feat: queue simulation validation — VBT vs realistic event-driven PnL comparison
Track 4 — quant/sim_validate.py:
SimulationValidator: runs VBT backtest, then replays signals through
Phase 3 event-driven simulator with queue position, partial fills,
adverse selection, and cancel latency.
Compares VBT (optimistic candle fill) vs simulator (realistic) PnL.
BatchSimValidator: runs multiple param combos through the pipeline.
Grid MM simulation results (all DISCARD):
VBT S=+0.78, ret=+1.5% → Sim S=0.00, ret=0.0%, 100% adverse
VBT S=-0.03, ret=-0.4% → Sim S=0.00, ret=0.0%, 100% adverse
VBT S=-2.74, ret=-6.1% → Sim S=0.00, ret=0.0%, 100% adverse
Grid fills happen during momentum (price crossing levels),
not mean reversion. Adverse selection rate approaches 100%.
VBT candle fills are fundamentally optimistic for grid strategies.
269 tests passing.
ramseshk
cdc7a01986