Thursday, May 3, 2007

An Email From Stockfetcher

I sent an email to Stockfetcher.com asking them if they had plans to
expand their backtesting capabilities. The following is their response.
The inability to backtest more than two years, and the inability to backtest
during a bear market is a fatal flaw. I can only guess that either their
subscribers or the company itself does not truly understand how useless their
service is without allowing backtesting over a much greater span of time.
I am not-so-secretly hoping that if enough potential subscribers or current
subscribers ask for expanded backtesting capability, that Stockfetcher will give
it greater consideration.

***Edit*** I should qualify what I mean by "useless." I am seeking development
of a longterm trend following system. Not being able to test over a long period
of time makes Stockfetcher useless to me. However, for short term swing trades
of a few days to a few weeks, their backtesting service may be very valuable.
If cornered, I would still argue that stocks will move differently 3 weeks after
an entry point in a bear market than they will during a bull market. Perhaps
Stockfetcher could include 2000-2001 in the periods covered by backtesting to
provide their users a downtrending market to test strategies. I would expect
that anyone testing short strategies is consistently stymied testing from
2002-2007.

Hi,

Thank you for the feedback. StockFetcher does limit a single back-test
to 2 years; however, the period covered by StockFetcher backtesting is
2002-2007 (i.e. you can test over any 2 year period in that range.)

Due to various performance requirements, at this time we do not have
plans to extend our backtesting period beyond the existing 5 years.

If you have any additional questions, please do not hesitate to send us
an email.

Thank you,

Tom
StockFetcher.com Support
Vestyl Software, L.L.C.

11 comments:

mdawsz said...

They should just change their name to 'stockfelcher.com' and call it a day.

Dogwood said...

I haven't experimented extensively with StockFetcher yet, but the service appears useful for developing swing trading strategies.

The two-year back testing feature using seven years of market data can help you develop robust trading strategies by using different start-stop dates to avoid curve fitting.

However, the back testing limitations make the service unsuitable for developing long-term trend following systems.

To develop these types of systems properly, you need to back test for at least five consecutive years, but 10 or 20 are much better and more reliable, almost guaranteeing that a significant bear market will be represented within the test results.

So, I don't think the service is useless, its just not a good fit for what you are trying to do. Its simply not the right tool for the task at hand, for that I believe the winners are still Wealth-Lab and Trading Blox...if only they had a StockFetcher price!

Woodshedder said...

Dogwood, you are correct, and I shouldn't have said "useless" or I should have qualified the term. In fact I will edit the post.

Flip It said...

Have you looked at AmiBroker. To backtest, you need some programming knowledge. I use it and find it works for me.

Woodshedder said...

I have not checked it out. Thanks flip it, I'll look at it tonight, although I have ZERO programming knowledge.

Tutor said...

Shedder, I recommend a look at TradeStation, now Tradestation.com. I was once a user group leader for them and was very pleased with the capability. Man, we wrote some pretty crazy systems. I strongly suggest a trial, though you will need at least 6 months or so of use before you'll actually have a handle on it, its that robust.

Your systems are written in what they call "Easy Language" and once you get the hang of it it really is simple, sort of a broken English syntax. You can even call outside code via .dll files.

The system testing capabilities are by far the best available; no limits. It goes for a monthly subscription with discounts for using their b/d for executions. Vast data available as well and charting is superior.

I see you over at Fly's now and then and respect your efforts.

If you're game you can email a piece or two of the system you want to construct and I might be able to give better direction. In the past I've traded with everything from basic moving averages all the way through chaos theory and neural networks (those were fun). If its been advertised in Stocks and Commodities magazine (the trader's bible) I've likely used it.

If you want something a bit easier to get your arms around I suggest Metastock or AIQ Tradingexpert. Both have system writing and backtesting capabilities, just not as robust. There is no reason to spend $2900 on a chart/system package these days.

Anyhow, good luck with your quest...

Woodshedder said...

Tutot, good stuff. I'll respond tomorrow. I'm falling asleep at the laptop right now.

Broker A said...

Hey, I got a fucking system too.

It's called buy the stock before it goes up. It seems to work for me. You should try it sometime.

Maybe you chart chomping imbeciles can write me a program.

Tutor said...

Fly, you're such a moron sometimes... While most of the newbies to trading gravitate toward "chart chomping" as the Holy Grail, more sophisticated and seasoned traders learn that the chart is just one part of the bigger story and should NEVER be ignored.

Furthermore, systems trading extends far beyond simple price volume relationships and can be extremely sophisticated quantitative and statistical trading engines.

Woodshedder said...

Tutor, good to see you still checking in here. I've got the tradestation demo, but I haven't got too deep into it yet. Some questions that I have are the extent of the statistical reporting. Some things I'm interested in running are Monte Carlo sims, maximum adverse excursion, and random entries. Do you recall if TS will generate these reports for a given strategy? Also, does TS allow the inclusion of a position sizing algorithm in the testing? Thanks!

Tutor said...

I think I mentioned that I haven't used TS for a number of years but as I recall if there exists a formula that can be written their systems engine will run it. You might Google "Tradestation" and see what hits you get as far as statistical uses go.

I've seen some pretty heavy coding by some very intelligent people. We used to have monthly user group meetings with on average 40 people in attendance and discussed every which way to trade the market.

Take a demo on it and I have a feeling you won't be dissappointed. In fact you'll probably be so intrigued you'll lock yourself in a dark room for days at a time testing ideas.

I long ago moved onto quantitative analysis and factor modeling. Discovered some cool methods of explaining many of the technical charting indicators, breakouts and so on by modeling fundamental data.

If you really want to go nuts take a look at MatLab (mathworks.com)...