In one line: it trails the index on raw
return, but draws down about half as much at ~42% average exposure — and its stock selection beats a
random book net of costs, most clearly out-of-sample.
CAGR, net
3.6%/yr
$100k book, 2% on idle cash
CAGR, gross
8.0%/yr
no costs
Max drawdown
−22.8%
index ~−55% over the same window
Avg exposure
~42%
in cash the rest of the time
| Kind of system | Mean-reversion stock-picking — buys a dividend-quality name after a sharp short-term sell-off, holds a few days, then returns to cash. Episodic, not always-on. |
|---|---|
| Universe | S&P 500 Dividend Aristocrats (25+ years of consecutive dividend growth), point-in-time membership including names later dropped. |
| Average period | ~4 trading days per trade — invested on roughly 40% of days, in cash the rest. Sell into the first close above the prior day’s high, or after about a week. |
| Concurrent symbols | Up to 10 slots, sized by ATR so each position carries a similar volatility exposure. |
| Return | CAGR 3.6% / yr net ($100k book, 2% earned on idle cash) — modest because the book is in cash most of the time and the per-trade edge is thin. 8.0% / yr gross before costs. The S&P 500 returned 10.8% / yr total-return over the same window; on raw return the system trails. Friction is 5 bps slippage per side plus a per-share commission. Full tables on the Deep Dive. |
| Exposure | ~42% of capital deployed on average, in cash the rest — so a raw-return comparison to a 100%-invested index is not like-for-like. |
| Distinguishing feature | A lower-drawdown profile — a worst loss of −22.8% against the index’s ~−55% in 2008–09; it gained 2.5% in 2008 and fell less than the index in 2022. Its stock selection beats a random book on the same universe and exits, net of costs (profit factor 1.07 vs 0.99), most clearly out-of-sample. |
| Who it’s for | A portfolio manager looking for a defensive, dividend-quality sleeve — valued for a shallower drawdown and the shape of its returns, not as a standalone engine meant to out-compound the index. |
| Out-of-sample | No parameters were fit (the rules are taken as supplied), so a 2005–17 / 2018–26 temporal split is a robustness check: the edge over a random book held on the later data (profit factor 1.10 vs 0.97 out-of-sample). A live walk-forward is a next test, not yet run. |
→ Details for how to read a part-time book against a full-time index, the skill-vs-luck test, and how it was stress-tested — or the full Deep Dive with methodology, the random baselines, the calendar-year record, and the entry-timing study.
See how it pairs with your book. Its low exposure and defensive
drawdown make it a natural sleeve — we can re-run it on your universe, or send you the Python, daily
trade logs, and data behind every number.
· start a conversation →