Next gen SnG tracker: brainstorm?

[eastbay]

I think poker tracker is a bit of an "almost there" tool for SnG statistics, but I think it's really kind of a first cut tool that could be improved immensely, especially for SnG players. So before I give some of my ideas, what would your dream SnG tracker do?

eastbay

[SuitedSixes]

I was just thinking about this today, actually. I think if it had te same functunality for SNGs that it does for ring games it would be much more useful.

For ring games it does a much better job of time of day and day of week analysis. There is no reason that it can't do the same for SNGs. It also computes SD for ring games.

I would like it to be able to replay entire tournaments, not just individual hands.

It should allow you to categorize players just as it does for ring games. It would be nice to sit at a table and have exported notes what a player's V$IP is in early rounds, hands raised with in later rounds, and see what the player's ITM or average place of finish is.

[Big Limpin']

I'd call it "Copernicus". A background app to track action from multiple tables/screen. All data is streamed to a database.

It observes all aspects of each play, every hand.
-bet sizes, all chip action
-every action's cadence... 4.5 sceonds...0.8 seconds...
-classifies all bet sequences into "genres",

some sort of algrortitm to rank each plays action with respect to Bring in/limp/pf raise....what flopped....next action....and so forth.

It will genrate huge amounts of data, queryable at a later time (and in my dreams, in real time). Any situation can be matched to previous examles of similar cicumstances,

Usefullness is limited to the intuitiveness of the user.

But now we're gettign into bot territory.

[Jman28]

I think you may be onto something with tracking bet sizes in NL.

I think most players unknowingly vary bet amounts between bluffs, semibluffs, and value bets.

-Jman28

[El Maximo]

They main improvements I would like to see is being able to break the stats out by number of players. Being able to see PFR%, VP$IP, Hands raised with, and a few others based on 7-10,4-6, and 3-2 would be useful. Than having an autorate feature that could be used in combination with this. Being able to overlay these features like Playerview would be excellent also.

[The Yugoslavian]

It should have you ICM calc at all times on your current equity. It could also show what your equity would be due to winning an allin either by pushing or calling a push.

It should also show you statistics vs. push hands shorthanded and how likely your hand is to be best with an open push given the number of players left to act.

It should smack you upside the head if you ever fold AA preflop. Better yet, it should lock up your PP account and not let you play poker ever again.

It should also occasionally show a pop-up at a weakish/tight table that says 'Strassa would push here with any two baby!!!' There could also be a pop up for 'who do you think you are, Gigabet??' when coming over the top of a bet with a seemingly marginal holding.

Oh, and to save Dali some energy it could automatically type in some obscenities and whining every time you get sucked out on. (And a 'Danger Mr. Robinson' when you're using your last few thousand $$ to play the 1k steps).

Yugoslav /images/graemlins/grin.gif

[Big Limpin']

[ QUOTE ]

I think most players unknowingly vary bet amounts between bluffs, semibluffs, and value bets.
-Jman28

[/ QUOTE ]

I think they (we) do also. Unwittingly. And the absolute key, KEY, i.m.o. is that they will always/usually play similar hands in a similar fashion.

A) Semi-bluffs
B) 2pr value bet
C) Protectionistic bet of TPbadK
**************
1) Bets the pot to the $
2) Bets 67% of pot
3) EP river leads in 10%

Connect the dots type thing, you know? Of course this is very crude, the above 3 factors. And corresponding type of bet pattern
will be somewhat player dependant, like a fingerprint.


Exploitable? Sure. But:

I would think long long long time would go by (years?) before your database was giving you a good signal-to-noise ratio. For a specific player. And his bet pattern so far in the hand. Pot odds of every call, every street. Thats alot of hands.

Just to get a, say, 80% read on what type of hand the villian is likely to hold.

Sub-calaculate chip EV in pot of your hand (known) and villian (educated guess). Re-query database for villains likely response to next streets betting pattern. Determine how to pull most chips out of him without folding him if you have likely got the best hand. Or how to keep the pot small, will he give free card if you need to hit an out?

This is the foundations of a bot. And thats a whole 'nother story. I have zero proof, but i bet if there arent already bots out there, someday there will be. And while i would think SnG bubble time needs a human "feel" to succeed, and in that sence, ring-bots would probably kill off that action before tacking SnG poker. But...who knows, eh?

Maybe i should dust up on my computer programming skills /images/graemlins/tongue.gif

P.S. Just kidding moderator! Dont censor/ban me please /images/graemlins/crazy.gif

[viennagreen]

yes-- breaking stats down by number of players is the most important feature in my book--- as it is now, PFR% and VP$IP stats are almost useless (except for stats particular to Rounds 1 and 2)

[johnnybeef]

allow me to import ub files

[Marcotte]

[ QUOTE ]
yes-- breaking stats down by number of players is the most important feature in my book--- as it is now, PFR% and VP$IP stats are almost useless (except for stats particular to Rounds 1 and 2)

[/ QUOTE ]

Breaking it down by # players is important, but it should also break it down by stack size:big blind ratio. You should be able to pull data based on either or both of these factors. eg: PFR% with 6 players left; or PFR% with 10xBB; or PFR% with 6 players left and <10BB. Of course each of these queries should accept ranges (<10BB, 7+ players). When using both, you might not have enough data on a particular player to be of any use (exception for $200 players with 1000+ tournies) but the data should help Hero identify leaks.

Also, it should track every hand shown or shown down by your opponents (but I guess that's obvious).

[(my name it is) Sam Hall]

[ QUOTE ]
Breaking it down by # players is important, but it should also break it down by stack size:big blind ratio.

[/ QUOTE ]

I'd like to expand this idea to say any program should be based on dimensionless parameters like this any time it is at all possible. Even if the quantity is a bit of a stretch, it is probably better than the raw numbers. I think "bet size/stack size" and "bet size/BB" would be useful quantities, which one depending on situation. Number of players is dimensionless (right?) so it could end up in those quantities too.

Sam

[eastbay]

[ QUOTE ]
[ QUOTE ]
Breaking it down by # players is important, but it should also break it down by stack size:big blind ratio.

[/ QUOTE ]

I'd like to expand this idea to say any program should be based on dimensionless parameters like this any time it is at all possible. Even if the quantity is a bit of a stretch, it is probably better than the raw numbers. I think "bet size/stack size" and "bet size/BB" would be useful quantities, which one depending on situation. Number of players is dimensionless (right?) so it could end up in those quantities too.

Sam

[/ QUOTE ]

You must be in an engineering program at Berkeley. And no, # players is not dimensionless (it has units of # people, of course), but fraction of players remainig is one way to nondimensionalize it.

eastbay

[(my name it is) Sam Hall]

[ QUOTE ]
You must be in an engineering program at Berkeley.

[/ QUOTE ]

Nice read. I am also a demonstration that smart people make terrible poker players until they quit outsmarting themselves.

[ QUOTE ]
And no, # players is not dimensionless (it has units of # people, of course), but fraction of players remainig is one way to nondimensionalize it.

eastbay

[/ QUOTE ]

OK, well, I won't argue that point on the grounds of irrelevance, but now I have more questions...

Are we only using this for 10-player STT's, which is all I ever play and pretty much the only thing anyone discusses here? If not, does anybody have an educated opinion as to whether or not players behave similarly in other tournament structures? I'm starting to think any calc involving the number of players has to be specific to 10-player STT's because of how players act in response to how long they've been playing, how many people they've seen go out, how close they are to money, and so on.

Sam