SportsQuant

SportsQuant is an NFL evidence workbench. Ask a question about what has already happened, and it answers from historical play-by-play data. Completed results show the sample and seasons used, plus an interval when the method supports one. It is built to hand you evidence you can examine, not a recommendation to follow. Three ways in. One body of evidence. SCENARIO — build the question Compare a player's per-game numbers with and without a teammate on the field. How did a back's yards per carry hold up without his usual quarterback? Scenario shows both sides. Or split one player's numbers by a condition: home or away, roof, surface, temperature, wind, down, field position, pass depth, or formation. Or set two players side by side and compare them directly. Every Scenario reports the number on each side, the difference between them, and a confidence interval around that difference. It also shows how many games sit on each side and the seasons it evaluated. It closes with one of three verdicts: "Real difference," "Probably noise," or "Not enough data." The last two are answers, not failures. A thin sample is information about your question. ASK — plain English in, Scenario out Ask is the plain-English mode inside Scenario. Type a question. It shows you exactly how it read it, then runs it as an ordinary Scenario you can inspect and run again. It knows its limits. An unknown player, a measure the app does not carry, or a question about what happens next gets a refusal that names the limit. Two players with the same name get a follow-up asking which one you meant. It does not invent an answer it cannot ground. ANALYZE — regression with a receipt Run a regression over filtered game and play data. The result assembles on screen as it computes, then settles into a receipt you can read. The receipt: coefficients with confidence intervals, parametric and bootstrap. P-values with significance markers. R² and adjusted R². An effect-size band. Average error in the units you modeled. A collinearity (VIF) reading for each input variable. A residual density strip with its skew. Models are linear, polynomial, exponential, and logistic. For supported model families, Lasso, Ridge, and Elastic Net regularization are optional. Quantile regression is available where supported and stays off unless you turn it on. Categorical inputs like quarter, down, and roof are encoded for you. You see per-level effects, such as "Quarter: Q2 (vs Q1)," instead of one averaged slope. Filter by season and week, team, temperature and wind, roof and surface, day of the week, home or away, and numeric ranges on the stats. Receipt fields appear on results that produce them. A run with too little data, or one the app declines, says so plainly in place of a result. SAVED WORK Analyses save to the Saved tab. Scenario cards save to the Scenario tab. Both reopen later, on either plan. WHO IT IS FOR Fantasy analysts and people who like NFL data. Maybe you have wondered how a receiver's targets looked in cold-weather games. Or whether a back's yards per carry changed without his usual quarterback. This is the place to check, and it will tell you when the sample is too thin to say. FREE AND ELITE Free to start. One subscription, Elite, gives you more room to run the same product. It does not unlock a different or better class of result. Every number here arrives with the window it came from. Terms of Use: https://appquant.app/sportsquant/terms