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)
Well luckily for me I do not know how to merge contracts and so all of my studies came from a single cl contract which should lend to it being robust since it tested well on all the others. On the downside, since then I have tried to install the continuous contract you have posted and it said "unexpected number of fields in line 1 " so I do not know how to get a singe result set to post for you to review. Any ideas RE the Error?
Can you help answer these questions from other members on NexusFi?
I couldn't agree more. I am trying to apply ideas like risk of ruin but I don't have a clear enough idea as to what data I should use because it speaks in terms of $per an hour.
Mike could please you elaborate on how you go about using StandardDev for risk assessment?
I primarily use expectancy plus heavily weight the number of trades and total sample size and then factor that into the total net profit for my custom fitness test.
Zach,
I am not math wizard, in fact quite the opposite. But I believe the general principal of the built-in standard deviation tests inside MultiCharts is that they use the total sample (trades) and compute the standard deviation from one trade to the other, then you can use that as a basis for formulating the standard deviation of the entire test set. I am the wrong person to ask when it comes to math stuff, I rely on trial and error or written examples from people smarter than me...
This is going to be a different kind of post. A different kind of thread. Mike, in fact, might not want it here, but I'm going to put it here, because it needs to be here, and what I have to say needs to be said.
I'd say a profit factor of 2:1 would seperate a very good system from most others. Next, I'd examine the drawdowns which as BM pointed out gets into ones personal risk thresholds rather than having a single number which is good or bad, but needlesss to say, the lower the better.
Also, for those who are using NT, don't forget that backtest results must be COBC =True to have an iota of relevance in your analysis.
All good points here. I'de add as a side note, that if you backtest on tick data (and you should) then you mostly likely will need to forward test and go live on that same tick data source. Tick data is normally different from different sources and it may affect the way your strategy performs (well depending on the strategy).
Also I'de say at least 2 years of tick data, more is better, but Ninja hates many years of tick data, even in NT7. Also many years of data is also hard to come by, hard to load and test with. By going back several years you are covering more market conditions, and the more conditions in the past your strategy handles, the more in the future it is likely to be able to handle (it is more robust).
The other thing is that 2008 was the 'GFC' (or whatever they call it) and this presents some fairly unique conditions, some strats love it, some don't, but it is often hard to have an optimal strategy running through that period and also the current period. If you need to do any optimizing (for example for price targets and stops), I would run it on 2009-10 then run the desired variables through 2008 and make sure 2008 comes out ok also (kind of like a walk back test). Trading in 2008 was like surfing Pipeline, but now the surf is flatter, so there is less momentum and I think you really need to reflect the current conditions for optimization. Maybe some don't agree, just my thoughts.
Also watch out for unusually smooth equitity curves and profit factors higher than about 1.5. When I see that I start to get suspicious ... if it's too good to be true .......
I optimize for min DD, and then look at the list of min DD's and look for the best compromise between DD and profit. Look for the number of consequtive loosers, look for the longest period where you won't make a cent, can you handle a week of loosers, can you handle 3 months of sideways equitity curve. Put the equitity graph on a big wide monitor so you can really see the dips and DD's. A small equitity graph can make a lot of pain and wheelspin look fairly insignificant.
Next I run it on market replays (a NT thing) and check that matches the backtesting results. Then I'll take it live and hope for some immediate postive results to buffer me from the first DD.
Finally I make sure that it has the best chance to continue run in the future ... ie don't go posting it on public forums. Don't do the market the favour of feeding it food to become efficient with.
One of the few nuggets of wisdom I've ever found online from Jim Simmons is that after everything has been passed through their teams of Phd's, the final question rentech asks themselves is "does this make sense?".
He didn't say anything more than that but I take that to mean does it make sense that they are exploiting some market property and not just falling into data snooping bias and spurious correlation but done from a speculative, non quantitative common sense perspective.
If you get some outrageous profit factor on a simple model you really need to first ask yourself if it makes sense...
You can make the arguement that any system who's next trade will be EV+ within your risk management you should trade, no matter how small the profit is. Obviously you want though a model that is going to have a life span but nothing is going to help you there beyond your own speculative common sense and skill.
