Data mining - logistics question

[SofaCoach]

I just started experimenting with "data mining" observed hands a couple days ago. I am trying to figure out the most effective way to do this and run into one big dilemma. I've read all I can find on the subject but can't find an answer.

The dilemma is this: Due to rapid file growth size many recommended putting observed hands in a seperate PT db. OK, I did that. Observed hands go into one db, hands from my table go into my regular db. Now, which db do you use in your gametime window? I want to take advantage of all the observed data, but also want constantly updating data on the players I am currently facing?

How do you guys resolve this?

[kevyk]

I think I've figured out a foolproof solution to this dilemma:

Given that you really only need data on the players you are actually playing against, and that the most relevant data for doing game selection are preflop looseness and aggression (two stats which rapidly converge), start datamining 4 tables a few hours before you plan to play.

You will end up with an approximately 300-hand sample on most of the players. Pick the loosest, most passive table you can, and mop the floor with your opponents!

Sure, you don't get great statistics on opponents' play on later streets, but you can gauge this by simple observation. I really don't think it's possible to compile a statistically significant database for all the players on Party.

[MS Sunshine]

I start a new DB at the start of each month, or when it hits 200K. I do one last export of the old DB to the notes.txt files. I setup the new DB as the default and direct playerview to new DB. For the next week I do no more export of notes. This leaves me with the old data in the note file and the new data displayed by playerview.

MS Sunshine