The three live momentum sleeves — USA (top 7 large caps), Canada (top 10 TSX) and Germany (top 10 XETRA) — combined. 2000-01 to 2026-04, 316 months. Rebuilt 2026-09-20 from the series each strategy publishes, after the previous version was found to be running a Canada series contaminated by a look-ahead in the fundamental filter, and over a window starting in 2006 rather than 2000. Every sleeve's own caveats carry through: survivorship windows, capacity limits, in-sample config selection.
| Sleeve | CAGR | Vol | Sharpe | Sortino | MaxDD | Worst month |
|---|---|---|---|---|---|---|
| USA | +25.3% | 26.4% | 0.85 | 1.80 | -35.5% | -22.9% |
| Canada | +34.8% | 25.9% | 1.23 | 2.85 | -26.0% | -18.9% |
| Germany | +20.6% | 22.1% | 0.79 | 1.55 | -39.1% | -16.9% |
Canada carries the book and Germany is the weakest leg on every measure. That is the starting point, not the conclusion — what follows is about whether the mix beats its parts.
| USA | Canada | Germany | |
|---|---|---|---|
| USA | 1.00 | 0.30 | 0.29 |
| Canada | 0.30 | 1.00 | 0.26 |
| Germany | 0.29 | 0.26 | 1.00 |
And in the worst fifth of months for the portfolio:
| USA | Canada | Germany | |
|---|---|---|---|
| USA | 1.00 | -0.22 | -0.13 |
| Canada | -0.22 | 1.00 | -0.28 |
| Germany | -0.13 | -0.28 | 1.00 |
The pairwise average falls from 0.29 to -0.21 in the tail — the sleeves do not sink together on a bad month. That is the whole reason the combination exists, and it is also the claim to distrust first: three long-equity momentum books will correlate in a genuine systemic crash however they behaved in this sample.
| USA/CAN/GER | CAGR | Sharpe | MaxDD |
|---|---|---|---|
| 25/50/25 | +30.4% | 1.459 | -20.3% |
| 20/50/30 | +30.1% | 1.458 | -21.2% |
| 25/45/30 | +29.7% | 1.457 | -20.6% |
| 20/55/25 | +30.8% | 1.455 | -20.9% |
| 30/45/25 | +29.9% | 1.453 | -19.7% |
| 25/55/20 | +31.0% | 1.452 | -20.1% |
| 30/50/20 | +30.6% | 1.451 | -19.5% |
| 20/45/35 | +29.4% | 1.451 | -21.4% |
| 15/55/30 | +30.5% | 1.447 | -21.8% |
| 15/50/35 | +29.8% | 1.446 | -22.0% |
| 30/40/30 | +29.3% | 1.445 | -20.0% |
| 25/40/35 | +29.0% | 1.445 | -20.8% |
| 20/60/20 | +31.4% | 1.444 | -20.7% |
| 30/55/15 | +31.3% | 1.441 | -19.2% |
Top of 231 combinations on a 5% grid. The best reads Sharpe 1.459 at 25/50/25, but the standard error of a Sharpe estimated from 316 months is 0.081 — so 65 of 231 combinations are statistically tied with it. Choosing between them on this evidence is choosing noise.
| USA/CAN/GER | CAGR | Vol | Sharpe | MaxDD | |
|---|---|---|---|---|---|
| equal | 33/33/34 | +28.4% | 18.0% | 1.41 | -20.0% |
| 35/40/25 | 35/40/25 | +29.5% | 18.5% | 1.44 | -19.5% |
| best | 25/50/25 | +30.4% | 18.8% | 1.46 | -20.3% |
| inv-vol | 31/32/37 | +28.2% | 17.8% | 1.41 | -20.3% |
All four sit inside the tie. Every one roughly halves the drawdown of the best single sleeve while landing between its CAGR and the weakest sleeve's.
Re-optimising the weights every January on the previous 60 months and holding them for the next twelve, repeated across the sample — 252 months out of sample:
| Rule | CAGR | Sharpe | MaxDD |
|---|---|---|---|
| otimizado | +28.7% | 1.30 | -23.6% |
| equal | +27.2% | 1.48 | -19.9% |
| 35/40/25 | +28.1% | 1.49 | -19.1% |
Re-optimising loses. It gives up 0.18 of Sharpe against simply holding equal weights and digs a drawdown -4.5 points deeper than the fixed 35/40/25. The optimiser chases whatever did well in the training window and arrives after it has stopped working. A fixed 35/40/25 beats both on every column.
