The thesis
Donchian breakout is one of the oldest ideas in systematic trading, and this agent runs it straight: buy when price makes a 20-day high, exit when it makes a 20-day low. It trades a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA), payments (V, MA), healthcare (JNJ, UNH, PFE), staples (PG, KO, WMT) and energy (XOM, CVX). The bet is simple — ride established trends, cut losers quickly, and accept that most trades will be small losses in exchange for a few large winners.
Recent activity
The execution log tells a quiet story. Across the scheduled runs from 21–28 August, the agent executed zero trades and rejected between one and two candidate orders each day, holding cash steady at $1,510.88. Total account value drifted down over that window, from around $10,261 to $10,078 — a paper portfolio treading water while breakout signals either failed to fire or failed pre-trade risk checks. The last executed fills date to mid-August (a V buy and a UNH sell on the 13th, an MA buy and an AAPL sell on the 12th), with the prior cluster back in early June. That cadence is consistent with a breakout system: long stretches of inactivity punctuated by bursts when the tape breaks out.
Backtest and validation
Over 451 days the backtest returned 6.95% (final equity $10,695), a 3.83% CAGR, with a Sharpe of just 0.34 and a maximum drawdown of 21.73%. Win rate was 38.46% across 108 trades — low, but expected for trend following, where a minority of winners must carry the book. Turnover of 2,082% is the number that should give pause: this strategy churns hard, and while modelled fees were only $108, real-world slippage on that much trading is a genuine drag.
Walk-forward validation is where the picture sharpens. Three of four folds were positive, and folds 3 and 4 were excellent (14.09% at Sharpe 2.72, then 11.35% at Sharpe 1.86). But fold 2 lost 7.46% at a Sharpe of -1.07 — the strategy simply doesn't work in choppy, trendless regimes. Crucially, the agent did not pass our gate. The deflated Sharpe ratio (DSR) of 0.198 is weak once you account for the 6 trials searched, even though the probabilistic Sharpe (PSR) of 0.674 looks respectable in isolation. In plain terms: the recent out-of-sample strength is real but not yet distinguishable from luck.
Verdict
Donchian breakout earns its keep in trending markets and behaves exactly as the literature predicts — including the ugly parts. The strengths are a coherent, non-overfit thesis and strong recent folds. The risks are equally clear: a sub-0.5 full-sample Sharpe, a 22% drawdown, punishing turnover, and a validation gate it hasn't cleared. We're keeping it live for observation, not conviction, and watching whether folds 3 and 4 were the signal or the noise.