Backtesting Software
- Thread starter minmike
- Start date
ppp:
>Other than possibly some speed issues...
I own and use WL every single day. The program has been worth every penny of it's price and more. That said, WL's biggest limitation IS speed. Yeah, you can code most anything you dream up in WL, but don't expect to multi-variable crunch those dreams on full NYSE and NAZ markets -- it's just not going to happen on any practical level (and yes, I've got a fast computer with lots of memory). I find that once your tests involve more than a couple million candles you are spitting into the wind with WL. Many of my tests involve 10 times that.
I really like some of the output options on WL and believe in testing any potentially viable system on separate platforms just to be sure so WL has become that secondary platform. Tests that can take 4-5 hours for one pass in WL can be done in just a few minutes or less on the proprietary system.
If you are not an expert framework programmer, WL is still a very valuable option.
JB
>Other than possibly some speed issues...
I own and use WL every single day. The program has been worth every penny of it's price and more. That said, WL's biggest limitation IS speed. Yeah, you can code most anything you dream up in WL, but don't expect to multi-variable crunch those dreams on full NYSE and NAZ markets -- it's just not going to happen on any practical level (and yes, I've got a fast computer with lots of memory). I find that once your tests involve more than a couple million candles you are spitting into the wind with WL. Many of my tests involve 10 times that.
I really like some of the output options on WL and believe in testing any potentially viable system on separate platforms just to be sure so WL has become that secondary platform. Tests that can take 4-5 hours for one pass in WL can be done in just a few minutes or less on the proprietary system.
If you are not an expert framework programmer, WL is still a very valuable option.
JB
Turok, Okay. I'll buy that. I test and trade all stocks with enough liquidity on the Naz, NYSE, AMEX - the list is something like 5000-6000 stocks - and over 10 or 20 years of EOD data, speed hasn't been a problem. But that is EOD. I can see where intraday stuff might jam you up.
I also test and use autotrading and backtesting option systems on home-built C++ and VB programs. But WLD saves much time in coding. You can literally code a profitable system in less than a minute.
Regards.
I also test and use autotrading and backtesting option systems on home-built C++ and VB programs. But WLD saves much time in coding. You can literally code a profitable system in less than a minute.
Regards.
Loosen Up:
When you characterize market data, you look for relationships between data points first. When (If) you find a stable relationship (that is repetitive) between data points, you can then choose indicators to signal you each time that specific relationship exists. In contrast, backtesters look at data, and without knowing if a stable pattern exists, they choose indicators, either at random, or in a purposeful way, and try to fit them to that data so that profitable signals are generated. These are two diametrically opposed ways of trying to make a dollar.
I hope this clears up your confusion.
Steve46
When you characterize market data, you look for relationships between data points first. When (If) you find a stable relationship (that is repetitive) between data points, you can then choose indicators to signal you each time that specific relationship exists. In contrast, backtesters look at data, and without knowing if a stable pattern exists, they choose indicators, either at random, or in a purposeful way, and try to fit them to that data so that profitable signals are generated. These are two diametrically opposed ways of trying to make a dollar.
I hope this clears up your confusion.
Steve46
Steve46,
Sorry, I am still confused. Ok, lets put it this way. If I find from eyeballing charts that a certain relationship exists and then try to design a system which is logical and consistent with my market knowledge, and then I backtest this over a period of back data, does this fall into your definition of backtesting? Else, is this some sort of finding correlation?
Thanks
In my opinion "eyeballing" data does not work. Search Acrary's posts for an excellent example of how correllation and dependency work. In my opinion, unless you can find a correllation between data points that is stable (tested) you are just guessing. Worse, you are usually betting your money on a something that may disappear quickly, taking a chunk of your trading account with it. Backtesting wont tell you whether this is the case. Testing to verify acceptable levels of corellation and dependency will. In my opinion that is the difference. I have to get some sleep. Take Care, Steve46
Steve,
Identifying stable, long term, correlations and dependencies is
EXACTLY what I do when I backtest.
Although im certain many people DONT do this in their backtesting,
its my #1 goal.
Not sure how you can claim that backtesting WONT tell you
whether this is the case or not.
There is nothing inherent in backtesting that someone prevents
this from happening.
I can do the SAME thing in a backtest that you can do in Excel.
Excel is just a more limited backtesting framework as far as
im concerned.
peace
axeman
Identifying stable, long term, correlations and dependencies is
EXACTLY what I do when I backtest.
Although im certain many people DONT do this in their backtesting,
its my #1 goal.
Not sure how you can claim that backtesting WONT tell you
whether this is the case or not.
There is nothing inherent in backtesting that someone prevents
this from happening.
I can do the SAME thing in a backtest that you can do in Excel.
Excel is just a more limited backtesting framework as far as
im concerned.
peace
axeman
Somehow (after reading many posts here) I share your feeling, as:
- Just becuase it doesn't work in theory doesn't mean it won't work in practice.
- Just becuase it does work in theory doesn't mean it'd work in practice.
- Just becuase it doesn't work in theory doesn't mean it'd work in practice.
- Just becuase it does work in theory doesn't mean it won't work in practice.
- Just becuase it doesn't work in practice doesn't mean it won't work in theory.
- Just becuase it does work in practice doesn't mean it'd work in theory.
- Just becuase it doesn't work in practice doesn't mean it'd work in theory.
- Just becuase it does work in practice doesn't mean it won't work in theory.
- Etc.
- Etc.
Propably you need to read this post first:
http://www.elitetrader.com/vb/showthread.php?s=&threadid=32206&perpage=6&pagenumber=50