MY LIFE · TRACK RECORD
We score our own forecasts
Every monthly forecast you open is written down before the period it covers, with a stated probability. Once the period ends we ask what actually happened, and grade ourselves on it.
Nothing on record yet
Open a monthly forecast and it gets written down here automatically, along with how likely we thought it was. Come back once the month has ended.
See a monthly forecastWhat happened?
These periods have ended. Answering them is what makes the score below mean anything — and answering the ones that went badly matters more than the ones that went well, because a record built only from happy memories is worthless.
The score
By kind of prediction
Scored separately on purpose. Skill is measured against how often a thing happens at all, and that differs by kind — pooling them would let one flatter or punish the other.
Are our probabilities honest?
When we say something is 70% likely it should happen about 70% of the time. Bars sitting above the marker mean we were too cautious; below means we were overconfident.
Everything on record
| Prediction | Period | We said | Outcome |
|---|
How this is scored
- Skill against the base rate is the headline, not accuracy. If things go well 70% of the time, predicting "well" every time is 70% accurate and worth nothing. Skill measures how much better we do than a forecaster who knows only that 70% and says it every time. Zero means we add nothing.
- Nothing is published below 20 answered predictions, or when fewer than half the finished periods were answered. A small or self-selected sample flatters whoever is holding it.
- Predictions cannot be edited or deleted once recorded, and an answer cannot be changed from happened to did-not. This is enforced in the database, not just in the page.
- Claims are always relative to your own usual, never "a good month" in the abstract — an absolute claim cannot be scored.