New Software Announcement - SIXTH SENSE

[2easy]

[ QUOTE ]

What version of Postgres do you have? What version is your PGOLEDB.dll file (In your system directory).

-Ben

[/ QUOTE ]

8.03 and 1.0.0.20

[SixthSense]

[ QUOTE ]
[ QUOTE ]

What version of Postgres do you have? What version is your PGOLEDB.dll file (In your system directory).

-Ben

[/ QUOTE ]

8.03 and 1.0.0.20

[/ QUOTE ]

Those are the right versions. I'm kind of stumped here. Maybe try turning off all the databases, check them back one by one to see which causes the problem?

-Ben

[teddyFBI]

Would you consider adding a "PLayers with data" column to the player data fields?

It should also be added to the searching / custom table formatting window. I like how the search function can highlight any tables with avg VPIP > 30, for example, but it's a lot less useful if that is only an average for 2 of the 10 people who are at the table (b/c I don't have stats on the rist)...I'd like for one of the criteria i can search by to be "min # of players w/ stats." Or at the very least you should be able to let the user add that as a column in the main table-finding window.

[teddyFBI]

sigh -- so 36 hours later, the PTAGG process finally finished without any errors, and I was excited to take it for a test run. Sure enough, the search only took 30 seconds, compared to the 8-10 mins it usually took to query the 5 DBs I have it drawing from.

Only problem: around 50% of the stats were completely missing. Yep, a bunch of zero's across the board, even for players who I KNOW I had thousands of hands for.

Ever the experimenter, I decided to do some trouble-shooting -- i have SS on both my laptop and desktop. My desktop is the one on which I let it run the PTAGG feature (and the one causing me headaches). I have SS on my laptop running the query the old-fashioned non-PTAGG way, but both are drawing from exactly the same DBs (I just copied them from my desktop over to my laptop 2 days ago, so I know they have exactly the same stat base.

Sure enough, 95% + of players have stats on my laptop (the non-ptAGG querying way), whereas a ton of players simply show zero's across the board on my desktop (the PTAGG one). Furthermore, even for players that DO have stats on my PTAGG machine, the # of hands between the SS on my laptop and dekstop is completely different...sometimes the desktop shows more hands than the laptop, and sometimes vice versa...so it's not simply a matter of the PTAGG DB having aggregated all the hands from across the DBs...if that were the case, it would ALWAYS show a higher # of hands than the laptop (non-PTAGG method), but this isn't the case.

I thought that I could check this by opening up the PTAGG database in PT to check on specific players, but PT wouldn't let me open it like any other DB, giving me some VERSION_INFo error.

Don't even know where to start trouble-shooting this, other than checking here to see if anyone else is experiencing something similar...

[SixthSense]

[ QUOTE ]
Would you consider adding a "PLayers with data" column to the player data fields?

It should also be added to the searching / custom table formatting window. I like how the search function can highlight any tables with avg VPIP > 30, for example, but it's a lot less useful if that is only an average for 2 of the 10 people who are at the table (b/c I don't have stats on the rist)...I'd like for one of the criteria i can search by to be "min # of players w/ stats." Or at the very least you should be able to let the user add that as a column in the main table-finding window.

[/ QUOTE ]

Its already there. [img]/images/graemlins/smile.gif[/img] The field is called "NumData".

-Ben

[teddyFBI]

[SixthSense]

[ QUOTE ]
sigh -- so 36 hours later, the PTAGG process finally finished without any errors, and I was excited to take it for a test run. Sure enough, the search only took 30 seconds, compared to the 8-10 mins it usually took to query the 5 DBs I have it drawing from.

Only problem: around 50% of the stats were completely missing. Yep, a bunch of zero's across the board, even for players who I KNOW I had thousands of hands for.

Ever the experimenter, I decided to do some trouble-shooting -- i have SS on both my laptop and desktop. My desktop is the one on which I let it run the PTAGG feature (and the one causing me headaches). I have SS on my laptop running the query the old-fashioned non-PTAGG way, but both are drawing from exactly the same DBs (I just copied them from my desktop over to my laptop 2 days ago, so I know they have exactly the same stat base.

