For me it is quite easy to come up with Mean Reversion-type strategies and have a mental block towards trend following/momentum-type strategies. For example, one of my mean reversion strategies consistently loses money (thankfully still in SIM) and made me wonder should be taking the opposite of buy/sell signal generated by my strategy. To expand on this further: I pairs trade index futures (ES VS NQ, ES VS YM, ES VS RTY etc...) in an unusual way (using net change in $dollar terms) and have a Z-score defined, enter long when Zscore is -2 or lower and go short when Zscore is +2 or higher. This almost consistently losses money - I was wondering if I were to do the opposite i.e., go long when Zscore is +2 or higher and go short when Zscore is -2 or lower would it make mine a trend following strategy? I'm going to make that change to my algo and see how it performs in simulated environment but I thought I would check here and see what others thought.
well, kinda. trend following/momentum trading is a bet on autocorrelation. If your asset class is autocorrelated, trendfollowing will work and vice versa
%% Initial impression is that could work well ; but UDOW[dow X 3] almost NEVER/never does as well as SPXL [spyX3]. However on an intraday it does sometimes. APRIL dow may do better than april spy.; but dow stiil tends to under perform spy/spxl/upro/es .GE or DAL shorts may have made more than longs; but they downtrended so much= most trends have been down...……………………... {Except for april dow,i watch NQ/SPY derivatives more; +momentum, which is a price speedometer has little to do with trend following in stocks/or stock based derivatives} If you have to/chose to be out by eod/early momo could help. But trend following is not really a mirror image of mean reversion/counter trend……………………………………………………...And most short systems dont make as much as long systems; if it was a mirror image; results would be much closer, like a mirror.
if u are reversing ur strategy then u r reversing your pnl less spread and commissions ofc. Plus your limit orders become stop orders (or both are simply market orders). Why would you expect any other result when you are reversing ur strategy ? I mean, when u think logically ? Or am i missing smth ?
%% Tq poster has good countertrend timing/LOL; much different from a trend...……………………………………………………………………………………………….Good gap down/ mean re/countertrend today.
I made the change last night and so far it is looking much better - basically go long when my Zscore Calculation is +1 or higher and go short when it is -1 or lower. This switch seems to make sense (particularly the way I construct my spread). I will monitor this for another week and if everything checks out I will take into Live.
%% Good brief counter-trend downtrend/QQQ/SPY/DIA. Re-read 2019 Stock Traders Almanac, last nite=SPY/DOW tend to be down AUG/SEPT. Can not, generally speaking-- let the profits ride on mean re[aka countertrend]; like the profits ride on trends...………………………………………………………………………….BUT if some thing goes down7/10%/or up in 2 days/summer; I aim for some of that, sometimes/LOL
Presence of autocorrelation says a time series can be approximated by some function of its several lags, while clearly momentum is short-term phenomenon which happens only sometimes (rarely). So I think presence/absence of autocorrelation says nothing about occurrence of momentum simple because the process of approximation "absorbs" or "hides" occurrence of anomalies. Any thoughts?
For stocks, there is some mean-reversion in the one-month horizon, some momentum in the one-year horizon, and then mean-reversion again in the five-year horizon. So it all depends on the timescale chosen.