Lesson
Why We Publish Every Rejection Too
Most of what the model looks at, it passes on. That's not a weak board. That's what an honest one actually looks like.
Quick answer
A release rate shows what share of evaluated markets a model actually plays. On a typical slate, only around 19% of evaluated markets get released; the rest are published as rejections, not hidden, so the track record can actually be checked.
A board that only shows hits isn't a board
If a service only ever shows you the bets it likes, you have no way to know whether it's actually selective or just quietly deletes anything that didn't work out. You're seeing a highlight reel, not a track record. There's no way to check the math because there's nothing to check it against.
What "release rate" actually tells you
On most days, the vast majority of markets we evaluate get rejected. That's not the model failing to find bets, that's the model doing what it's supposed to do. A real edge is rare. If most of what showed up on the board were "playable," that would be the warning sign, not the reassurance.
A typical slate
| Markets evaluated | 78 |
| Released | 15 |
| Release rate | 19.2% |
The other 80% weren't hidden. They're on the board, marked as what they are: evaluated and passed on.
Why the rejections matter as much as the releases
A rejection is a real data point. It tells you the model looked at that market and decided the price wasn't good enough, or the lineup wasn't confirmed, or the edge didn't clear the bar. Publishing it means you can go check our work on the ones we passed on, not just the ones we liked.
What this protects you from
It's easy to build a "model" that looks great by only ever showing the good calls after the fact. Publishing every rejection, permanently, removes that option entirely. Whatever the record actually is, wins and losses, hits and passes, it's the same record whether it makes us look good that week or not.