Performance
Backtested across 16 stress episodes, 6 regions, 28 years of historical data (1998–2024). All figures are net of transaction costs and opportunity cost deductions.
Developed 2026. Live signal generation began April 2026; the formal prospective validation window opens after the current re-testing program concludes. Past performance does not guarantee future results.
6.0:1
Net preservation ratio
15.9%
Avg. capital preserved per episode
16
Stress episodes covered
28 yrs
Historical data, 6 regions
Episode-by-episode results
Benchmark is a 60/40 equity/bond portfolio using regional equity indices and sovereign bonds, rebalanced monthly.
| Episode | Region | Period | 60/40 | RegimeR | Net preserved | Ratio |
|---|---|---|---|---|---|---|
| US GFC | US | 2007–2009 | -22.6% | +16.1% | +35.8pp | 13.5:1 |
| AU GFC | AU | 2008–2009 | -25.2% | +12.1% | +31.4pp | 6.4:1 |
| DE GFC | DE | 2008–2009 | -31.8% | +7.0% | +35.5pp | 11.6:1 |
| GB GFC | GB | 2007–2009 | -29.1% | +10.4% | +36.6pp | 13.3:1 |
| DE Eurozone | DE | 2011–2012 | +3.4% | +23.0% | +7.3pp | 1.6:1 |
| US COVID | US | 2020 | +2.8% | +8.0% | +1.2pp | 1.3:1 |
| AU COVID | AU | 2020 | -5.5% | +6.4% | +8.9pp | 4.0:1 |
| DE COVID | DE | 2020 | -2.5% | +6.7% | +6.1pp | 3.0:1 |
| GB COVID | GB | 2020 | -4.5% | +8.1% | +11.0pp | 7.8:1 |
| US 2022 | US | 2022 | -24.2% | -15.6% | +6.0pp | 3.3:1 |
| AU 2022 | AU | 2022 | -20.3% | -13.7% | +3.3pp | 2.0:1 |
| DE 2022 | DE | 2022 | -31.6% | -18.4% | +10.0pp | 4.1:1 |
| GB Gilt Crisis | GB | 2022 | -21.0% | -5.3% | +13.1pp | 6.3:1 |
| CA GFC | CA | 2008 | -22.5% | +1.4% | +23.9pp | 8.2:1 |
| CA COVID | CA | 2020 | -5.4% | +3.0% | +8.4pp | 4.1:1 |
| CA Oil Rout | CA | 2014–2016 | -2.6% | -4.0% | +-1.4pp | 0.0:1 |
14.3%
US avg (3 episodes)
14.5%
AU avg (3 episodes)
14.7%
DE avg (4 episodes)
20.2%
GB avg (3 episodes)
Robustness
Scaling
Tested from $100M to $10B AUM. At $10B the system retains over 98% of the preservation benefit in major episodes. Regime switches are monthly and use liquid instruments, limiting market impact.
Sensitivity
Statistical validation
Withdrawn, honestly. Our original permutation tests shuffled individual months, which overstates significance for trending market data. When we re-tested with block-based methods that preserve market trends, the regime boundaries were not statistically distinguishable from chance placements. We are re-testing under a stricter pre-registered protocol rather than publishing numbers we can no longer stand behind.
We publish this finding because an audit-grade record that only reports favourable results is not audit-grade. The episode table above is a descriptive backtest; whether the classification skill behind it exceeds chance is exactly what the re-testing program and the prospective record are designed to establish.
Recovery compounding
Capital preserved during a drawdown compounds throughout the subsequent recovery. This second-order effect added 61% on top of direct preservation across the 16-episode sample.
$268M
Direct preservation per $100M
$163M
Recovery compounding
−$46M
Recovery basket drag
Compound sequential by region
Episodes chained chronologically within each region. The ending balance after each crisis-and-recovery cycle becomes the starting capital for the next.
| Region | Episodes | 60/40 terminal | RegimeR terminal | Advantage |
|---|---|---|---|---|
| US | GFC → COVID → 2022 | $143.6M | $265.8M | +85% |
| AU | GFC → COVID → 2022 | $133.0M | $250.4M | +88% |
| DE | GFC → Euro → COVID → 2022 | $183.2M | $465.9M | +154% |
| GB | GFC → COVID → Gilt | $142.1M | $243.5M | +71% |
| Average (per $100M initial) | $150.5M | $306.4M | +104% | |
Methodology notes
Benchmark. 60/40 equity/bond portfolio using regional indices, rebalanced monthly. This is a standard institutional proxy — actual allocations vary by mandate. RegimeR provides the regime classification signal; clients define their own defensive baskets. The backtested results use example baskets to demonstrate signal value. RegimeR applies to the liquid portion of the portfolio.
Regime assignment. Each month is assigned its classified regime using walk-forward percentile thresholds on named macroeconomic indicators. No future data is used in any classification decision. The regime-specific basket for that month determines portfolio weights.
Recovery compounding. The preserved capital gap at the crisis trough is multiplied by the benchmark's recovery-period return to isolate the pure compounding advantage. Recovery basket alpha (the difference between RegimeR's expansion-regime basket and the benchmark during recovery) is tracked separately and is net negative overall (−$46M per $100M), reducing the headline figure. The $268M and $385M figures are simple sums across episodes spanning 2007–2024, not present-value discounted.
Compound sequential. Episodes within each region are chained chronologically: the ending balance after one crisis-and-recovery cycle becomes starting capital for the next. Both crisis and recovery returns are applied. Between-episode periods assume capital remains invested at the prevailing regime allocation.
Limitations. In three of four 2022 episodes, the 60/40 benchmark had not fully recovered to its pre-crisis peak by the data cutoff. All returns are gross of management fees. Returns are computed on monthly adjusted close prices. Past performance is backtested and does not guarantee future results. Live signal generation began April 2026; the formal prospective validation window opens after the current re-testing program concludes.