Sure enough, 95% + of players have stats on my laptop (the non-ptAGG querying way), whereas a ton of players simply show zero's across the board on my desktop (the PTAGG one). Furthermore, even for players that DO have stats on my PTAGG machine, the # of hands between the SS on my laptop and dekstop is completely different...sometimes the desktop shows more hands than the laptop, and sometimes vice versa...so it's not simply a matter of the PTAGG DB having aggregated all the hands from across the DBs...if that were the case, it would ALWAYS show a higher # of hands than the laptop (non-PTAGG method), but this isn't the case.

I thought that I could check this by opening up the PTAGG database in PT to check on specific players, but PT wouldn't let me open it like any other DB, giving me some VERSION_INFo error.

Don't even know where to start trouble-shooting this, other than checking here to see if anyone else is experiencing something similar...

[/ QUOTE ]

I'm pretty sure this is just a simple conversion bug in SS, so don't worry you won't have to re-import your PTAgg database.

In my PT Database I didn't have all types of games, and I'm thinking its just a conversion error from the SS game_type to the PT game_type. Run the steps below to get your list of game_types, then you can email them to me at ([email protected]) or PM them to me. Also let me know which limits you were having a problem with. Then I can fix it.

There are differences between the direct queries and the aggregated queries. The aggregated query is limited to games with certain # of players, and ONLY for games of the exact same game type (which is the proper way of doing it). The direct query I think just pulls all hands for that player from the first database they have data in. This is bad because its for any game type, any # of players at the table, and not aggregated across db's.

Off the top of my head, the # of players required is 8-10 for full seated games, and 5-6 for 6-max games.

I'm hoping everyone will switch over to PTAgg to make the problems with the direct queries a moot point. Once PTAgg is setup and completed its 1st scan, your SS stats should hopefully match up to your PAHud stats (assuming you use the same logic for # of players).

-Ben


** Steps for Extracting your Game Types **

Go to Start Menu / Programs / PostgreSQL 8.0 / PGAdmin III. Double click on the first server listed on the left, enter your Postgres password. Under "Databases" select "ptagg", an d then click the SQL button (looks like a red pencil with SQL below it). In the top half enter:

select * from game_types

and then select Query Menu / Execute. Then do a File/Export and export it to some file (on your desktop is easiest). Then please send that to me.

[SixthSense]

[ QUOTE ]
(snipped screenshot)

[/ QUOTE ]

See my reply above about how SS's direct queries and aggregated queries are different.

On your screenshots, I'm going to call the left side "direct" and the right side "aggregated".

For player BIGDIGOUT, I'm guessing he's played a lot of games either at other game types, or at this game type but short handed. The direct queries just get ALL the players hands, which is why you'd see 392 hands. On the aggregated side, we only see 110 hands. This is because that query looks for the exact same game type (which will narrow down hands), and only games with 5-6 players (narrows it down even more).

"lotus18" is showing no hands on the aggregated side. I think this is because you have data on him at other limits, or only for short handed games. Same for "bimbambusse" and "poker_shen".

"ipflivecp" and "cleaner97" has same pattern as BIGDIGOUT.

It is possible for the direct side to have FEWER hands. This would happen if it just happened that one of your first databases in order had a low # of hands. The direct mode quits searching the db's after it finds data in one of them.


The goal here is to get SS PTAgg mode to match PAHud. I'll probably have to fixup a conversion error, but I think it should be good after that.

It would be interesting to know what stats these players show in PAHud. And also what logic you use in PAHud to query for # of players. SS uses 8-10 players for 10-seaters, and 5-6 players for 6 seaters.

Thanks for putting the effort into helping us get this out of Beta!

-Ben

EDited: game time -> game type

[teddyFBI]

OK, I see why the Direct might sometimes show more hands than the aggregated DB, and sometimes less (which previously made no sense to me). Here's some more comment. As I mentioned in my email, I probably have some of the biggest DBs of people on this forum. If guys w/ DBs as robust as mine are still having issues w/ known players showing "zero's" it indicates that some rejigging of PTAGG's logic is required. I was active in early discussions for both Playerview and PAH, so I've been around the block with these issues, and wouls strongly suggest the following:

The biggest issue with your method I think is reflected in your statement that:
"The aggregated query is limited to games with certain # of players, and ONLY for games of the exact same game type (which is the proper way of doing it)" emphasis added [img]/images/graemlins/smile.gif[/img]

The above is incorrect, in my opinion.

