Saturday, May 05, 2007

StockFetcher Setup

I use a certain configuration in my back testing. It is not unique to the test but closely matches my own trading style and methods.

First when I trade I try not to have more than 4 stocks in play at any one time - I prefer to have a lot of shares of a winner than a few shares of a winner and a few shares of several losers. And realistically - they all can't be winners - you've all seen that I pick my share of losers - I publish them on this site. You've also seen me add to day after day of those that were running.

Second I tend to bail quickly on winners and losers. Although recently with some stocks such as AKS and GGB I stayed with through their run. But again that was a market conditions type of play - I was playing the February 27th correction and now it seems to have been played out.

So, bottom line, when I set up my testing in stockfetcher I attempt to match my hit and run style. Here are my system settings -

Name: First Approach Test
Approach Type: Long
Start Date: 12/05/2006
End Date: 04/03/2007
Benchmark Symbol: ^SPX

Exit Setup
Stop Loss: 8%
Profit Stop: N/A
Trailing Stop Loss: N/A
Minimum Holding Days: 1
Maximum holding days: 4
Exit Trigger #1: close 1 day ago > close 2 days ago
and close > close 1 day ago

Extra Indicators
Entry Columns:
Show Performance After: after 1 day
after 4 days
after 10 days
after 20 days
after 30 days

Advanced Options
Selection Method: select by volume descending
Entry Price: open
Conditional Entry: No
Exit Price: close
Maximum Trades Per Day: 1
Maximum Open Positions: 4


The dates shown are what I call my "current period" - I normally step back and forward by 1 month when I test for my own personal use but only show the current period.

Some additional thoughts on testing while I've got your attention -

While a long term test (greater than 6 months) is interesting its only value is to show long cycle changes. For example - I report on the INDU ATR every day and am convinced that every time it dips below 80 that within the next several days we will have a mini (or maxi) market correction. In order to see that you have to look at years and years worth of data - that is a macro change. I have run tests on this (using my old fashioned Excel laboratory) and it seems to hold true. I also use the NewMoMo indicator in my trading - if it goes above .005 the market goes down (most of the time) if it goes below -.005 the market goes up (most of the time) - once more I use my Excel lab for that testing. These things inform me regarding the condition of the market - NewMoMo is why I say the current rally is "soft" and the ATR seems to be telling the same story as it approaches 80. What would I do with such information? - Well I'll sell all of my long term holdings (I started yesterday) and switch back to an exclusive day trading model in the next week or so. Do I recommend this for you? Oh hell no - each trader has to take responsibility for his or her own actions.

I am a short term trader at heart - day or swing. I am interested in a 4-day window and I am interested in what is working now - not what was working last fall. I can run tests that prove conclusively that certain filters (approaches is maybe a better word) work better during some market conditions and poorly during others. I've discussed this numerous times during my tenure posting here.

The market is fluid and your approach to trading should be fluid. Yes there are certain things that seem to be always ready - things like the RSI(2) (for short term trades) and the ATR moving up and down above and below average and tweezer bottoms and tops. These signals are paramount and should be in your toolbox at all times.

Somethings like cross overs and BOBs only work well during some market conditions and not so well during others. Cross over works best most of the time since it precedes the moving average cross over that also works well most of the time (a moving average cross over is not bad - it's only late).

But most approaches only work some of the time. Yes you can invent a method that will give you super returns with a massive win percentage and that finds 4 stocks a year - would I trust that filter the next time out - not on your life - or as I say - I wouldn't trade those stocks with your money - never mind mine.

So what I am looking for is not nirvana nor the Rosetta stone nor the alchemists magic formula for turning lead into gold - what I am looking for is a reasonable chance to make a buck. And the hardest part of that calculus is finding a stock that hasn't already been found or finding one that is one day away from being discovered.

And that's why I have methods such as the BOB where you find the set-up, set the stop (physical or mental) below the low of the second candle and if it goes up you take your profits graciously and if it goes down you take a small loss and move on to the next one.

3 comments:

Anonymous said...

Thanks!

L.J. said...

The 4 day maximum hold makes perfect sense. I've generally been using 15 days, but that's not really the way I want to trade. With most of these filters my intuition is that they are most powerful in the days immediately following the trigger and the longer you hold them, the less predictable it is that they'll continue to go up.

I've been thinking a lot about needing to trade more like the backtester. If I would just stick with taking small gains (and small losses) quickly my returns would be a ton better.

One other thing I wish Stockfetcher would do? Let me have an option for "don't hold through earnings". I almost always sell before the earnings release and without being able to model that, I'm thinking it's hard to tell if the backtest results are meaningful for me.

As always, thanks for sharing Marlyn.

Anonymous said...

Marlyn mensioned in one of her posts, "Excel laboratory".
http://filteringwallstreet.blogspot.com/search?q=%22Excel+laboratory%22

Does anyone know what Marlyn was doing with Excel Laboratory?