stats: inaccuracy problems

Let us know what features not currently in the PT4 should be added.

Moderators: WhiteRider, kraada, Flag_Hippo, morny, Moderators

stats: inaccuracy problems

Postby Fox128 » Sun Apr 08, 2012 1:29 pm

First I wasn't sure if I should make a thread about this, but I came to the conclusion that this is indeed really important stuff for me..

You made the first step by re-defining VPIP, but there is still a lot of inaccurate stats, some of which are really important to me

The two main problems:

1. Raising opportunities are currently defined as "facing an action to defend against..." for many stats.
This is the case for most stats that refer to 4bets, raise flop/turn/river etc.
The problem here is, that they are devided by def-opportunities.
4bets get devided by faced 3bets in some stats
raises get devided by faced bets etc.
While this seems to be logical at first, there is one big problem in terms of accuracy with this definition:

For folding or calling vs an action, this is no problem and outputs the exact correct stat... but not for the opportunity to raise (eg raise, 4bet, 5bet+ etc.)
Whenever a player faces an all-in, it still counts as an opportunity to raise against! which is false, as you obv can't reraise vs an all-in.

4Bet+ Preflop After Raising for example:

(cnt_p_4bet_after_raising / cnt_p_3bet_def_opp_when_open_raised) * 100

While for the most stats this little inaccury doesn't change too much and I don't care for them, I certainly care for 4bets. I think here the inaccuracy really has a big impact:
Whenever you face a 3bet all-in by a shortstack (or maybe you play shortstacked yourself), it lowers your 4bet stats which it should not...! This situation occurs a lot so it has a pretty huge impact on the stats, really making it kinda hard to rely on them (I tested this a while ago).

Also I already pointed this kind of thing out in the PT3 boards (for raise F/T/R bet) back then.



2. 4Bet vs 4Bet+ in general:
The second problem:
I'm not sure about this, as it is a little confusing, but:
4Bet+ means 4bet or 5bet or 6bet etc. right? So..
For example for "4Bet+ Preflop" the definition is "cnt_p_4bet / cnt_p_4bet_opp"
... so "cnt_p_4bet" etc. counts all the 5bets etc too?
(this is what confuses me, because it doesn't say so in the column descrptions)

If it is the case I don't like this at all. 4Bets should be a seperate category from 5bets!
:

An extreme example (but to get my point): 4Bet ranges often times are COMPLETELY different from 5bet ranges. There are people who 4bet bluff like 15% ranges but kinda chicken out when they are the one's facing a 4bet and 5bet just 2% or something like that so the stats would give me the impression that they do both with like 5-10% ranges, making this a useless stat to rely on for both situations.

You do have "stand-alone" 3bet-stats. Why don't you just have 4bet stats?
There even IS a 5bet+ stat too so I really don't see why 5bet+ should be included in every 4bet stat.

(For 5bet+ counting 5bets and 6bets+: This is a different story as 6bets occur VERY rarely anyway.)

But for 4bets... I just want them to count the 4BETS. Please!



For all these reasons the 4Bet+ Ratio stat for example is completely useless for me as it is now. =/



Please consider re-defining some of the default stats to make them more accurate. :(

...after all, this is still what PT is all about for the majority of users: Looking up numbers in your HUD while playing and making decisions with the help of them.
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Re: stats: inaccuracy problems

Postby Fox128 » Mon Apr 09, 2012 4:16 pm

I recommend banning spambots.
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Re: stats: inaccuracy problems

Postby Fox128 » Mon Apr 09, 2012 4:46 pm

Also:

How is Open 4bet+ calculated? I thought it was like COLD 4bet+ ing!? Seems I'm wrong but what is it than!?
I thought you havent done anything in the hand yet and face a 2bet and a 3bet+!

How is it possible I get results for MP (in 6max), where I could only possibly face a 2bet without me acting first!?

Also it gives me very odd results even exceeding my overall 4bet range for a big sample and can't be true for cold 4bets.
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Re: stats: inaccuracy problems

Postby kindevil » Tue Apr 10, 2012 7:48 am

Why nobody answers TS questions? It's been here 2,5 days already!
kindevil
 
Posts: 25
Joined: Thu Dec 29, 2011 4:42 pm

Re: stats: inaccuracy problems

Postby Barlin63 » Tue Apr 10, 2012 9:02 am

i agree that stats should be accurate as possible even if this means that we end up with more stats (5Bet, 6Bet etc). Maybe with the combined stats all together for the less professional player.

kindevil wrote:Why nobody answers TS questions? It's been here 2,5 days already!

I guess that Eastern was a day off or maybe two so give them a break. I am also looking forward to the swift release of a new Betas and hope that it is a bit more than only Zoom adjustments :D
Barlin63
 
Posts: 221
Joined: Sun Mar 25, 2012 5:44 pm

Re: stats: inaccuracy problems

Postby kraada » Tue Apr 10, 2012 3:30 pm

I apologize for taking so long to get back to you. I just wanted some time to research this in further depth for you.

Regarding (1) - we changed definitions (in the database) so that this should not happen. If you can give example hands where the opportunities aren't valid opportunities, we will make certain these get fixed.

