Does Offense or Defense Carry Over More? Year-to-Year Reliability, 2021-2024
Across 90 season pairs (2021-2024), runs allowed per game carried over at r=0.50 while runs scored carried at just r=0.28. Run prevention is the stickier skill.
Category
Projections & Methodology
How the models think. The machinery behind win probability, playoff odds, and forecasting — why models disagree, and what their numbers really mean.
Does Last Year Predict This Year? Season-to-Season Carryover in MLB
A team's winning percentage carries to the next season at just r=0.54 — last year explains under a third of this year, and teams regress about halfway to .500.
Regression to the Mean, Explained: Why Hot Starts Cool Off
A .400 April never lasts, and there's a reason. What regression to the mean really is, which stats stabilize fastest, and how projection systems lean on it.
Does Spring Training Performance Predict Anything? What the Research Says
Every March, hot spring stats spark hope. What the research actually says about whether spring-training performance carries into the regular season.
How Win-Probability Models Work (And Why They Disagree)
That live win-probability percentage has a model behind it. How win expectancy is built, why two models disagree on the same play, and what the number really means.
Why Playoff Odds Models Disagree: A Look Under the Hood
FanGraphs says 62%, another model says 48%. A look under the hood at how playoff-odds simulations are built — and why reasonable models reach different answers.
How Projection Systems Work: Building a Simple Marcel Forecast
Projection systems aren't crystal balls — they're weighted averages with regression. Build a simple Marcel-style forecast from real data and see how they think.