Fees eat the edge: 26,154 indicator setups by timeframe
We ran 26,154 indicator setups on 12 coins traded in Korean won. Before costs, short timeframes looked fine. After fees and one tick of slippage, only 2.5% of five-minute setups made money.
How this was made: the code, the research runs and the first drafts of this text were produced by an AI assistant (Claude) working under the operator's direction; the operator made every decision and checked the numbers. All trading was on paper: real prices, no real money. Nothing here is investment advice.
What we measured first
Our first live-price paper trader used market orders on small Korean-won coins. Its median round-trip cost was 0.46%. Its signals were right about direction 58–63% of the time, but the average move they caught was +0.04% to +0.09%. A trade that wins 60% of the time and makes 0.06% before costs loses about 0.4% after costs, every time.
The big test
Sixteen indicator families (moving-average crosses, RSI, MACD, Bollinger, Donchian, Keltner, stochastic, SuperTrend, CCI, Williams %R, ROC, VWAP, ADX and more): 307 single settings, plus combinations with filters and exit rules, for 4,359 setups per timeframe and 26,154 in total. Twelve coins on Upbit (BTC, ETH, XRP, SOL, DOGE, NEAR, SUI, XLM, ADA, ARB, BCH, LSK), January 2024 to October 2026 (five- and ten-minute bars from January 2025). Orders filled at the next bar's open. Costs: Upbit's 0.05% fee per trade, plus one price tick of slippage per trade (at least 0.05%).
Why short timeframes lose
- Costs are fixed per trade but moves shrink with the timeframe. Five-minute moves are small, so 0.1% to 0.2% per round trip takes a large share of every trade.
- Cheap coins have coarse price steps. A coin trading at 133 won moves in 1-won ticks, so one tick is 0.75%. Crossing the spread once can erase a day of edge.
- More trades means more chances to pay. The best five-minute setup made +3.6% a year after costs; none passed our consistency test on both the training and the test period.
Does that mean four-hour rules work?
In backtests, more of them survive costs: 69.5% of four-hour setups were positive after slippage, during a period when simply holding the same coins also made money. But backtests have their own traps (survivorship bias), and the best four-hour rule we ran live still lost over two weeks (server test). Surviving costs is necessary, not sufficient.
Check your own numbers with the bot vs. fees calculator.