About Edge or Not
Trading content is full of signals that look convincing on a chart. We take them one by one and check, with numbers, whether they beat random entry after costs.
How every test works
- Rules first. Entry, exit, timeframe and the pass criterion are written down before any run. No tuning after seeing results.
- Controls. The same signal is compared with random entries, the opposite side and shuffled data. A signal is alive only if it beats all three.
- Costs included. Spreads and fees are subtracted from every trade.
- Out of sample. History is split; a result that holds only on the part used for design is treated as overfitting.
- Published as is. "No edge" is a result, not a reason to stay quiet.
Data
Public exchange data collected by our own servers: Binance and Bybit liquidations, open interest, taker volume and candles every 5 minutes, and order book depth from Binance spot and OKX perpetuals every minute.
What this site is not
Not investment advice and not a signal service. Models are estimates. A tool can be useful for reading the market and still have no predictive edge - we say which is which.