Every prediction is recorded the day it's made, scored when it matures, and dead companies stay in our numbers. This page is the scoreboard — it updates nightly from the same ledger our models are judged against.
Nothing backfilled, nothing hand-picked, nothing quietly deleted.
A prediction only counts when its full horizon has elapsed — 60, 90, or 180 days after it was made. Until then it sits on the record, in flight, waiting to be scored. Nothing here is backfilled or hand-picked.
This is a reliability diagram — the standard honesty test for probabilistic forecasts. We group matured predictions by the probability we originally published, then plot what actually happened. A perfectly calibrated model sits on the diagonal.
Worked example: suppose we publish 100 picks, each with a “60% probability of rising.” If about 60 of them actually rise by maturity, the dot lands on the diagonal — our 60% really meant 60%. Dots above the line mean we were too cautious; below, overconfident.
Most services never show you this chart, because it can't be massaged: the probabilities were published in advance and the outcomes are just arithmetic. We plot it for you either way.
Plain language, no fine print. These rules are enforced by the pipeline itself, not by an editor choosing what to show.
The scoreboard updates nightly, and it doesn't forget. See the product behind the record, or take the 3-minute tour first — no signup needed.