Window: 2006-08-14 .. 2026-08-10 (5017 days) | universe: top-100 as of 2007-06-01 | vol: trailing 60d realized | rebalance: 21d | gross returns
| Quintile | Ret | Sharpe | MaxDD | PF |
|---|---|---|---|---|
| SPY buy-and-hold | +496.62% | 0.67 | -56.47% | 1.12 |
| Q1 (low) | +191.74% | 0.63 | -25.74% | 1.10 |
| Q2 | +218.83% | 0.58 | -34.97% | 1.10 |
| Q3 | +173.84% | 0.47 | -40.77% | 1.08 |
| Q4 | +81.07% | 0.31 | -49.68% | 1.05 |
| Q5 (high) | -61.24% | -0.06 | -84.19% | 0.99 |
| Q1-Q5 spread | +145.82% | 0.35 | -365.60% |
IS/OOS split at 2016-01-01
| Variant | Segment | Ret | Sharpe | MaxDD | PF | |---|---|---|---|---|---|---| | Q1 | IS SPY | +60.39% | 0.41 | -56.47% | 1.07 | | Q1 | IS Q1 | +81.28% | 0.70 | -25.74% | 1.12 | | Q1 | OOS SPY | +271.98% | 0.95 | -34.10% | 1.16 | | Q1 | OOS Q1 | +60.93% | 0.57 | -16.92% | 1.09 | | Q5 | IS SPY | +60.39% | 0.41 | -56.47% | 1.07 | | Q5 | IS Q5 | -78.65% | -0.48 | -82.09% | 0.92 | | Q5 | OOS SPY | +271.98% | 0.95 | -34.10% | 1.16 | | Q5 | OOS Q5 | +81.51% | 0.44 | -48.09% | 1.07 |
Q1 vs SPY: OOS dret -94.52%, bootstrap p=0.982
Decision memo
Low-volatility anomaly: classic factor finding. Tests whether it holds in our PIT universe.
- Adopt if: Q1 Sharpe > SPY, spread is significant.
- Otherwise: reject and ledger.
VERDICT (2026-08-25): REJECT -- anomaly real but fails OOS
The low-volatility anomaly is confirmed in the full window:
- Perfect monotonicity: Q1(+192%/Sharpe 0.63) > Q2(+219%/0.58) > Q3(+174%/0.47) > Q4(+81%/0.31) > Q5(-61%/-0.06)
- Q1-Q5 spread: +146%, Sharpe 0.35
But OOS (2016-2026) kills it:
- Q1 OOS: +61% vs SPY +272% (Sharpe 0.57 vs 0.95)
- Q5 OSS: +82% vs SPY +272% (recovered from GFC losses)
- Q1 vs SPY OOS: -95%, bootstrap p=0.982
The anomaly works in crash protection (Q5 destroyed in GFC) but underperforms in the bull-market OOS period. Not a viable standalone strategy. Could be useful as a DEFENSIVE sleeve in bear regimes, but that's a different deployment than a standalone factor strategy.