We backtest to Jan 1 2006, and we’ve actually tested the engine back to Jan 1 2000 using proxies (since few ETFs were available then). Was helpful to see that the basic techniques worked well during the dot-com bust.
There were meaningful momentum drawdowns in 2009/2010, Summer of 2006, 4Q2007 (quant crash), and y2k burst. Looks like your models did fine during these periods.
Thanks for noticing! Most of those are in our window (2006 forward), and it's mechanical, not luck. The model isn't one fixed rule in all weather, it's regime-adaptive. In a normal uptrend it ranks on ~12-month momentum and won't hedge proactively. Once the market has spent long enough below its long-term trend, the model restricts its own memory to recent strength, so it isn't ranking on stale pre-crash highs, which allows us to catch the new trend quicker. Similarly, a full defensive posture unlocks only as a real bear market deepens.
It’s worth noting that the ETF menu was more limited during the early years, so potentially the performance would have been *better* during that period if the current menu was available.
This is very interesting. Up to what time period does the backtest run to? Thank you.
We backtest to Jan 1 2006, and we’ve actually tested the engine back to Jan 1 2000 using proxies (since few ETFs were available then). Was helpful to see that the basic techniques worked well during the dot-com bust.
There were meaningful momentum drawdowns in 2009/2010, Summer of 2006, 4Q2007 (quant crash), and y2k burst. Looks like your models did fine during these periods.
Thanks for noticing! Most of those are in our window (2006 forward), and it's mechanical, not luck. The model isn't one fixed rule in all weather, it's regime-adaptive. In a normal uptrend it ranks on ~12-month momentum and won't hedge proactively. Once the market has spent long enough below its long-term trend, the model restricts its own memory to recent strength, so it isn't ranking on stale pre-crash highs, which allows us to catch the new trend quicker. Similarly, a full defensive posture unlocks only as a real bear market deepens.
It’s worth noting that the ETF menu was more limited during the early years, so potentially the performance would have been *better* during that period if the current menu was available.