|Posted: March 19, 2013, 12:19 pm - IP Logged|
If people find it helpful, there's more than can be done with it. I can add more filters in there like assuming a position. Say if you have a system that typically nails Position 1. You could pick that number for position 1 and it would run the pairs for the other two positions.
Some of the more interesting testing has been when outs are applied to that list. For example, if any given number (for each position) is under the out average for the pasy 180 days, add 10% of the total column to itself. Then look and see if a number has been out more than it's previous out maximum over the last 180 days. If that happens, add 20% to the total column back to itself. This method turns the Total column into more of a ranking system. I've done light testing and it typically produces several wins a week. Sometimes the hit is in the first 5 results, others times its number 15 in the list. So it's just a question as to how many sets you want to play.
I've also been toying with the idea of generating data where the code will find how well a number peformed, week by week, from the previous year. That will give some data points like what we could expect out of a number, an average on a hot week vs a cold week. Overall performance for the year. Stuff like that. Taking it a step further, the data could be generated for previous years to get a sense of trending for each number on each position. So year by year (or month by month), we would be able to see what numbers might be considered hot or cold. All of that could be applied to the pair ranking system. It's a bit of work but could be interesting to see how well it works. I'm just not sure how I would present the data. Maybe just go the obvious route and make a chart for it similar to what I have on my site already.
Anyone else have ideas for filters, different data points, or anything I'm not thinking of?