The most robust backtesting framework for Microsoft Excel. Practitioners that use technical indicators tend to use the following ones to exploit momentum patterns: Values close to 4 indicate strong negative autocorrelation (mean-reversion).Values close to 2 indicate the absence of autocorrelation.Values close to 0 indicate strong positive autocorrelation (momentum).This formula of this test results in a single value that ranges from 0 to 4, which indicate the following: There are multiple ways to quantify the autocorrelation of a series of historical prices (or any other time series), the most common being the Durbin-Watson test. Informally, this means that an estimation error with a given sign (positive or negative) tends to be followed by an estimation error of the same sign. Momentum is defined as “ positive autocorrelation” in statistics. Thus, using an intraday mean-reverting strategy together with a momentum interday strategy is plausible. Previous research might show that a given asset features short-term mean-reversion, but momentum over a longer time horizon. It is not uncommon to have two algorithms trading at the same time, one exploiting mean-reverting patterns and the other one momentum ones. Different assets tend to have different behaviors, and market regime changes also determine the overall preponderance of one over another, it is not contradictory to use both techniques simultaneously. There are copious amounts of academic papers rigorously analyzing how price increases tend to be followed, on average, by further subsequent price increases, and vice-versa.Īt first, the hypothesis behind momentum trading seems to contradict the thesis of mean-reversion algorithms. Momentum strategies consist of a set of rules that aim to exploit the tendency of asset prices to continue changing in a given direction. I provide a brief explanation of this test in the following section. Statistically, it is possible to test for either mean-reversion or momentum of a series of prices using the Durbin-Watson test. The details of the Ornstein-Uhlenbeck are definitely out of the scope of this article, but you can further read about it in this excellent article. In a nutshell, this process has three main components: the rate of mean-reversion, the long-term mean, and the average volatility of the process (Brownian motion). Mean-reversion can be mathematically modeled with an Ornstein-Uhlenbeck process. You can refer to this paper in order to review a robust analysis of mean-reversion. Having said that, the academic literature suggests that there is significant evidence of mean-reversion patterns in the prices of stocks. Thus, it reacts slower to current changes in price.Ī mean-reverting algorithm will go long whenever the fast-moving average crosses the slow moving average from above, as a result of a strong price decline.Ĭonversely, it will sell the asset when the fast moving average crosses the slow moving average from below due to a strong positive price change.Īs expected, such a simple algorithm with an arbitrary set of parameters (30 bars, 90 bars) will most probably perform rather poorly if implemented. Slow Moving Average: a simple average that goes back in time further than the fast-moving one.A popular value for the fast-moving average is 30 (bars). Fast Moving Average: average of the most recent closing prices.These strategies are used not only for buying but also for selling an asset, since most mean-reverting algorithms are instructed to go short after substantial price increases, and long after declines.Ī simple mean reversion strategy can be created by means of two moving averages: Mean reverting strategies try to profit from sudden and big price changes and their tendency to revert to their “original” price, or at least in that direction. Strategies Based on News and Sentiment Analysis.Pure Arbitrage Strategies (High-Frequency Trading).
0 Comments
Leave a Reply. |