Textual exploitation stats for short NL, comment & discuss

[kraada]

I can believe it . . . I'm looking forward to setting up my own 1TB 7200.11 as ext4 and seeing what sort of speed boosts I can get

[PeteX]

Took around 4 hours to format the 1,5T HD.

I created the new analysis database like this:
- Created a d:\database-folder (D-drive is the whole new 1,5T volume)
- Using pgAdmin III, created a new tablespace "pokerdata" and set location as "d:\database"
- Created a new empty database into default tablespace with PT3 (just to get the schema)
- Created a new "ShortNLAnalysis"-database using the just created empty PT3 database as a template and setting "pokerdata" as the tablespace.
- Created a new ShortNLAnalysis-database to PT3 and mapping it to the existing database that was just created.

And then I started importing the ~85k handhistory files. In the beginning PT3 was importing around ~266h/s and said ~35h being the remaining time. This morning it was importing around ~155h/s and I think it was something like ~58h time remaining.

So, I guess it'll be few days to go before I can start running queries...

[bloodndef]

A question for ya,

How can we find the top/bottom 10% or 25% of a stat?

Take VPIP for example: how do I create a report that shows the players with the highest 10% VPIP? And can we have a stat reference this calculation?

Thank you for shedding the light on this.

[PeteX]

A question for ya,

How can we find the top/bottom 10% or 25% of a stat?

Take VPIP for example: how do I create a report that shows the players with the highest 10% VPIP? And can we have a stat reference this calculation?

Thank you for shedding the light on this.

First of all, I don't use PT3 custom reports, I use pgAdmin III and SQL.

But it goes roughly like this:
a) I create a query that retrieves a list of players and their VPIP values, and here I would use a HAVING-filter to get only players that have, say, >500 hands because that certainly would be enough to have a stable VPIP value for a player. On the query I have sorting done descending by VPIP-value so that the highest values are on the top.
b) Let's say I get back 1000 rows, each having the VPIP value for a single player. Of this 1000 I obv calculate that top 10% = 100 players and top 25% = 250 players.
c) Then I modify my query a bit, adding LIMIT 100 at the and and running the query again. Now I have 100 rows returned and the LAST ROW now shows me the LOWEST VPIP that the TOP-10% players have. Let's say that it is 40% -> Now I know that 40% is the threshold that separates the top-10% players from the other players. And then repeat the same with 25.

Once I have those threshold values figured out, I use them in the PT3 custom stats to set the display thresholds for my mnemonic stats.

Kraada might respond your particular questions regarding PT3 reporting, but that's how I do it manually.

[WhiteRider]

There is a way to do this more visually using PT3 reports and Excel.
You can create a Custom Report in the Holdem Cash Player Statistics section and add the stat that you want to view, along with the player name.
Then in the report Filters window, uncheck 'filter to active player' so that all players are shown, then add a filter for #Hands# > 1000 to limit the report to players with over 1000 hands.
Now when you run the report you will see the stat value for those players.
You can then Expand and Export the report, and open it in Excel.
If you sort it in Excel by the the value of the stat column then you will get all the players' values for that stat in order, and then if you graph that you can see where the bulk of the values lie and pick out what range you want to use.

[bloodndef]

Thank you!

There is a way to do this more visually using PT3 reports and Excel.

This is how I've been doing it, but with OpenOffice. It works fine, just wondering if PT3 had a filter I was missing or something.

Thanks again.

[PeteX]

Well, ~30mil short NL hands imported in ~99 hours. Now doing the housekeeping and the db has been clustering for 43 hours now...

[kraada]

30 million hands is going to take a while . . . I hope you have plenty of extra hard drive space, or that's going to make the cluster take even longer.

[PeteX]

30 million hands is going to take a while . . . I hope you have plenty of extra hard drive space, or that's going to make the cluster take even longer.

Well, still over 1,2T free on my new HD... I'd say that falls into the "plenty" category...

[kraada]

That would most certainly qualify

The only reason I asked is that if you don't have as much free space as the size of the database, the cluster can take 3-5x longer. 30M full ring hands would be ~300G, so probably 180-200G for short handed, and thus it's not an entirely unreasonable question to make sure you have 300G+ free.