Inside Benchlytics: how our player projections are built
A plain-language look at the inputs, the modeling approach and, importantly, the limits of what a projection can tell you.
Priya Sharma
Head of Analytics
We get asked a lot how the numbers are made. We think that's a healthy question, you should understand any tool you rely on. Here's the honest version, without the black-box mystique.
The inputs
Every projection blends several categories of publicly observable information: historical and recent performance, role within the team, expected opportunity (minutes, touches, usage), matchup context, pace and game environment, and, where relevant, venue or park factors. When lineup news or injury reports land, those feed in too.
The approach
We model each player's expected production as a distribution rather than a single point, which is why you see floors and ceilings, not just a midpoint. The models are re-fit as new data accumulates through a season, and they're calibrated so that, over many games, the ranges mean what they say, an outcome we described as a 'ceiling' should actually be a ceiling most of the time.
We deliberately favor approaches we can explain over ones we can't. If a recommendation can't be reasoned about, it's much less useful to you.
The limits
Here's the part some companies skip: projections are estimates, and sports are genuinely unpredictable. No model, ours or anyone's, can promise a result. Injuries, coaching decisions and plain randomness will always defeat perfect foresight. We report our numbers as research inputs, and we're explicit that the decisions, and the outcomes, are yours.
That honesty is a feature, not a disclaimer. A tool that pretends to certainty is a tool you can't trust.
This article is for informational and entertainment purposes only. Benchlytics provides analytics and research, not betting or wagering services, and does not guarantee any result. Play responsibly.