A decorrelated trend core is mathematically off the table
The idea that we can reduce system drawdown by grouping currency pairs with low correlation is fundamentally flawed because the pairs that actually…
The idea that we can reduce system drawdown by grouping currency pairs with low correlation is fundamentally flawed because the pairs that actually trend are inherently correlated. I spent time testing a hypothesis that I could lower the total drawdown (the peak-to-trough decline in account value) by building a core portfolio based on low inter-pair correlation rather than just individual pair strength. I calculated the daily correlation and the trend-following edge for 19 currency pairs and gold, using a breakout strategy on higher timeframes. The results, however, pointed to a structural reality that contradicts the diversification dream.
| Currency Pair | Profit Factor (PF) |
|---|---|
| USDJPY | 1.56 |
| XAUUSD (Gold) | 1.34 |
| GBPJPY | 1.25 |
| EURJPY | 1.21 |
| AUDJPY | 1.19 |
| CHFJPY | 1.15 |
| NZDJPY | 1.08 |
| CADJPY | 1.01 |
| A profit factor (PF) is simply the ratio of gross profit to gross loss. A value greater than 1 means the system is profitable. As shown above, only 8 of the 19 assets I tested (all of which are either Yen crosses or gold) showed a positive trend-following edge. Pairs like EURUSD or GBPUSD simply did not exhibit a trend edge. In other words, the assets that actually trend are already a giant, correlated cluster. There is no such thing as an uncorrelated trending pair in the FX market. | |
| When I tried to force a portfolio with minimized correlation, the system performed worse. By swapping a pair like EURJPY for AUDJPY to lower the internal correlation from 0.21 to 0.16, the system saw its drawdown worsen from -21.7% to -24.8%. The return-to-drawdown ratio dropped from 8.85 to 7.63. The reason is simple: the individual drawdown of the new pair was worse than the minor benefit gained from the reduced correlation. My existing “robust5” selection, which chooses pairs based on their individual edge, is already effectively optimized. | |
| This leads to a firm conclusion regarding my research track: you cannot reduce system-wide drawdown through diversification within the FX market because the necessary ingredients (uncorrelated trending pairs) do not exist. | |
| The only real path to diversification is moving into entirely different asset classes. I have already integrated stock indices into the system, which hold a correlation of 0.16 against the FX core. Since the path of rebuilding the core based on low-correlation FX pairs is closed, I am marking version 1.4.1 as the current frontier. Further drawdown reduction will have to come from outside the FX and gold universe. |