Some changes in price action dynamics this year have impacted the performance of certain classes of quant strategies. This is in addition to changes to risk and money management strategies in the quant domain that also impact performance.
The most important change this year was an increase in market volatility. Below is a daily chart of SPY ETF showing the rolling 21-day standard deviation of daily returns and the 14-day ATR expressed as a percentage of closing price:
Both measures of short-term volatility in above chart show values nearly double the longer-term average. Especially the 14-day ATR rose to levels previously seen during the major correction of 2011. More importantly, there were two peaks in volatility this year, which has happened only during bear markets in the past. Quant strategies often experience difficulties when there are large swings in volatility since most of them are designed to benefit from either low or high volatility but not from both.
Below is a variation of the Momersion indicator that measures the percentage of down days followed by up days in a 252-day rolling window.
This indicator is essentially a measure of the effectiveness of “buying the dips”, which is a strategy implicit in a large class of quant strategies. It may be seen that the indicator value collapsed in 2018 from a high of around 65% to nearly neutral 50%. This means that during this year, as most traders have already realized, the “buying the dip” strategy has lost its effectiveness.
In addition, note that rapid declines in the above indicator have occurred in the past along bear markets and during major corrections. When the market is on a strong uptrend, this indicator usually rises to above 60%.
In general, quant strategy problems arise from at least two sets of issues:
- Changes in price action dynamics
- Nature of quant strategies
(1) Includes the following:
- Higher volatility
- Increased noise in price action
(2) Includes among other things
- Tendency of quants to minimize risk
- Decreasing influx of dumb money
- Use of “pedestrian” quant strategies
In dealing with the above issues we have developed techniques to extract “idiosyncratic” alpha from noisy and non-stationary price action. This applies mainly to long/short equity but also to directional strategies and it is based on features developed by DLPAL LS software that have economic value. An additional layer of classification based on machine learning (such as SVM) can offer significant alpha boost as our customers around the world report. For an outline of the methodology and examples of results see our recent presentation in M4 Conference and specifically slides 37 – 44.
If you found this article interesting, you may follow this blog via any of the methods below.
If you have any questions or comments, happy to connect on Twitter: @mikeharrisNY
Charting and backtesting program: Amibroker