It’s safe to say that in 2014, between $4T and $5T will have been traded each day in the global FX markets, based on the BIS’ Triennial survey from April 2013. Volume is lower than in previous years, but institutional electronic platforms like FXall are still reporting daily flows north of $100B.
While most trading strategies in other asset classes have limits in terms of liquidity and most, if not all, regulated exchanges impose position limits, this model is scalable and can be executed with no significant constraints…as long as the trade size is less than $10M at a time, that is. (The upshot is that this model scales in anyway, so achieving an initial price point with $100M or more over the course of a few minutes is no problem.)
The Model in a Nutshell
Strategy. Over long time horizons, markets are driven by cyclic, structural and risk events which produce fairly stable statistical patterns in price. Reversion to mean (measured on a logarithmic scale) is almost a certainty. Scaling into a position over long excursions, commonly confused by most retail traders as “adding to a loser”, makes use of high probability reversions while at the same time providing the added benefit of “value discovery” (my term). At key price levels and in alignment with events, initiating a trade idea may require more than one position, each with several trades. Although this strategy isn’t a precise science, all inputs and decision triggers are precisely defined, statistically driven, and consistently applied.
Execution. Trading frequently across a broad, actively balanced basket of currency pairs, in very small sizes, greatly increases the model’s edge. More volatile currency pairs with larger interest rate differentials are favored. Optimal position size is between 1/50,000 and 1/40,000 of account balance, depending on volatility. This small size safely allows scaling in over longer time horizons, until the instrument eventually reverts to mean (relative to cost basis), and/or long enough to recover losses with profitable trades. Although most trades are usually open for several weeks, the time horizon for scaling in during mean reversion is usually measured in months. It is possible that one could carry a losing position for years, but in most cases profitable trades will have allowed for the recovery of losses much sooner. In many cases, losing trades will be earning yield as well.
Account Management. The optimal commitment of available margin is 17-18%, which is equivalent to carrying about 25 trades (depending on the pairs selected and their established margin requirements). Margin in the basket can be as much as 30% of account balance and still allow room to maneuver comfortably. 50% is probably too high, but the model is still healthy if the basket correlation value is 0.65 or less.
(A note about correlations: Most currency pairs present different figures for different time horizons. For example, as of July 30, the NZDJPY-EURUSD correlations are 0.84/1hr, -0.87/1d, 0.70/1wk, 0.91/1m, and 0.11/3m. In a week or a month, the numbers will again be different, with no discernible pattern to the shift. This makes the effective use of correlations as a strict rule set very difficult, but actively balancing trades across the basket certainly mitigates equity curve volatility.)
Profit Targets. The target for realized profits is between 1.5% and 2.5% per month. The annual target is between 18% and 28%. This assumes that at the end of each quarter and at the end of the year, NAV will be (-3%) or less. (See below.)
Volatility. Target volatility in NAV is +3%/-3%, as measured against account balance. Since maturation, the maximum drawdown in NAV is (-8%). With increases in volatility this figure could change. Again, low correlation across the basket will mitigate maximum drawdown.
As of June 2014, the currency model has produced gains in every month since 2010. Maturation was achieved in September 2012. Full operation was achieved in September 2013.