never moves.

New adjustments are tested before they count. The past is replayed one week at a time, using only what was known before each game, and an idea stays out unless the predictions get closer to what happened. Plenty fail: weighting missing NBA players by their minutes alone, extra rest days, three games in four nights and long road trips added nothing measurable beyond back-to-backs.

12What True Strength does not do

It does not guarantee anything. Projections, probabilities and EV are estimates, and picks lose. Past results, including closing line value, do not promise future results.

It works from the data it can read: league scoreboards and box scores, official injury reports where they exist, posted lineups, forecasts and sportsbook prices. College basketball has almost no public injury reports, so tonight's availability there is estimated from the previous game's rotation and labeled as an estimate. When a source is late or missing, the game waits or is passed rather than filled in.

Use Slam Wager only where betting is legal for you, with money set aside for entertainment and limits set in advance. Never chase losses or bet money you need.