Question about automated data mining, possibly distributed...

[ReMMy]

Posted under Internet as well, maybe this is more appropriate?

I'm wondering about a couple things.

1. Would a program that simply keeps opening different tables violate Party's T+C's? Basically it would allow you to continuously gather hands when away from the computer without having to worry about the tables closing.

What about going 1 step farther and collecting those hands in one place and then giving users of the program the ability to request hands be sent to them that other people gathered.

2. Is there any interest in a program that does the above?

Thanks,
ReMMy

[Dazarath]

There's a lot of interest in a program that would automate datamining. There's already someone working on one, but I don't know what happened to that project. It was called "Partymine", if you want to search for the original post.

[ReMMy]

We're getting ready to release a distributed data mining client in the next week or so.

Anyone have any feedback as to what you'd like to see?

Basically you can select the limit you want a certain client to watch and it will keep switching between different tables every 10 minutes or so which will allow you to run it overnight.

Hands will then be transmitted back to a central server where you will be given credit for the # of hands submitted. Sending 100 hands at a certain limit would allow you to request 200 or 300 hands back at that limit mined by someone else.

Is this what you guys are looking for? Any feedback would be appreciated. We've already got all the components in place, just need to piece it together...

Thanks,

[lawrence]

Can you datamine by just opening the client and tables, and not being logged in?

[Orpheus]

IIRC, you can only have one table open at a time when you are not logged in. This would dramatically slow down your data acquisition

[Dazarath]

[ QUOTE ]
We're getting ready to release a distributed data mining client in the next week or so.

Anyone have any feedback as to what you'd like to see?

Basically you can select the limit you want a certain client to watch and it will keep switching between different tables every 10 minutes or so which will allow you to run it overnight.

Hands will then be transmitted back to a central server where you will be given credit for the # of hands submitted. Sending 100 hands at a certain limit would allow you to request 200 or 300 hands back at that limit mined by someone else.

Is this what you guys are looking for? Any feedback would be appreciated. We've already got all the components in place, just need to piece it together...

Thanks,

[/ QUOTE ]

Here's a couple suggestions I have. First, it'd be nice if there was an option to select how many players there need to be for the program to continue mining. I'm not sure what the purpose of having the program switch every 10 minutes is, but if a table kind of breaks apart and only 5 people are left, the data isn't as useful to someone who plays full ring games. Also, the ability to use multiple skins would be nice. I'd run all the party skins if possible.

[TheTROLL]

Poker Edge (http://www.poker-edge.com ) provides the same sort of result without you having to run any datamining yourself, surely? What is the advantage of this approach?

[ReMMy]

Poker edge doesn't provide raw hand histories for use in PokerTracker.

[ReMMy]

The program will only select tables w/ more than 6 people for full tables, more than 5 for 9max and more than 3 for 6max. The purpose of switching regularly is so you don't get a ton of stats on a few people but rather a decent amount of stats on lots of people. We could make things like this customizable if need be.

It will run on all the skins and each skin will have the ability to watch a different limit or game type.

[TheTROLL]

That's true enough, it's only intended as a realtime tool, I guess. I don't think I've ever used an individual observed hand history in PT - only ever thought of datamining as a source of data for GT+. I'm interested in how you would use those and what benefit they would provide over looking at players' aggregated stats?

[KJ o]

Automated datamining of SnG's would be great. Those would have to be complete though. Just getting ten minutes per table i sessentially worthless for a SnG.

[Entity]

There was some great software called Partymine that did this a while ago, but it apparently has dropped off the face of the earth (beta expired, no contact from author). I'm looking forward to another project that does this well.

Rob

[Mizzles]

There is ALOT of interest in a program like this. Sounds great. There was once apon a time a program called partymine which was supposed to be released ages ago but it turned into "vaporware"

[xLukex]

[ QUOTE ]
There was some great software called Partymine that did this a while ago, but it apparently has dropped off the face of the earth (beta expired, no contact from author). I'm looking forward to another project that does this well.

Rob

[/ QUOTE ]

If you still have the program, why not let someone else work on it?

And we should petition the author to come back to planet Earth.

[tor]

[ QUOTE ]
Anyone have any feedback as to what you'd like to see?

[/ QUOTE ]

How are you guys handling data poisoning, hand history collisions and the like? What's to keep someone from reporting falsified hand histories?

[Orpheus]

Man, I *SO* wish that this point wasn't as insightful as it is -- but I know too many examples from other collaborative efforts in the past decade.

Ladies and Gentlemen, the next wave in thuggery....

[illunious]

[ QUOTE ]
distributed data mining client
Sending 100 hands at a certain limit would allow you to request 200 or 300 hands back at that limit mined by someone else.



[/ QUOTE ]

That's really really really cool...

[Prod1gy]

Its only been a month people. Settle down. PartyMine is coming, hope to release in the next week or so.

Oh yea and PartyMine is client only, we don't pipe back hand to a central server to profit off it later.

[Prod1gy]

Why was this vaporware? The last beta was about a month ago. Software turns into vaporware now after a month? Wow.

[bostonvex]

"we don't pipe back hand to a central server to profit off it later"

There are 2 approaches:
1) Distributed Datamining
2) Datamining

I believe both methods have their pros and cons.
Distributed datamining would give users access to X times as many hands as they would get if they datamined by themselves.
With the distribited dataminer it is a community effort rather than a "I only mine for myself" effort. So you get so much more for so much less.
The distributed dataminer will have no fee or subscription, as long as you contribute you can benefit from it.