mt5-paper-tradelisted

# Automated demo trading ```bash python tools/mt5_paper.py --once # dry run, sends nothing python tools/mt5_paper.py --rule sma_cross --live --once # one live demo cycle python tools/mt5_paper.py --rule random --live --interval 300 python tools/mt5_paper.py --close-all --live # flatten its own positions ``` This is the only module in the project that sends orders, kept separate from `mt5_account.py`, which reads and cannot write. ## The fence 1. **DEMO accounts only.** `account_info().trade_mode` must be 0. REAL, CONTEST and any unrecognised value all abort before an order is constructed. Verified against all four cases. 2. **Dry-run by default.** `--live` is required to send anything. 3. **MAGIC-tagged orders**, so it only ever closes positions it opened. 4. **Caps**: `--lot`, `--max-positions`, `--max-daily-loss` (halts the session). Never relax these to make a run work. If a user asks to point it at a live account, the answer is that this tool cannot, and a live executor is a separate decision needing explicit discussion of size, drawdown and failure handling. ## What to tell the user about results **None of the rules has a measured edge.** This repository backtested its own composite at IC 0.002 and found zero of 105 tested patterns surviving multiple- testing correction. Expect roughly break-even before costs and negative after. The `random` rule is the benchmark, and comparing against it is the point: a rule that cann