MM, thank you for the link!
I have been very busy, and I was not able to implement the Darvas method yet, but the Oddball system looked intriguing, and much easier to test.
As I didn't have historical data available (till now) I did a smal program to "forward test" it. (See more details below) When I have some free time I'll adapt it to test the historical data you provided.
This is what I did with the Oddbal System so far:
My program "forward tests". That means that it collects RT data and applies Oddball to it as it comes.
I run two independent tests, one for NYSE, and another for Nasdaq.
I use &ADVN and &ADVNQ (as numbers of advancing issues for NYSE and Nasdaq) and SPY and QQQ as the stocks that are bought and sold.
I keep data points each 15 minutes for archival purposes, but the percentage of change is calculated every minute or so against yesterday's value closest in time. (i.e. today's 10:12 &ADVN against yesterdays 10:12 &ADVN value)
I place trades as they present themselves, and not in restrospect. (Except for the first day) At first I followed the rule of not operating before 10AM but I broke it today as the % was way over 100%.
I added an arbitrary rule to avoid whipsaws: (and pattern day-trading) Only one trade per day. So far it has served well.
In the following EMails I'll post data and trades for the last 6 days. (All I have so far...)