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)
Do you mean this: if you were profitable at the end of the Combine, that would encourage you to trade the strategy live?
- Wouldn't that still be a 'not enough data' scenario?
- And why attribute that much importance to profitable vs not profitable? Looking at your bands, I'd estimate breaking even as roughly 'lower 28%' performance (whether it's 27% or 30% does not matter that much). I understand that the everyday view of trading results treats $0 as a special borderline - but why would your statistical approach put that much emphasis on being above the 'lower 28%'? Why not 'lower 20%' (which would encourage trading even when not profitable by that time) or 'lower 40%' (which would advocate stopping even if some profit has been made)?
I hope this is not coming through as annoying. I'm just trying to make my point clear so that I can learn from your reply. (And I've got a lot to learn. I'm nowhere close to your knowledge and experience)
Call me stupid, but if it is just a strategy test, you could have paid for the practice account, only $100 for a month, just to see if the strategy is viable. You don't need the Combine for that.
Even with the free 2 weeks trial you get the stats.... But, hey it is your money, TST would like to thank you...
No, I understand what you are speaking about. I wasn't very clear. For me to actually start trading it with real money, the performance would have to be a lot closer to the average line. The zero line is really psychological. I can't imagine trading a strategy whose "live" performance was negative, even if it was well within the expected range. That is one of those mental hangups I have, like everyone else.
Bottom line is right now this strategy has not convinced me at all that it deserves a real money allocation. BUT, it is early in the process. At least, as of right now, the strategy has not fallen off the proverbial cliff.
No, I will not call you stupid! It is a good point you raise.
I developed the strategy specifically for the Combine. Only afterwards did I entertain thoughts of maybe trading it for myself, if it proves itself. So, you are right - I don't need the Combine to "prove" my strategy - I normally track incubating strategies differently.
So for me, I get 2 benefits at the same time : 1) I get to try the Combine, and 2) I get to incubate my strategy.
Where this all ends up (pass/fail Combine, pass/fail incubation, trade/not trade with real money) really depends on what the strategy does in these 2 months, and probably the next 2-3 months after that.
3 trades, 3 wins today. One trade for each system, the trade on the farthest right of each chart. 2 were really small wins. One should have been a loser, but I exited early so I could play outside with my kids on the first 80 degree day of 2013 (I live in Cleveland Ohio, so you learn to take full advantage of beautiful days like this).
I'll see if I can post the more complicated Monte Carlo analysis tomorrow.
Sorry for the lack of drama lately. Good trading (even if it is losing, like I am so far), is pretty boring...
(Just remember, based on past statistics, I am "owed" 3 days of $500+ profit, with one of them being over $1,000. That should make it interesting, if it ever occurs...)
Hi Kevin,
I would be interested in the Statistical Process Control you mentioned. Would you give us some more clarification, and/or reading suggestions?
Thanks
Thanks for the comment. I do not think I could I know I cannot explain it well enough in a post to do it justice. There are some good books on Amazon that look like SPC for beginners or novices. I recommend starting there.
With factory machines, the idea is you record the dimensions of parts the machine produces, and following certain rules, you either leave the machine settings alone or make changes. This is accomplished thru use of a run chart.
With a trading system, you can treat it as a machine. You only make changes (or shut down the system) when the statistics (from individual trades) tell you to do so.
Sorry I could not be of more help, but an intro text will get you going.
No Trades today, although I missed one during the last 30 minutes of the day (I had some errands to run). That trade lost 2 ticks, so I'll include it on the "perfect" curve.
I will try to post the Monte Carlo chart this weekend.
Here are some curves of where I am at with the Combine:
Average - thick dark blue line - if strategy was performing as it has historically, equity should be close to this
Actual - light blue dots and line - my actual performance, as measured by TST simulator
Perfect - dark blue dots and line - performance according to Tradestation backtest engine, includes $17.50 per round trip for commissions and slippage.
Before I continue this discussion, I realize that this short post is probably way too involved for the casual reader. So, I'll summarize this work in a sentence: Before you change or quit a poorly performing trading system, make sure that you aren't neglecting an unlucky runs of trades - you might be ditching a good system at just the wrong time.
A few posts ago, I showed a fairly simple way for you to determine if you strategy is performing to expectations:
2. Create upper and lower bound equity curves, based on whatever confidence levels you want (68% for 1 std dev, 95% for 2 std dev, etc).
3. If your strategy equity curve is within the bounds, chances are the strategy is performing as expected (although maybe not to your liking)
My point with this kind of chart was to show that even winning strategies may start off on the wrong foot, and lose money. You have to let the long term positive expectancy kick in before you make any decisions.
Unfortunately, this is hard to do. I see people abandon a strategy after 4 or 5 losers at the start, or who subscribe to a strategy signal service for 1 month, and if it lost money, they bail. Such actions may be crazy!
To win at a game of probabilities like trading, you really have to know what the probabilities tell you.
*********************
So, here's a sample case. This is a real world system I currently trade with my own money.
Curve 1: The system performance after 140 days. Barely positive, and close to the lower 10% line. That means, at that point, only 10% of the randomly generated equity curves, based on the system trades, would have been worse than this.
I'm sure most people would have stopped trading at this point.
I did not, since I knew there was a chance that the system was not fundamentally broken.
Curve 2: The curve for the next 200+ days. I kept the system running, and now it is performing close to its expected value. (The thin lines you see are the standard deviation calculation lines, so you can see how close they are to Monte Carlo simulation lines.)
***************
As I explained earlier, I use Monte Carlo simulation to determine if a strategy is performing to expectations or not. It is basically the same result as with using mean and standard deviation (described in earlier post), except that you can include "boundary condition" effects with Monte Carlo (like quitting after % drawdown, or quitting when account gets blown out).
So, here is my Combine system, using Monte Carlo. As you can see, my actual results, while negative, are within expectations. This would tell me the system, while underperforming, is NOT broken.
To create these curves, you can just stop the Monte Carlo sim after X trades, and plot the percentile results that you want. For most people, it probably is easier to use the earlier method I presented, and frankly, the results are just as good.