We backtested 80+ tactical allocation strategies over 30 years, here’s what actually holds up

So about a year ago a few of us went down a massive rabbit hole trying to figure out if tactical asset allocation actually works or if its just market timing with extra steps. We come from quant finance and software engineering backgrounds, and we started implementing strategies from academic papers. Meb Faber's trend following stuff, Wouter Keller's work, a bunch of others. We weren't planning to do 80+ of them but every paper references three more and you know how that goes.

After running all of them through 30 years of data, a few things stood out that we really didn't expect.

First, the boring stuff works best. The single best predictor across almost every strategy is some version of the 10-month simple moving average. Faber published this in 2006 and its still the foundation of like half the strategies out there. Asset above its 10mo SMA, hold it. Below, move to cash or bonds. A lot of the more complex approaches are basically just dressing this up with extra steps and more parameters to overfit.

Second, TAA doesnt really beat buy and hold on raw returns. The median strategy we tested returned about 9.2% CAGR vs 8.5% for a plain 60/40. Not exactly life changing. But the max drawdown was -16% vs -34%. Thats the real argument for TAA, not making more money but not panicking and selling at the bottom because your portfolio just dropped 30%.

The thing is no single strategy works all the time. Momentum strategies crush it in trending markets but get destroyed in choppy sideways stuff (2015, parts of 2018). Defensive strategies protect you in crashes but drag in bull runs. The only thing that actually smoothed the equity curve was blending 4-6 strategies with different approaches together. Which makes sense but we didn't appreciate how much of a difference it makes until we saw the numbers.

Walk-forward testing is humbling btw. A lot of these strategies look incredible when you backtest them in-sample. The moment you do proper out-of-sample validation, sharpe ratios drop 20-40%. If you're looking at TAA backtests and nobody mentions whether they were tested out-of-sample, you're probably looking at overfitted results.

One thing that surprised us is signal timing. A signal dated "Feb 28" could mean "this is what you should have held during February" or "this is what you should hold starting in March." Getting it wrong means you're trading a month early or late, and on momentum strategies thats like 1-2% annually. We've seen some well known platforms get this wrong which is kind of scary.

Also if you're a European investor, good luck. Almost everything assumes you're trading SPY, TLT, GLD. UCITS alternatives exist for nearly all of it but nobody bothers documenting the mapping. This was honestly one of the things that frustrated us enough to build our own tool.

We ended up building our own tool to track all of this because nothing out there did what we needed. Happy to share if anyone's interested but didnt want this to turn into a promo post.

Curious if anyone else here runs any kind of tactical approach or if its mostly buy and hold in this sub.