I’m a bootstrapped tech entrepreneur by background, so my mindset is simple: try a lot of things, keep what works, and don’t go broke in the process.
When I moved away from discretionary trading, I spent time on Allocate Smartly. Some of the models resonated with me, but most felt too muted. At P123, I found the opposite problem — much more powerful models, but many with volatility and drawdowns larger than I’m comfortable with.
That pushed me to explore:
-
Blending lower-correlation, more stable strategies with higher-return ones
-
Hedging with SPX puts
-
TZA pairings
-
Timing sleeves
Those timing sleeves are largely inspired by the risk-off logic found in tactical approaches like those on Allocate Smartly.
Are they overfit? Maybe.
Will they work as well going forward? Nobody knows.
I recently watched Jim O’Shaughnessy and Cliff Asness discuss how new factor ideas need two things: supportive data and some intuitive logic. That framing stuck with me. Interestingly (and related to some other discussions here), they noted that with machine learning, the balance may shift more toward trusting data even when intuition is harder to discern.
Do these timing overlays make sense as a way to reduce exposure during poor regimes? Does the data look reasonable? That’s about all I can ask myself before giving it a shot.
I’m not betting the farm on them — I’m combining several different timing mechanisms as a composite and applying that composite only to a portion of the portfolio. Even if the outcome is neutral, trimming exposure to a strong model by ~20% across CAGR, risk, and drawdown isn’t necessarily a bad outcome. That said, I obviously hope for — and expect — much better than “success-neutral.”
Using multiple variations of these overlays can provide diversification across how and when exposure gets reduced, rather than relying on any single timing signal.
I think there’s meaningful potential in cross-hybridizing models with tactical asset allocation approaches. The same may be true of thoughtfully blending in certain discretionary techniques, which is why the recent discussion around loss truncation caught my attention so strongly.