I was working with an older filter last night and I came up with a new twist - 4 sequential days with a lower low each day with the last day's low of the sequence making a new 20-day low. The other days can also be making new 20 day lows but it doesn't matter. This changes the emphasis from the close to the low which actually makes more sense than using the close.
The output of this filter looks like this -
And yes I did test 5, 6, and 7 sequential lower low days with mixed results - mixed because the more stress you put on the filter the fewer hits it gets.
I tested this against three periods - the one I've been using lately, a more up to date period and that period last March through June when the market was going down. The results for that test were poor as was expected so this is another filter that can't find solid plays in a down market. But since the market is always up - no matter.
Anyway my two positive test periods yielded the following results:
9/01-12/22 - 71% win rate,169.53% ROI, 8% net change over 30 days
10/06-02/07 - 63% win rate, 113.51% ROI, 5% net change over 30 days
These are not bad results and as you can see we are entering a weaker time in the stock market. But for what it is worth the three top selections from the filter for last week were: KBR, DBRN, and PDLI whose chart is above.
I said last week that I would be looking at the possibility of using a filter or two to attempt to predict the market and I have also obtained mixed results. Mixed in the fact that yes you can use a filter in this way but you already know the answer before the filter answers the question. I.E. there is no real forecasting ability. At least none that I've been able to discern. I'll keep looking and if I find a filter that seems to have some ability in this regard I will share it with you.