Last 30 days: not enough graded calls yet. (n = 10, we wait for 30)
Every hour a model looks at BTC, ETH and SOL and decides whether a move big enough to pay trading fees is coming in the next 12 hours. The call is written to the database before the outcome exists, then graded against the real Binance close. Wins and losses stay up. Mostly it says nothing, and we print that too.
since the first live receipt, 2026-11-04 11:00 UTC. A coin-hour is one coin, one hourly check.
The model only calls when P(big up) minus P(big down) clears 0.16 either way. Inside the hatched band it stays quiet, because in testing smaller leans did not pay for the 0.2% in fees.
BTC lean +0.04. Needs 0.16 to call. Not even close.
ETH lean +0.12. Needs 0.16 to call. Getting warmer.
SOL lean +0.08. Needs 0.16 to call. Getting warmer.
Mild hopium detected.
Not financial advice. Crypto is unregulated and can be highly risky. Predictions are probabilistic and often wrong.
Mild hopium detected.
Not financial advice. Crypto is unregulated and can be highly risky. Predictions are probabilistic and often wrong.
Mild hopium detected.
Not financial advice. Crypto is unregulated and can be highly risky. Predictions are probabilistic and often wrong.
The model sat out all of the last 24 hours. Discipline or cowardice? You decide.