The small-cap test
The capacity argument said a small account's structural edge lives where institutions cannot trade. Every failure above happened in 112 mega-caps, so that argument was never actually tested. 109 small/mid caps, hand-picked and therefore maximally survivorship-biased, at 23x lower dollar volume (median $39.8M/day against $919M).
| momentum | equal-weight | index | edge vs EW | maxDD | Sharpe | |
|---|---|---|---|---|---|---|
| large cap (111) | +361.7% | +324.3% | SPY +262.7% | +37.4% | 27.5% | 1.04 |
| small cap (109) | +374.1% | +360.9% | IWM +134.4% | +13.3% | 44.5% | 0.82 |
The edge is smaller, not larger. Drawdown is far worse and Sharpe is lower.
And the survivorship contamination is worse too: equal-weighting my hand-picked small caps returned +360.9% against IWM's +134.4% — a +226 point construction premium, against +150 for the large-cap set. Hand-recalling small caps selects survivors hard, and 6 of 116 tickers had already delisted, roughly 3x the large-cap attrition.
Why the capacity argument fails
Breakeven half-spread against equal-weighting the same universe:
| breakeven | measured | |
|---|---|---|
| large cap | 6.76 bps | 2.40 bps — clears it |
| small cap | 4.08 bps | 14.27 bps — 3.5x over |
Measured 2026-09-09 on the live feed. Quoted half-spread by universe:
| median | p25 | p75 | p90 | max | |
|---|---|---|---|---|---|
| large cap | 1.91 | 1.12 | 3.12 | 4.50 | 9.16 |
| small cap | 10.02 | 4.88 | 18.79 | 29.56 | 53.49 (ATNI) |
Trade-weighted for what momentum actually buys: 14.27 bps — worse than the universe median, because momentum's picks skew toward the thinner, more volatile names. Against a 4.08 bps breakeven, the +13.3% edge becomes −73.5%.
The argument was self-defeating and it took running it to see why. Institutions avoid these names because the trading costs are prohibitive. Those same costs apply to a $100k account running 26x annual turnover. The friction that protects an anomaly from arbitrage also consumes it for you — you do not get to collect an edge that exists only because collecting it is too expensive.
A low-turnover strategy might reach it. A weekly rebalance cannot.