What TheHomeRuns.org is trying to prove
A first field note on the point of the project: clearer home run predictions, public accountability, and a model that gets sharper because it is forced to explain itself.
Field notes from the home run model.
TheHomeRuns.org exists because home run prediction should feel less like a black box and more like a living argument.
Every slate is noisy. Weather changes. Lineups drift. Pitchers get scratched. A hitter can square three balls and still leave with nothing but loud outs. That is the cruel part of baseball, and it is also the reason the work is interesting.
The goal here is not to pretend the model sees the future. It does not. The goal is to make a better daily case for where home runs are most likely to come from, then keep score honestly enough that the model has nowhere to hide.
The thesis
A good home run slate needs more than one number.
Raw power matters. Barrel rate matters. ISO matters. Pitcher weakness matters. Park and weather matter. Bullpen context matters. Recent form sometimes matters, but only when it does not trick you into chasing noise.
The useful thing is the combination: batter power against pitcher vulnerability inside a real game environment, with enough humility to say, "This looks strong, but here is why it could fail."
That is what the score is trying to become.
Why the site is public
The model gets better when its misses are visible.
If a pick worked, we should be able to explain whether the signal was real or whether the outcome bailed us out. If a pick failed, we should be able to ask whether the process was still sound or whether the model overweighted the wrong thing.
That is why this blog exists. Some posts will explain how the model works. Some will be postmortems. Some will be product notes. The common thread is accountability.
What comes next
The next layer is sharper editorial proof:
- model explainers for every major signal
- weekly postmortems on what the slate got right and wrong
- playbooks for using the agents, simulator, and conviction tools
- field reports when the model finds something worth unpacking
The promise is simple: if the site says a matchup has an edge, it should also help you understand why.
The home run is the loudest event in baseball. The work now is making the prediction process worthy of that noise.