Original fantasy football studies, run on our own database and required to replicate across seasons before we publish them. Every claim here traces back to a sample size and an effect size — including the ones that failed.
We tested 272 stat fields across four positions. Only 41 replicated. The complete record of what predicts fantasy production and what doesn’t — including the advanced metrics that fail, what consensus betting lines know that box scores don’t, the matchup myths that died in testing, and a shelf of our own discarded ideas.
Replication or it doesn’t ship. An effect has to hold the same direction across every season we test before it informs anything on this site. One good season is a coincidence.
We publish the failures. Every study carries a section for what didn’t work, including our own models losing to simpler baselines. Findings you can’t check against their failures aren’t findings.
Effect sizes, not adjectives. “Target share matters” is a take. “+0.217 across 3,222 player stretches, same sign five straight seasons” is a result.
Research, not picks. We study consensus betting lines because they are the most honest public forecast of player production available. Nothing we publish is betting advice.