GodzillaBTCSignal methodology · v4.2
An auditable rules-based research engine; profitable performance is not yet established.
Data and indicators
Only completed, positive, internally consistent OHLC bars are used, with a 60-second settlement buffer. Wilder-smoothed ADX/ATR replace the old calculations. Breakouts compare the current close against prior candles. New entries require a verified bid/ask quote, fresh setup bars, daily context and hourly timing. Price-only fallback feeds are display references.
Setup families
Trend and range-fade rules remain separate. Trend targets are 2R Confirmed, 1.5R Standard and 1.25R Tactical. Range targets are 0.75R, 0.65R and 0.50R. Stops define 1R before costs; gaps can exceed that loss. Hurst, efficiency and entropy are descriptive features, not independent proof of an edge.
Historical research replay
Replay enters at the next bar open, handles adverse gaps, retains both-hit bars as conservative ambiguous losses and deducts a 0.05R cost floor. It removes overlapping trades and purges the chronological 70/30 split. Positive net holdout expectancy must survive doubling the assumed cost. Higher-timeframe context is applied where coverage exists. The technical gate is shared, but live macro events, portfolio filters, actual spreads and publication latency are not fully replayed.
What the percentage means
The percentage is a smoothed historical target-hit rate from comparable technical setups, not a calibrated live win probability. The holdout is reused for model gating; it is not a pristine, once-only test set. Small samples, serial dependence, incomplete context history and repeated selection remain limitations. The engine withholds percentages when evidence is inadequate.
Risk and publication
Kelly-inspired caps use the holdout sample, uncertainty and assumed costs. Caps are 1.00% Confirmed, 0.65% Standard and 0.35% Tactical; zero calculated risk withholds publication. Atomic recording prevents duplicate setups, overlapping signals on one symbol, duplicate FX exposure in the same USD direction, and more than three open paper positions. Correlation comparisons align return intervals by timestamp. Unavailable calendar or ledger information blocks new entries.
Forward accounting
Publication freezes the entry, stop and target. Outcomes use hourly candles; expiry rounds up to the next UTC hourly close. Crossings inside the partial publication candle, both-hit bars and unexplained missing coverage are AMBIGUOUS and charged a conservative -1R gross plus assumed costs. That is an accounting assumption, not an observed broker loss. Gap stops can lose more. Original closed records are preserved; pending legacy rows receive corrected timing.
Health and operation
A -2R current-model daily circuit breaker includes ambiguous and cancelled accounting results. A rolling likelihood health heuristic can pause the model; it is not a formal SPRT with guaranteed error rates. Pauses require review. The package generates and evaluates signals when its API is requested; uploading this ZIP does not configure an unattended scheduler.