Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I have been working with the MultiCharts Portfolio Trader and I am starting to run into some of it's limitations. Here are several things I have issue with:
I cannot mix and match symbols with signals (a/k/a systems or strategies)
Optimization is done at the portfolio level, not the symbol level
I can deal with #1 by adding this to my code, though it is awkward. Thankfully for what I am doing now, it is the same (equities and ETF's).
The other two I will just have to deal with and find another solution.
This does not mean it does not provide some utility to me. For a system I am not planning to optimize, I can do my backtest and simply filter out the symbols that will not work well. Here is a list of S&P symbols in my test set (criteria: currently in S&P and started in the S&P before Jan 2, 2000; beginning with the letters A and B):
So I can at least take these and work with the ones I know are more likely to work with this system.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,057 since Dec 2013
Thanks Given: 4,409
Thanks Received: 10,225
By this I assume you mean that if you are applying a single strategy to twenty instruments that the optimization finds the single set of variables that yield the highest portfolio results and not a set of variables for each individual instrument. I've always wondered about this. If it works with one set of variables across multiple instruments that has to make you feel really good. Question is if you run different optimizations for different instruments do you keep the instruments that have losing optimizations? I've back tested some published successful trend following strategies, where they use the same variable set across all instruments (futures) and have been surprised to find they continue to trade contracts that have continually lost money. It's interesting because when you look at the individual contract equity curves, many/most would make you cringe, but when combined they look nice. The power of uncorrelated (or even negatively correlated) instruments I assume.
Your assumption is correct: one set of optimized parameters for the entire portfolio. At least, that is all I do for now in MC. In TradeStation, using MultiOpt, as you know, we can optimize each instrument individually. So this is my limitation here with MC.
I think I understand why the classic trend followers do this (I am thinking of Jerry Parker, one of the turtles, and Niels Kaastrup Larsen, specifically). Jerry Parker talks a lot about catching that rare fat tail. They are okay with a 30% win rate, but they will catch that one fat tail trade that makes riding through all the bad trades worth it.
You hit on the two things that makes those equity curves look so good: diversification and negative correlation.
Small wins turning to losses, big wins turning to losses, big wins turning to small wins – get used to that in Trend Following.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,057 since Dec 2013
Thanks Given: 4,409
Thanks Received: 10,225
I agree. I think they have confidence in their methodology/process and believe that the next big fat tail is just as likely to happen in a instrument that has never worked as it is one that has worked several times. But if you follow that thought process, shouldn't you be testing strategies across all instruments not just single ones?
Those trend followers, like Jerry Parker, just run what they have been running for decades across multiple instruments. They may have an instrument with multiple losing years, but then catch that one really good trade (I am thinking of RB's long beautiful run from last November (2020) running into this year as one example). They may optimize parameters, but they are still trend followers. Some of the systems you and I have worked with by way of @kevinkdog's classes require testing across multiple instruments.
The conclusion I came to was that the idea doesn't work on individual instruments, but works when combining instruments. The guy who runs the 40-in 20-out experiment commented on the system, so his thoughts add a lot of insight.
@SMCJB, just for academic purposes, here is an equity curve of the 31 instruments in a portfolio from 1999 to 2021 on the same system, including the ones that would have failed walk-forward analysis:
...and the Monte Carlo, using @kevinkdog's worksheet, just for fun:
Well happy new year to everyone and anyone! I wanted to share a couple interesting things.
I have been reading Algorithmic Short Selling with Python, by Laurent Bernut. This is an excellent book for those more experienced traders, particularly in the algo space. Forget about Python or short-selling in the title. This book is excellent for all of us algos and covers both long and short sides.
Bernut, who I first heard on Andrew Swanscott's Better System Trader podcast about a year ago, is a short selling expert. I've been following him on Quora and LinkedIn, and the man is simply brilliant.
In other news, I am developing a system based on John Ehlers latest article in TASC magazine (Jan 2022), the RSIH (H is for Hann window filter thingy; it is a finite response (FIR) filter originally used in meteorology). Ehlers' tireless work in trying to identify cycles and smooth indicators is on display here:
RSIH and RSI on Daily Chart - MTZ
The calculation period is 13 for both. The smoothing is noticeable immediately, but then we also see a lot of lag.
SPY - Weekly
This is another look using the SPY weekly chart. The system I am building will look for significant changes in RSI direction. The big advantage with the RSIH is the reduced number of signals. The hope is that it is enough to overcome lag. I'll post some of the results after I develop and test.
Cool stuff! I agree Bernut's book is good. He was on Better System Trader podcast in mid December, in case you missed that new interview. He was a guest right after me.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,057 since Dec 2013
Thanks Given: 4,409
Thanks Received: 10,225
Happy New Year vmodus. Thanks for the interesting posts.
Not a cheap book. That normally doesn't bother me but since its about Python (unfortunately I'm a R guy) I was wondering how relevant it would be to a non-Python user?