Screenshots? We've been discussing about systems and etc. over the past few weeks in ET. And you expect me to have charts? How about a bunch of csv files that are not labelled? How many do you want? For each set of tests I've mentioned there's about 800+ csv generated. Should I have them zipped and post it here? Hold on... I'll post it...
How can I post a 15MB zip file? Split it over... 15-16 archives? This site needs bigger storage space!!!
Before it goes off topic with TA discussions, which I have no intention to do so... What I wanted to say is that, "There is a high possibility that people think similarly and figure out the same thing as you, no matter which tool you use. Most likely, there's someone else trying to trade a similar way as you fighting for a piece of the "edge's" pie." That said... I'll get back to the original topic. Dealing with changing markets.
Here's an excerpt of the output file... for the TA Correlation... Dunno if it really helps people but if you insist...
Few comments: 1. I haven't posted one system. I'm only mentioning about the results I got from the tests I've ran in the past. A test result and a system are 2 different things. 2. Robustness is a myth. ALL models are curve-fit. Seriously, this "debate" is done like 5 years ago. Some people need to keep up with the recent trend and developments regarding systematic trading. There is only 2 forms of robust trading models: A. One that does not trade. B. One that never makes or loses money. Meaning... one that gets killed by the costs of trading. 3. I haven't seen one post in all of ET about how to backtest properly. The fact that people are talking about systems and models using conventional backtest reports to assess a trading model tells me that they just won't get it. You backtest to assess the risks of a model, or a portfolio of a series of models. Not what you can expect from trading them. 2 different things. Seriously, why do you backtest? What are you trying to acheive from doing so... Why do people stop using vendor testers and go custom?
I deal with changing markets by having a simple, effective algorithm to switch off a model when it stops working. No good model lasts forever, my goal is to exploit it while it lasts, and be constantly re-evaluating/tweaking it and be ready to react quickly when I notice the market condition/premise on which it is designed is shifting. By having a portfolio of models, I only need a reasonable number to hold up in production to get a decent overall result. Well designed robust models don't fall apart spectacularly, they tend to flatline first giving opportunity to switch them off before they do too much damage.
I've got plenty to bite but I'll stick with the last paragraph... What surprises me is that people believe that putting a lot of effort into developing a model automatically leads to a higher probability of having a favorable future. You mention that a well designed robust models don't fall apart spectacularly... etc. etc.... If there was a logical / rational way of flatlining the future performance of a system, wouldn't it be reasonable possible that you can predict a downfall? In another words, you are simply stating that is possible to predict the effectiveness of the system to a point where you can flatline it. Another irony is you mention that systems won't last and you try to exploit it while it lasts... though you mention about robustness. You are contradicting yourself. Do you actually have all your thoughts coded or are you using discretion regarding these topics. Added: I'm very sure, you are talking more of what you want to do rather than doing what you mentioned. I skimmed through your past posts in ET. Of course, you can deny and argue... Argument 1( out of 4) that I can pull out.... It's very interesting how you are a systematic trader monitoring multiple models but you don't have your Sharpe correctly calculated...
Yes, on re-reading I did contradict myself. Remove "robust" and I stand by my statement. I originally found this thread interesting and thought I'd share my experiences, yet your reply is so combative ("of course you can deny and argue") there is no point for me continuing with this.