Entry radar & alarms
Start paper trading for live ICT bias, strength meter, and entry alarms (sound + optional desktop notification).
Each successful backtest saves metrics + vol/trend regime tag to
data/runs.sqlite3 for comparison across market conditions.
Opens, scale-outs, and stops from paper and backtest with entry/exit reasons for review —
data/playbook.sqlite3. Edit strategy/spec.yaml →
market.watchlist for the 10 pairs polled in paper (capital split evenly).
| When (UTC) | Mode | Sym | Event | P&L | Reason |
|---|---|---|---|---|---|
| Loading… | |||||
Walk-forward: 70/30 train vs held-out test, composite score (OOS Sharpe weighted over in-sample). Stress: same ICT rules on crisis date slices. Evolve: small random search + selection; top genomes re-scored with crisis penalty (silent backtests — no playbook spam). Uses backtest fields on the left. Apply pushes rank-1 genes into the running server (backtest + paper + lab) until Reset or restart — does not write strategy/spec.yaml. Multi-symbol evolve: parallel backtests (full INITIAL_CAPITAL each) — robustness aggregate is min Sharpe / PF and worst DD across coins, not a single pooled portfolio. Chief QE adds segmented stability-sweep (rolling min-Sharpe / min-PF span) and cost-stress (scaled commission + slippage); VPS: python scripts/chief_qe_sweep.py.
On KZ_AUTO_RESEARCH=1, the server queues evolution when UTC session leaves a kill zone (debounced). Runs are stored in data/kz_research.sqlite3. Manual run uses the same lighter pipeline as the auto job.
| UTC | Trigger | Sym | TF | Fit fast | Fit+crisis | Status |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Save GO / HOLD with notes plus an OOS metrics snapshot from rank #1.
Stored in data/promotion_decisions.sqlite3 — append-only style history for desk sign-off.
- Run Evolve genes above and wait until it finishes — this panel binds to the latest successful run.
- Add a decision note (why GO/HOLD, window, policy e.g. regime gate).
- Chief QE (required for GO): run stability sweep and cost stress in Research lab for the same window/symbols, then tick the acknowledgment below. Server rejects GO without it.
- Save GO only logs the decision. GO + Apply runtime logs GO and writes rank #1 genes to runtime (stop paper trading first).
- Refresh log reloads the table from SQLite. If metric cells show —, that row predates a full snapshot or evolve wasn’t used — run Evolve again before the next GO to fill Sharpe / PF / DD / trades.
Table columns & symbols
- min Sharpe / min PF — worst coin in the OOS test window (multi-coin robustness).
- DD % — worst max drawdown % across symbols (negative = drawdown).
- Trades — total trades summed across symbols in that snapshot.
- GO saves are confirmed if the numeric snapshot is incomplete, so you don’t silently log empty metrics.
- Table vs note: cells are the evolve OOS test snapshot (rank #1). Your note may cite a different window (e.g. full-history verification) — different numbers are expected and both can be valid.
- Capital model: multi-coin lab runs are parallel sleeves (each symbol gets full configured capital), not one cross-margin portfolio.
| UTC | Decision | min Sharpe | min PF | DD % | Trades | CQE | Note |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
Total P&L
$0.00
0.00%
Win rate
0%
0W / 0L
Sharpe
0.00
Risk-adjusted
Max drawdown
0%
Peak to trough
Total trades
0
Final —
Profit factor
—
E —
Price
Run a backtest to load OHLC from Binance for this range.
Equity curve
Signal strength (last 100 bars)
Recent trades
| Time | Side | Entry | Exit | P&L | P&L % |
|---|---|---|---|---|---|
| Run backtest to see trades | |||||