CTAs - Performance and replicating signals - Trend identification works

A recent strategy paper from a large bank quant group focuses on the benefit from holding CTAs in a portfolio and the signal generation from CTAs. Diversification is enhanced and drawdowns are reduced when CTAs are added to an equity portfolio and to a lesser extent fixed income portfolios. CTAs show consistent long-term performance albeit there will be periods of negative return. Of course, this work does not make the key distinction between CTAs and trend-followers. Trend-followers are CTAs, but CTAs are not always trend-followers. Nevertheless, the research further documents the value of trend-following when added to a portfolio.

These conclusions are well-known and reinforce the value-added from CTAs; however, this is not the main purpose of this research. This quant group attempts to engineer CTA signals in order to help investors better manage asset class exposures. Take what CTAs do and convert the strategy into signals on flow and crowdedness so that investors can use these signals without investing directly into a CTA. 

The quants suggest that these signals provide a better tool for investors that trying to determine CTA positioning from commitment of traders reports as generated by the CTFC or making the direct investment. Follow CTA signals and you may get the best of both worlds, signals that enhance returns and offer diversification. 

This type of signaling research can be useful, but the analysis presented will generate reader confusion. First, the research does not define terms correctly. Trend and momentum are not the same, and this research does not focus on this difference. Second, the presentation of signal development is not clear; consequently, the quality of the signal is suspect. Trend followers may be similar, but there are large differences in the look-back period and processing of data. Without clarity on trend identification, the signal does not tell us clearly what is being modeled. Third, the signals are structured to replicate the thinking of trend-followers and not find the best trend signal for markets. The signals are a combination of exponentially weighted moving averages adjusted by market volatility and converted in signals between 1 and -1 while accounting for liquidity and portfolio volatility. They then assume that these signals represent the trading positions of CTAs.

Nevertheless, we can learn from this work. Trend signals have forecasting power and do not serve as just indications of flow from CTA strategies. Second, trend signals are noisy indicators of crowdedness just like the commitment of traders information. Strong identified trends like large trader commitments do not reverse but show continuity. We cannot make judgments on reversals. 

The bank quant group spends significant time linking CTAs (trend-followers) with their CTA trend identification model when what they are just doing is providing trend signals that can allow investors to replicate the behavior of trend-followers. All they had to do was say that they generated a good trend signal model that is similar to a classic trend-following manager.