Open: diversifying ideas that use the economic and physical-economy data creatively, and that could add value alongside 60/40 and trend following
Conclusion
Recommended portfolio
No new strategy is admitted this cycle. Two of the three got 'needs_work' and one got 'reject'. All three combinations were tested on train-only inverse-vol weights.
(1) 60/40 alone. Train: Sharpe 0.55, max drawdown -32.5%. Validation: Sharpe 1.09, CAGR 11.9%, max drawdown -18.9%.
(2) 60/40 + Trend (40.8/59.2). Train: Sharpe 0.81, max drawdown -15.3%. Validation: Sharpe 0.87, CAGR 7.4%, vol 7.5%, max drawdown -14.0%. Correlation between the two is 0.49. This blend gives up about 0.2 validation Sharpe compared with 60/40 in its best era. In exchange it roughly halves the GFC drawdown (-15% vs -33% in train) and cuts the COVID drawdown by 5 points. That crisis behaviour is why it stays as the core.
(3) 60/40 + macro_quadrant (42.9/57.1). Validation: Sharpe 1.40, CAGR 10.7%, vol 6.7%, max drawdown -8.7%, correlation 0.22. This is the best-looking result, but I am not admitting it, for four reasons:
- A 1.40 Sharpe is above the lab's caution line.
- The macro_quadrant input (validation Sharpe 1.11) is the default run. Its honest walk-forward OOS Sharpe is 0.74, with a -13.6% drawdown and 991 days underwater.
- Most of its validation return comes from 2019-2021: +12.7%, +20.3%, +15.0%, against 2-5% a year in 2016-18. Average gross exposure is 1.43, mostly SHY and TIP, so part of the gain is leveraged duration in a bond bull market.
- The family has 14 trials, and the deflated Sharpe has not been recomputed for that.
(4) 60/40 + Trend + macro_quadrant. Validation: Sharpe 1.08, max drawdown -9.9%. Train: Sharpe 0.78, max drawdown -10.1%. In validation macro_quadrant correlates 0.66 with Trend, so it overlaps heavily with the trend sleeve and adds little beyond it.
labor_turn_duration is rejected. A constant 0.6-duration basket beats it on train, 0.61 vs 0.33. fear_vs_fundamentals is not admitted: its validation return is essentially 2019 alone, and it loses in the scares it is designed to fade (-13.2% in 2015-16). Neither was blended, to avoid spending trials on them.
Decision: keep 60/40 + Trend as the core. macro_quadrant is the leading candidate. If it clears the Skeptic's checks, it should enter as a capped sleeve of at most 25% risk, not the 57% that inverse-vol would give it.
Results
Validation-period results for each idea, with the Skeptic's verdict.
| Strategy | Verdict | Val. Sharpe | Val. CAGR | Val. max DD | Corr. 60/40 | DSR | 2007–21 |
|---|---|---|---|---|---|---|---|
| macro quadrant Growth x inflation quadrant allocation from point-in-time hard data and breakevens | Needs work | 1.11 | 9.5% | -9.6% | 0.22 | 100% | |
| labor turn duration Labor-turn duration timing: buy Treasuries when the labor market cracks, because the Fed cuts more than markets price | Rejected | 0.38 | 3.4% | -16.1% | -0.21 | 82% | |
| fear vs fundamentals Fade market growth scares that the physical economy does not confirm | Needs work | 0.47 | 3.3% | -9.3% | 0.35 | 87% |
Portfolio tests
Combinations the Constructor tried. Weights fit on train; ★ marks the recommendation.
| Combination | Train Sharpe | Val. Sharpe | Val. CAGR | Val. max DD |
|---|---|---|---|---|
| 60/40 stocks/bonds 100% | 0.55 | 1.09 | 11.9% | -18.9% |
| ★ 60/40 stocks/bonds 41% + Trend following 59% | 0.81 | 0.87 | 7.4% | -14.0% |
| 60/40 stocks/bonds 43% + macro quadrant 57% | 0.62 | 1.40 | 10.7% | -8.7% |
| 60/40 stocks/bonds 26% + Trend following 38% + macro quadrant 35% | 0.78 | 1.08 | 8.2% | -9.9% |
The highest validation Sharpe isn't automatically chosen: the Constructor only admits strategies the Skeptic has cleared.
The ideas
Growth x inflation quadrant allocation from point-in-time hard data and breakevens
Each asset class does best in a different regime.
RejectedLabor-turn duration timing: buy Treasuries when the labor market cracks, because the Fed cuts more than markets price
Labor-market turns are non-linear.
Needs workFade market growth scares that the physical economy does not confirm
Markets regularly price recession probabilities well above what the real economy later delivers.
Next research
Highest priority: run the Skeptic's checks on macro_quadrant, using one pre-registered trial each.
(a) Rerun with max_gross 1.0 and a version with no SHY in G-/I+, to measure how much comes from leveraged duration.
(b) Regress it on RiskParity and long-only Trend. Admit it only if the residual Sharpe is at least 0.3.
(c) Report Sharpe excluding 2020 and for 2016-18 alone, plus time share and P&L by quadrant.
(d) Fix the z-window at 1260 days and recompute the deflated Sharpe at N=14.
If it passes, re-blend at a 25% risk cap and paper-trade it in Amber Terminal. Pre-agreed shutdown rule: stop at -12% drawdown or a negative 18-month Sharpe.
Data to acquire:
(1) ALFRED vintages for ICSA, INDPRO, TOTALSA, HOUST and UNRATE. All three new ideas rely on revised growth data, and this is the main remaining leakage-of-hindsight risk.
(2) A longer-history inflation input (breakevens or CPI surprises, plus commodity futures history before 2007). The inflation-up quadrants have almost no sample in 2007-2021, and the 2022-style inflation shock is the untested failure mode for both 60/40 and macro_quadrant.
(3) CFTC positioning (Milestone 3). It is a candidate for a trend/carry signal with low correlation to the current trend sleeve, which overlaps 0.66 with macro_quadrant.
Also fix the UNRATE de-duplication bug in _releases before any labor-based idea is reused.