0a93af10a90801e92ced3cf3ebc8db73b3bbc76f
- indicator_service.py: _detect_order_blocks() now requires the confirming candles to actually displace price past the OB candle's own high/low (the "imbalance" a real SMC order block is supposed to leave behind) instead of accepting any 2-candle same-direction sequence, even one that stays entirely inside the OB candle's range. strength (1-3) now reflects real displacement/volume confirmation instead of always 1. - indicator_service.py: removed a stray "+ 0.001" from Tweezer Bottom's zero-division guard that Tweezer Top's otherwise-identical formula didn't have -- both already guard via an explicit > 0 check, so the epsilon just made the two directions inconsistent for no reason. 237 backend tests pass (+5). This closes out the full 13-algorithm audit started in v10 -- only (tt) parameter re-optimization (needs empirical walk-forward runs, not a code change) and (yy) multi-position portfolio backtesting (needs a new architecture) remain, both flagged as separate larger investment decisions. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Description
Test repo