Saturday, January 20, 2007

Filter Madness

Dogwood and I are currently working on some filters trying to determine what the best method for finding stocks at their bottom. I started this with my 4-day down method where I skip a day and look at lows. He uses a sequential method looking at closes (although I might get him to convert) and has had great success in 4 and 5 day increments. Go take a look - Dogwood is worth a read anyway.

So this morning I was working on my version of the filter looking at 6 and 7 low sequences when, for fun, I made a new, three line filter that said "give me stocks making a new 3 month low." And that would be a close at the three month low - no sequence or anything else - in other words a brute force approach to the problem.

There were only two things that amazed me about the back test results - first that there never was a day without a stock making a new three month low - in fact most days there were many of them and second - it returned 121% ROI. We're on the right track. It only had a 60% win percentage however. In an effort to tighten that up I offset the requirement by 15 days - that is I said give me all stocks that hit their 3 month low 15 days ago. That brought the win percentage up to 66% and I like that. It did lower the ROI a bit but that's OK since ROI in any given period is only for comparison purposes.

Now just to keep things on the straight edge I reversed both filters - for the first one I substituted new 3 month high and for the second, new 3 month high 15 days ago.

The win percentage was 47 and 55 respectively and the ROI's were 9.45 and 49.2. I'd like to say again for those who are still with me - if you learn nothing else from this Blog - inscribe this on your monitor - Buy the dips - Sell the rips.

I tried 4, 5 and 6 month lows and not surprising there are a lot of selections out of any category on any day. But the longer they go going down the more they go down. Still the returns on each of those in both win percentage and ROI-wise were better than the new high versions.

I'm going to work on this for a bit and see if I can't modify the 15 day offset 3 month low to look for something in that 15 day time period that will improve both win and ROI.



Anonymous said...

Three-line filter?

You getting lazy on me!? Ha!

Bullish Jim said...

Good stuff, Marlyn.

When you tested three month highs you backtested them as buys not shorts, right? It seems like it might be a good tool for finding short opportunities if one were so inclined.

I know I've asked you before but what software or website do you use to filter and backtest?


Nothing like a good filter.