One method for every setup
- Rules as live. Each backtest replays the live scanner's own rule code (Trend Breakout: the rules of a published research paper, with the pass test written before the first run; Smart Trend: the same breakout with two skip rules picked from what its losing trades had in common; Regime Trend: the same breakout with a market switch and a regime exit, chosen in the third of three research rounds with the pass rules written before each round). Backtest trades replayed through the live scanner give the same entry and stop.
- Costs. 0.1% Binance taker fee on each side and 0.02% slippage on the entry and on the exit, both market orders at the daily close.
- R. Every result is in R: 1R is the loss if the starting stop is hit. "Before costs" shows the same trades without fees and slippage, which tells you how much of a setup's edge goes to the exchange.
- Periods. History from 1 January 2019 (or a coin's Binance listing) to today, so it covers the 2021 bull market, the 2022 bear market and the years since; results are shown per calendar year and for 2019-2022 against 2023-2026. Smart Trend's skip rules were picked on all of these years, so its numbers are in-sample and its page adds a walk-forward test (rules picked only on the years before each test year). Regime Trend was found on the third look at the same history, so its numbers are in-sample too; its live record from 19 October 2026 is the clean test.
- Live caps. In strong markets the raw backtest has dozens of trades open at once, which no account could follow. The "live caps" figures replay only the trades the live signals allow: at most 3 new a day and 6 open, most liquid first.
- What can't be replayed. Live signals pass an AI review (approve, reduce or veto). The backtests are the raw rules.
- Survivorship. Trend Breakout and Smart Trend use today's Binance pairs (the liquid pairs DanioX scans; today's top 100), so coins that died are missing, which flatters a long-only system. Regime Trend uses every Binance USDT pair ever listed, so coins that later collapsed or were delisted (LUNA, FTT …) are in its test while they were among the 20 biggest. On that honest list the plain 20-day breakout lost money in 2022, 2025 and 2026.
Retired setups
Six earlier setups were backtested the same way over 2023-2026 (15-minute to 4-hour candles) and none was clearly profitable after fees: CTP and Spot Swing were about break-even, FGA looked good on its design data but lost on the earlier, unseen years, and AA, AMD and AMD Structural lost money in every year. They were retired on 19 October 2026 and no longer publish signals. Their backtests and signal history are kept, and the AI reviewer still learns from them.
The scripts live in the DanioX signal bot (rtrend_backtest.py, trend_backtest.py, strend_backtest.py, with unit tests for every rule function) and produce the JSON these pages read. Back to the live signal board.