I wanted to gather up some insight and literature on creating a portfolio of multiple trading systems. Diversification is very important, and I am looking to gather some info on various ways to smooth out the equity curve. Diversification= smoother equity curve Trading Methods: Trendfollowing - ORB - channel/band break out (donchian, %bands, Bolinger, Keltner or ATR bands...... - moving averages Buying on dips with longer time frame trend- Mean Reverting Strategies- cyclical analysis, stat arb (pair trading), VWAP band (intraday), seasonal cycle trading, index arb, intermarket strategies.......... There certainly are many ways to skin the cat. Anyways, I am looking for recommendations on what some of you have found works best, and what works best for smaller account guys <250k . Trying to diversify by trading a basket of commodities makes it very very difficult since fixed % betting is near impossible. Soo suggestion on books that are written for designing portfolios with multiple trading systems, trading a basket of instruments, on various time frames. With proper money management allocation. For smaller accounts that cant trade 100+ instruments, can we find same diversification by trading variations of trend following systems , perhaps on different timeframes? How does one determine position size for two systems with very different trade frequencies, say 1 system trades on avg 2x/week while another 20 ? All suggestions are welcome. Thx
I focus first on finding profitable strategies that are easy for me to trade. I have a small diversified buy and hold portfolio, a long term (months) momentum portfolio (401K with 30 day trading restrictions), factor based stock models (long and short) with hold time of a couple weeks, and mean reversion (long and short) with a hold time of a couple days. Now looking into a short term (days) trend following strategy). I won't win any trading contests this way but it is a way to control risk that works for me.
Here's one way to trade hundreds (or thousands) of different trading systems, out of a modest sized account: Trading a large number of systems without a large account (link) One way to decide which systems to put in your harem, is to choose a set of systems whose equity curve RETURNS have low mutual correlation.
tradingblox forum is one of the best out there, especially for trendfollowing the idea is great especially if you were do divide the subsets of trading systems based on method, fundamental,trendfollowing, contrarian(sentiment,breadth), seasonal etc... and than cast 1 vote per subset and if you have 10 different subsets your max vote count can be 10, in which case you would have a very diversified but same expectation of market direction for the given instrument traded.lol.... has anybody tried something similar to this?? say you run 50 trading systems, 25 of which are trend following in nature differing in timeframe , method of entry,exit, etc... 25 systems based on reversion to the mean/ contrarian methods also varying on different timeframes these systems could vary greatly in terms of profit factor, #of trades , % win or whatever metric you choose to rank them by. if you ran contrarian/reversion to the mean system along with trendfollowing system say on daily bars both with positive expectancy and roughly yielding same amount of trades wouldn't they cancel each others votes most of the time? also would it be better to not weigh the systems equally but base their vote strength on their actual performance? I don't know, this could get very complicated, I don't even know how to begin to test something like this on a platform like multicharts 7. I guess it may not be that complex except I honestly am having difficult time finding good reversion to the mean systems. A few sesonality based systems for futures can have good outcomes but don't have a large sample of trades to make the long term reliable.
You can test multiple systems on Tradingblox. Also, if you search the TB forums, you will also find a great thread showing the benefits of combining all trendfollowing systems, even if they have correlation of 0.6 to 0.7 with each other.