ERROR
The miner failed
type str doesn't define __round__ method · Refreshing in the background; reload shortly.
These are the live 1-minute capture snapshots, not the 5-year option-chain backfill. They are two different datasets: this one fills every minute the market is open; the other is what /thesis is still waiting on.
No pattern cleared the sample floor. That is a result, not a failure — it means the captured history does not contain a repeatable setup at this horizon.
Train is how often the pattern worked on the data it was found in. Holdout is how often it worked on data it had never seen. A big drop between them means the pattern was fitted to noise — and a miner reporting only the training number will find a beautiful pattern in pure randomness, every time.
The 50% line is drawn on every bar because that is the coin flip these have to beat. A mined pattern is a hypothesis. None of them reaches a trade, a score, or a size until it survives a real forward test.