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)
Thanks again for your previous answer. Now something came up that are related to previous question. When Building portfolios i see that some strategies from some instruments are dominating. Now it is no problem if they do as long as they perform but if they do not perform it becomes a problem. Below you see some live traded results from a portfolio that have been running for i think 8 weeks or maybe 9 (160 trades). The 2 strategies on the far right are gold and they have hurt the performance badly even thought totally it is still in about 9% profit this morning . When building a portfolio what should i do. Remove the dominating strategies or leave them in. Or maybe instead change and individualize the moneymanagement from the start to even them up as per their historical performance.
If the two strategies on the right would have been winning it would actually also look strange since then the other strategies profits would have been really small in comparison.
The answer to the question "what should I do" depends a lot on what you want to accomplish with the portfolio. Are you going for max returns, or maybe minimum drawdown, or smooth equity curve, or excellent return/drawdown ratio, etc.
Once you figure out what your goal is, you can examine the pieces of your portfolio, and construct a portfolio to meet your goals. Maybe Gold should be taken out, or maybe you should trade mini Gold. Or maybe you trade 2 or 3 contracts of everything else to 1 contract of Gold. Or maybe you just have a heavily weighted Gold portfolio, which maybe offsets some other investments you have.
I hope you see the point - what you have found is something correctable, but only if you know where you want to get to.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals, U308 and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,059 since Dec 2013
Thanks Given: 4,410
Thanks Received: 10,226
Great question @Mabi, I'm only trading a few systems (as a systematic trader) and have found a similar situation with one system dominating prtfolio performance.
Assuming your confident in the system I think this is the best answer... but only if you have the capital to do it.!
Comparing with how you run a company you never want your whole existence depend on one or 2 costumers and you would like to be able to budget next years result but you still want the profit from the highvalue cotumers though but they shall not be able to pull you down if they leave. So i guess i will try to even them out in some way prefeably increasing contract size on the lesser performers,
I can't answer the decision of whether to include or exclude systems, but I can give some insight into portfolio allocation methodologies that you could use. When I was still trading a suite of systems, I used either one of the following methods to assign risk / weightings to my systems.
You can view your systems as independent systems and allocate capital evenly across them at the beginning. The benefit of this method is that winning systems will have a greater allocation of capital and losing systems will have a smaller allocation of capital as time progresses.
You can view your systems as all part of one portfolio and constant re-weight them. An easy way to do this is to always risk x% per trade of your total portfolio value. Downsides are the opposite of the positives of the independent system approach, but when a system starts to become profitable it recovers losses much quicker.
If you wish, you could also add drawdown-reducing techniques to both methods. Some drawdown-reducing techniques are things like a) halve size every with every 10% in drawdown and bring it back only once you recover the drawdown-threshold, b) halve size after winning % drops to x%, or c) halve size after a certain number of losing trades in a row. You would need to keep track of the individual metrics of each system for this to work.
Final thought - with regards to gold dominating your portfolio - if you are basing your stop-loss (and then sizing your positions accordingly) on ATR or something similar, then the allocation to gold should be equal to an allocation to treasuries. I am guessing here, but I would suspect that your sizing on gold is larger than it should be in terms of ATR. If that is the case, you need to evaluate whether the added risk is worth it to you. If not, then you need to use Kevin's criteria to determine whether it should still be included.
@grausch
Thanks again for your reply !!. I have been busy at work but I had time to shortly after your post lowered the contract size on the less performing strategies and increased them equally on the performing ones.
System shortly after went in to a drawdown and Gold kept loosing but with last couple of days due to increased volatility portfolio has been doing pretty well and even though Gold has continued to loose it's impact is now much less on the portfolio as you can see from the screen dump. I might ditch the Gold if it keeps loosing but it is not time for that yet due to it's historical performance. Portfolio has now done around 330 trades since October 10. I plan to launch 2 more portfolios by January 1 and if I find time I might start a thread how I did it. However I rather start a thread if it is doing good with some longer history.
Hi Kevin,
First I would like to thank you for taking the time to answer many posts very thoroughly.
I apologize in advanced if the following has been discussed here before but the thread is getting large
When optimizing a system there are a few types of strategy parameters. There are parameters you wish to check for a wide range of values and there are parameters that you only want to check 2-3 values. Sometimes it can be an "on off" switch parameter.
Do I need to differentiate between those types of parameters while determining the minimum trade count that is sufficient per in sample? In your book you say that 30 trades per parameter is sufficient, is it true for "on off" switches as well?
Another thing I was wondering about, if I am optimizing an intraday strategy and I do have more than enough trades for "avoiding overfitting" in 3 months, should I use more in sample data just in order to avoid optimizing (or exploring the idea) in the in sample to a certain "state" of the market which can pass shortly (like a 30 days of trend for instance)?
And one last subject I have issues with, should the data that I explore my strategy initially be a part of the in sample in a walk forward optimization?
You wrote before about not to try to optimize the backtest but if I do optimize by adding more filters (in the first in sample) and still have enough trades per parameter is it OK? Will the walk forward tell me the answer or should I avoid running too many walk forward optimizations in order not to look too much at the out of sample data?
I have never treated those as different, but I guess you could. Remember though that the 30 number is just a rule of thumb, ideally you'd want 50 or more. So, I'd be conservative and treat on/off the same as any other.
I would definitely test an intraday strategy over more than 3 months. You want something that works over the long haul, with a lot of different market conditions. You only see limited market conditions in 3 months.
I usually leave the initial exploration time period in the walkforward. I would try running one walkforward optimization, and that is it. Everytime you make a change and rerun the walkforward, you are getting less and less "out of sample - ness". Rerun walkforward enough and you get an in-sample test. So be careful here.
One other thing I have a problem with is that when we are testing over a few years an intraday strategy it means that we disqualify the possibility of new strategies that become relevant only in the last few months. For example, recently the iceberg orders in forex futures were cancelled. After such an event new strategies that take advantage of level 2 that were not valid before can suddenly have a merit. Should I not test these kind of strategies only because I do not have a few years to test on?
Also, if my strategy has only 3-4 ticks of target per trade, the "trendiness" or many of the other "states" of the market do not effect it. So a strategy like that cannot be optimized over a period of 3 months or so (with ~50 trades per parameter)?
About the walk forward optimization, if I make on walk forward run, discover I overfitted the data, remove one filter and rerun it, will it be valid?
If I make an agreement with myself not to look at the results thoroughly, only look at the cumulative profit graphs and if they do not satisfy me I will readjust the strategy, does it hurt the out of sampleness? I will not derive anything from the out of sample data but the shape of this specific graph.
Of course you can do all these things if you want to. I personally don't, but that doesn't mean you can't. The big keys to strategy development:
1) develop strategies in a way that you feel comfortable with
2) validate your approach with real money trading, and adjust your approach as necessary.
I know other traders who laugh at and tease me about my development process. They do things differently, and if it works for them, who am I to judge? I do what works for me. You should do what works for you.