EVAL PASS BOT
version 1.1
built 2026-08-22

PURPOSE
  Pass a Topstep $50K Trading Combine evaluation. Optimized for
  probability of passing, not for long-run profit.

STATUS
  Risk engine ...... TESTED   32 unit tests against Topstep's rules
  API adapter ...... WORKING  TopstepX / ProjectX, dry run default
  Order flow ....... TESTED   OFI, delta, book imbalance, gate
  Strategies ....... UNVALIDATED  no real-data evidence of edge

KNOWN GAP
  The three included strategies have never been tested on real MES
  data. Pass probability ranges from 0.2% to 100% depending entirely
  on whether the strategy has an edge, which is unmeasured. Run
  fetch_data.py and backtest before funding an evaluation.

WHAT IS PROVEN
  - Trailing drawdown, daily loss limit and consistency rules are
    implemented correctly and enforced as hard constraints
  - Position sizing shrinks to zero as the drawdown buffer shrinks
  - A negative-expectancy strategy is stopped by the risk layer
    rather than being allowed to run the account down
  - Required edge to pass: roughly 40% win rate at 2:1 reward/risk
  - Optimal risk per trade: $75-100 with no evaluation deadline