1. Don't make the default setting "use hands from this limit only." This is important. I play 20/40 thru 100/200, for example. I might have 1,000 hands on a guy at 20/40 in my DB, but I don't want SS to think i've got no stats on the guy just b/c he happens to be sitting at 30/60. You need to make the default logic to use ALL the hands a user has on a player, regardless of what stakes they're at. Trust me on this one. Players don't play substantially differently between 30/60 and 50/100. And while players might play differently from .50/1 to 100/200, no players are going to be playing both those stakes...see what i'm getting at? No players are going to be playing stakes so different that they'll be playing drastically differently. Make the default treatment to use ALL of a player's hands, but you can provide a checkbox (like PAH does) to allow the user to specify 'only use hands from current limit'. But I think most ppl who actually use that box would be misguided in their efforts. Don't forget that most users are only going to have DBs with stakes that they typically play (and this is usually going to fall within a fairly narrow range). I doubt very much that there are ANY players in any of my DBs who play drastically differently within the stakes that I actually have PT data for. And even if some do play slightly differently b/t 30/60 and 100/200, this difference is vastly, vastly, vastly outweighed by the drawback of PTAGG not incorporating the vast majority of a player's stats in a query simply because they're at a slightly different stakes.
But wait, you might object: what if a successful 30/60 dude comes home wasted one night, and decides he's going to sit at 2/4 and play like a complete LAG maniac with a 95 VPIP for a few hundred hands. That's the most extreme example you could think of, and you woulnd't want that session to poison your PTAGG query. But think about it would require: Regular 30/60 players, like myself for example, won't even have any of those 2/4 hands in our DBs, so that LAG-maniac session is like it never even happened. And 2/4 regulars who might actually now have said solid 30/60 player in their DB as a LAG-maniac will never be running queries that will find that guy, b/c he'll presumably be back up at 30/60, and no 2/4 regular is going to be playing 30/60, or using PTAGG to query those tables. That is why the current PTAGG logic is incorrect, IMHO.

The above is the single bigest improvement you can make to PTAGG. I will go ahead and say, in fact, that it's imperative.

2. My player filter logic in PT is 2-6 and 7-10. BUT, I don't have any hands in my DBs with fewer than 4 players (or a negligible amount). What this means is that any time I sit at a table w/ between 2 and 6 players at it, PAH shows me the stats for those players from my DBs in hands that had 4 to 6 players, and at a table w/ 7-10 players I see stats from hands that had 7-10. PAH's system of filtering is close to perfect. Let the USER PICK what he wants his filters to be. Although if that's too hard, then your chosen filters of 5-6 and 8-10 are the ones I would have picked as the one-size-fit-all choice too.

If your goal is to get SS PTAGG mode to match PAHUD, then you're going to have to let the user select his filters of choice, b/c that's exactly what PAH allows.

The biggest thing, I stress, is the stakes-specific way PtAGG aggregates, which needs changing. There's a big difference between game TYPE (e.g. limit vs. No-Limit) and game level (aka stakes). The latter (game level) should not be distinguished by PTAGG, as I wrote above: just have it show ALL a player's hands (pursuant to the #-of-players filter, but irrespective of the stakes). I personally don't even think that you should worry about PTAGG aggregating LHE with NL stats. Yes, listen, I understand that players play differently at LHE than at NL, and I won't argue with someone who wants to say that they shouldn't be aggregated, but here is the opinion from someone who has been around the datamining block, and does play consistently at every stake from 20/40 to 100/200 and 5/10NL to 25/50NL >> I don't think stats need to be kept separate for LHE vs. NL -- there are relatively few players who play consistently at both LHE AND NLHE. A negligble number, at least. That being said, if it's an easy task for you to query only limit stats for limit tables, and NL stats for NL tables, then I don't think it will hurt. But I don't think it's worth the effort if it will take a while...there are more important fish to fry on this project, IMHO. The only people who would find this advantageous are those who have huge DBs of both LHE and NLHE, and who play both regularly...I am one of those, and even I don't think it's crucial.

[SixthSense]

Thanks Teddy, that makes a lot of sense.

I think I'll add an option to SS (which will be used in PTAgg), to query the players only based on # of players (using my defaults) and limit vs. no-limit (so regardless of stakes). This will be the default option. I'll add another option to just do that game-type/stakes. And finally an option to pull from all game-types/stakes.

I don't want to change the # of players queried if I can help it. Its some work, and doesn't have much payoff.

I think that should get SS 98% there.

-Ben