Strategy lab

Every strategy replayed with the live trading rules on years of history the model never saw, including fees, slippage and pessimistic fills. Auto-trade only uses what is proven here.

Strategies tested

—

coin × horizon × rule

Proven families

—

of — tested across coins

Usable for buying

—

auto-trade can use these

Not usable

—

auto-trade ignores these

Strategy families — the same rules on every coin

One coin's record is rarely enough to tell skill from luck. A family pools the same rules across all coins, so it has hundreds of trades to judge. A coin may use a proven family only if its own record with it is positive too.

How big the wins are. Every trade risks 1R (the distance to its stop-loss, 0.8σ of the horizon's expected move). “Win / loss” is the average winning and losing trade after fees, so you can see the margin behind the average, not just the win rate. Targets were set by replaying each family with several exits on the same unseen data: a wider target made the average win of every buy family bigger (trends run further than the old 1.3σ target), so the proven 30-day buy family now aims for 2.6σ with the stop moved to entry at +1R. Wider still earned more per trade but its drawdown failed the proof limit, and shorts got worse, so they keep 1.3σ. Fewer trades hit a wider target, which is why its win rate is lower while its average is higher.

Per coin: the coin's own record, judged separately for buy and short trades

Backtests are simulations: markets change, and a strategy that worked on past data can stop working. Results assume 0.1% fees and 0.05% slippage per side, stops filled at the worse price on gaps, and 1% of equity risked per trade.