# BTC 5m vs 15m vs 1h signal comparison — frozen protocol

Recorded before the first comparison run on 2026-09-25 UTC. Purpose: answer whether a retail trader should expect a fixed EMA9/EMA21 signal to behave differently on completed BTC 5-minute, 15-minute and 1-hour candles after modeled costs. This does not select the historically best timeframe as a future recommendation.

## Main design

- Same Binance BTCUSDT spot market and main UTC interval `[2025-09-25T06:00:00Z, 2026-09-25T06:00:00Z)`. Require every expected 5m, 15m and 1h bar; otherwise stop publication rather than silently compare different windows.
- 15m candles are aggregated only from three contiguous archived 5m bars. 5m and 1h are archived; they were retrieved retrospectively and are not claimed as real-time point-in-time capture for the whole period.
- A completed EMA9/EMA21 upward cross opens one long at the next bar's open. Downward cross closes at next open. Stop 2%, target 4%, maximum hold 48 hours. One position maximum; initial $10,000 cash, $500 margin, 1×, 0.05% fee and 0.02% adverse slippage per fill. Stop wins ambiguous same-bar stop/target touches.
- EMA periods count bars, so their calendar lookbacks differ by timeframe. This is a comparison of three commonly selected chart presets, **not** a controlled experiment isolating only sampling frequency.
- Main outcomes: signal/completed-trade count, net account return, fee total, max sampled equity drawdown, median trade P&L, first/second-half net trade P&L. Double fee and slippage is a preset sensitivity. Fully invested BTC holding is contextual only because the strategy usually holds much less capital at risk.
- Trade-level rows and sampled equity paths are public. Monthly trade-P&L totals are a descriptive dependence check; no independent-trial p-value is inferred. A later release may examine RSI or higher-timeframe confirmation separately, without rewriting this prespecified example.

## Invalid interpretation

This one BTC asset and one authored EMA rule do not establish which timeframe is best for all indicators, market regimes or investors. Historical bar prices do not prove exchange fills, limit queue priority, funding or live model decisions. Zero missing bars does not remove survivor, execution or overfitting risk.