35/40/25 — USA, Canada, Germany. It is not the in-sample peak (25/50/25 is, by 0.024 of Sharpe, a quarter of one standard error). It is the weight that wins the only test that looks forward: Sharpe 1.49 against 1.30 for the optimiser and 1.48 for equal weights, with the shallowest drawdown of the three.
In-sample it delivers CAGR +29.5%, Sharpe 1.44, MaxDD -19.5% — against -26.0% for the best sleeve standing alone. Roughly two thirds of the return of the strongest leg, at half its drawdown.
Two things this does not say. It does not say 35/40/25 is better than 30/40/30 or 35/45/20 — those are inside the tie and the data cannot separate them. And the drawdowns here are measured on month-end values: the intra-month low is worse, by a factor of roughly 1.5 on the one sleeve where daily data exists to check it.
| Year | USA | Canada | Germany | Portfolio |
|---|---|---|---|---|
| 2000 | +16.9% | +45.4% | +21.1% | +32.5% |
| 2001 | +0.0% | -6.3% | -2.0% | -3.0% |
| 2002 | +0.0% | +51.1% | -0.1% | +18.6% |
| 2003 | +33.0% | +143.0% | +58.3% | +79.2% |
| 2004 | +14.4% | +35.3% | +33.3% | +28.9% |
| 2005 | +32.7% | +29.2% | +45.7% | +35.6% |
| 2006 | -8.5% | +66.8% | +42.3% | +32.6% |
| 2007 | +81.7% | +40.7% | -1.3% | +42.2% |
| 2008 | -13.2% | -9.5% | -18.2% | -12.6% |
| 2009 | +14.8% | +26.0% | +34.3% | +24.7% |
| 2010 | +20.3% | +73.6% | +37.0% | +45.3% |
| 2011 | -3.9% | +1.9% | -16.8% | -4.5% |
| 2012 | +29.1% | +27.9% | +36.8% | +31.0% |
| 2013 | +46.4% | +71.2% | +68.0% | +62.3% |
| 2014 | +23.1% | +5.8% | +2.3% | +11.7% |
| 2015 | -1.0% | +14.9% | +46.9% | +16.5% |
| 2016 | +13.9% | +47.2% | +2.1% | +24.3% |
| 2017 | +43.0% | +91.0% | +69.7% | +71.0% |
| 2018 | -9.7% | +21.7% | +8.2% | +8.5% |
| 2019 | +4.7% | +9.9% | +47.1% | +17.0% |
| 2020 | +69.8% | +51.4% | +18.7% | +50.7% |
| 2021 | +26.4% | +28.8% | +12.6% | +26.7% |
| 2022 | +11.0% | +25.9% | -7.9% | +13.4% |
| 2023 | +20.0% | -14.2% | -16.2% | -3.1% |
| 2024 | +149.8% | +69.2% | +37.8% | +87.5% |
| 2025 | +51.9% | +68.0% | +48.4% | +60.2% |
| 2026 | +120.0% | +5.4% | +7.3% | +41.4% |
23 of 27 years positive.
| Month | USA | Canada | Germany | Portfolio | Sleeves down |
|---|---|---|---|---|---|
| 2004-04 | -13.1% | -14.1% | -3.2% | -11.0% | 3/3 |
| 2008-01 | -15.4% | -0.1% | -16.9% | -9.6% | 3/3 |
| 2026-03 | -8.0% | -18.9% | +3.5% | -9.5% | 2/3 |
| 2000-03 | -17.6% | +1.2% | -13.4% | -9.0% | 2/3 |
| 2000-05 | -13.7% | -1.5% | -13.8% | -8.8% | 3/3 |
| 2018-10 | -22.9% | -0.0% | -0.0% | -8.1% | 3/3 |
| 2019-09 | -11.5% | -11.1% | +2.0% | -8.0% | 2/3 |
| 2020-02 | -6.5% | -11.4% | -3.8% | -7.8% | 3/3 |
Read the last column. Most bad months are one sleeve having a bad time; the months where all three fall together are the ones that set the drawdown.