Backtesting Software

mmillar · May 20, 2004

Quote from hypostomus:

...are you trying to bait your betters? Go aggravate Jack, he's an easier target. E-Signal is perfectly fine for the simple strategies which are the proper purview of backtesting. My best codes have maybe 50 instructions in them. With a modest speed PC you can have several of them running in real-time as "helpers" if you're too stupid to recognize your own signals. If you're such experts, what do YOU use? Or are you just flies buzzing around the white trash?
More...

I'd love to bait Jack but he would then bamboozle, confabulate and generally confuse the shit out of me.

If you're my better AND white trash what does that make me :eek:

TradeStation 8 for me. I've tried both eSignal and TS and - trust me here - TS is MUCH better.

hypostomus · May 20, 2004

...I yield. Please, please tell me why TS is better. And why a pre-senescent old man shoud learn a new BT language. E-Signal's Un-Easy language was bad enough.

steve46 · May 20, 2004

Hello:
First I would like to confess that I don't engage in traditional backtesting. For those who want to hear my argument, go to "YM in a can" where I present that argument.
I am posting here because I wonder why more traders do not use Bob Linn's backtesting package "Linnsoft"? It seems to be a nice alternative to TS, wealthlab, Esignal and others. The language is simple to learn and Bob supports the product pretty well.
By the way, I am not currently a user, and I have no relationship to the vendor. I use Excel and find it works just fine.
I look forward to your comments and opinions.
steve46

OddTrader · May 20, 2004

Quote from steve46:

Hello:
First I would like to confess that I don't engage in traditional backtesting.
More...

Would like to learn what's Non-traditional backtesting.

Any pointers (sites, books, posts)?

What are the main differences against traditional BT.

Is it better than traditional BT, how do we know?

How to do it (non-traditional one), if we want to try?

Thanks in advance.

steve46 · May 20, 2004

Oddtrader:
If you read the post, it points you to the thread where I have already made my argument. It is the next sentence after the one you quote.

OddTrader · May 20, 2004

Do you mean this post?

http://www.elitetrader.com/vb/showthread.php?s=&threadid=32206&perpage=6&pagenumber=50

Quote from steve46:

Hello Michael:
Reading your comment about backtesting, I have to confess my bias. I believe that backtesting does not work. The reason is that the basic premise is not valid. What you do when backtesting is to start with a sequence of actions in response to market price (when the market does A, then enter long, when it does B, exit). There is an implicit relationship between the two sides of the equation. Unfortunately that relationship of one price point to another is a dynamic rather than a stable relationship. That is why you can backtest and develop a system that seems profitable on paper. But when you go to real time trading, you find that the drawdowns are greater than you expected, the number of consecutive losers escalates, and the per trade profits start to slip until you no longer have the confidence to pull the trigger. So what do you do then? Well surprisingly the answer is stop backtesting and start testing for correlation and dependency. What you want to do is to find a relationship between to points on a price chart whose correlation is as close to 1 as possible (This is basic stats Michael, if I can learn it anyone can), Then you measure the strength of that relationship by testing for dependency (there are two kinds of dependency, and again this is something you can learn about if you want to just using google). Once you find a stable correlation between two points that exhibits dependency, you can then propose a method of trading that says, "if I see price move to point A, then go long". "When it moves to point b, exit". If you think about it, this is the opposite of what people are doing with Tradestation and other backtesting platforms. Instead of trying to make the market fit your system, you are finding out how the market moves and designing your system around that movement. I appreciate the opportunity to offer my opinion. Hope it helps in some way. Steve46

P.S. By the way, this has been discussed before on ET, by Acrary. look at his posts and you will find examples of correllation and dependency that may help make it easier to understand.
More...

steve46 · May 20, 2004

Yes

nitro · May 20, 2004

I have looked at everything in existence, and I am completely disillusioned with all retail packages.

I am building my own from scratch in C#. I am finding the effort really beginning to pay dividends, but it's a huge amount of work.

nitro

OddTrader · May 20, 2004

Quote from steve46:

Yes
More...

Thanks Steve46. Would there be any other pointers and keywords so that I can search them with Google?

So far I can only find the Acrary/ET thread you mentioned: http://www.elitetrader.com/vb/showthread.php?s=&threadid=14794

Your help would be appreciated.

drsteph · May 20, 2004

I'm a big fan of backtesting after having seen someone use it to design a *simple* trading strategy which operated more or less like a cash mint in the early 90's.

After using backtesting on proprietary systems running on a 286!!!! (all night for a simple moving average with bands) and then Supercharts/Tradestation, I've broken with TS and ported over to Amibroker. I think Amibroker is a superior product for backtesting, although it does require you to have some inkling of programming skills to make it work. I looked at wealth lab, which was interesting, but didn't like the web-based interface. I prefer something that sits on my computer that I can use WHEN I want and if it goes under, I didn't spend all my time and effort learning a programming language which I can't even access (while I port over to something else).

Amibroker is a steal for the price - just expect a fairly steep learning curve.