i.
Half-Kelly only.
Full-Kelly assumes zero estimation error. The realized edge is always noisier than the backtest. Every adapter in trading-algo sizes at half.
Math & Economics at the University of Toronto. Builds quantitative trading systems.
Pricing the priceless & proving the obvious.
Full-Kelly assumes zero estimation error. The realized edge is always noisier than the backtest. Every adapter in trading-algo sizes at half.
Deflated Sharpe (Bailey & López de Prado), not a point Sharpe. PBO and White's Reality Check on every config before it leaves research. The 12 documented limitations are in the repo.
TRADING_ALLOW_LIVE in env. --allow-live on the command line. Paper-only enforcement explicitly disabled. Per-order YES at the terminal. Confirmation callback wired through the CLI. Missing any of them blocks the call.
ibkr_tool, flex_tool, kite_tool. Structured envelopes, stable exit codes, idempotency keys. SEBI-compliant jsonl for kite-algo. An agent action that isn't in the log can't have happened.
The shelved Gemini trader is still in the repo. The crypto book lags BTC buy-and-hold and the README says so. Trailing stops hurt Sharpe on every config; off in V11.