Regarding (2) - We believe that 5bet and higher pots are exceedingly rare and that, barring extremely deep play, these definitions will work well for the vast majority of our customers. Regarding 4Bet+ ratio, that stat in particular requires that you have made the first raise in order to be counted - so the only way a 5bet is counted is in a strange multi-way spot (you raise, someone 3bets, another person 4bets and you 5bet) and by in the large when you 6bet or higher, you've already 4bet which already triggers this statistic as it is. Each hand can only trigger for this stat once, so 4betting and 6betting doesn't make it count double.

That said there is a 4Bet Preflop stat exactly - it is called "4Bet Preflop". We think that the + stats will work better for most people, but if you want to use this stat, please feel free to use this one - that is why we provided it in the default stats package. We tailored our HUD to what we thought the average user would find most useful - but it's configurable because we recognize that not everyone has the same needs.
kraada
Moderator
 
Posts: 54430
Joined: Wed Mar 05, 2008 2:32 am
Location: NY

Re: stats: inaccuracy problems

Postby Fox128 » Tue Apr 10, 2012 5:31 pm

(1) good to hear. If I still find errors I'll let you know.

(2) 5bet pots aren't that rare at all nowadays especially where some people also work with bluff 5bet all-ins vs wide 4bet ranges in LP etc. on higher stakes or go broke super light 2-4 handed.
Also this is actually about the OPPORTUNITY to 5bet which occurs a lot. So.. sorry, I have to disagree on that.

Ok I see you have a raise_3bet definition for exactly 4betting and "cnt_p_4bet(_opp)" counts in 4bet+s too.
--> you should update the description of some columns than (and maybe there are more). Right now it just says:

"Number of times player 4 bet preflop." (cnt_p_4bet) - used for the 4B+ stat
"Number of times player had the opportunity to 4 bet preflop." (cnt_p_4bet_opp) - used for the 4B+ stat
"Number of times player 4 bet preflop after having made the first raise." (cnt_p_4bet_after_raising) - used for the 4B+ after r stat --> which is also devided by 3bet def opp so 5bets really shouldn't be counted in here!! (make no sense to include 5bet++ here anyway as in the case of being first raiser and then facing a 3bet AND a cold 4bet one would be acting VERY differently).
And about the "rareness" of these situations. I can tell you that it's not THAT rare where I play and is pretty common for mid+stakes action and deep action!

(3) Please answer my question about Open 4Bet+ and some more questions about my awkward results:

I get totals of:

4Bet+ Ratio: 2.40
4B Range after RFI: 3.55 (custom stat: I can send a ticket if you want it, but it's actually the one from PT3 repository I think)
--> where comes the big difference from?

I see walks are discounted from the opportunities now but shouldnt that actually boost the 4Bet+ Ratio stat more (or am I making a logical mistake here.. hmm)

Edit: I think I see now that Ratio is based on PFR and 4B Range is based on RFI so the latter would be logically higher. I don't really like the idea of PFR though. PFR can be 3bets, cold 4bets etc. too...

Open 4Bet+: 5.85 (and 13.56 in MP in 6max. How!? How can there even be a result for this position!?)
-->seems very off how can this be even higher than the two others!?
The sample is 170k hands.

For reference/calculating:

4Bet PF = 7.99
4Bet+ PF = 9.19.. lol
Raise & 4Bet+ PF = 12.69

Raise First = 28.12
PFR = 18.87

Soo...
4bet(+) range after RFI should be ~3.57 (matches more with the custom stat)
4bet(+) range after PFR should be ~ 2.39 (matches more with the ratio stat)

I think there should be built in "after RFI" stats if they are that different.

I mean playing against a 3.6 (~like TT+,AK,AQs) range is different from playing against a 2.4 or less range (KK+,AK) after all!
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Re: stats: inaccuracy problems

Postby Fox128 » Tue Apr 10, 2012 6:23 pm

About (1)

Ok I filtered for hands where I faced an all-in 3bet and called and filtered UTG so there are only few hands to review..

Every 4bet-related stat is 0.00 obv.

Times/Opp:

4Bet PF: 0/7
4Bet+ PF: 0/2 (!) How can this be less for the same hands?
Raise & 4Bet+ PF: 0/7



Many of my 4bet opportunities look like this:

PartyGaming - $2 NL (6 max) - Holdem - 6 players
Hand converted by PokerTracker 4

SB: $208.62
BB: $71.58
Hero (UTG): $219.05
MP: $216.32
CO: $114.60
BTN: $241.36

SB posts SB $1.00, BB posts BB $2.00

Pre Flop: (pot: $3.00) Hero has ?:spade: ?:heart:

Hero raises to $6.00, fold, fold, fold, fold, BB raises to $71.58 and is all-in, Hero calls $65.58

--> yes I can 4bet here, because I have a larger stack. However, effectively I can only call as he is all-in and there are no players behind.







Also some more things/questions about 4Bet+ Ratio:

((cnt_pfr / (cnt_hands - cnt_walks)) * (cnt_p_4bet_after_raising / cnt_p_3bet_def_opp_when_open_raised)) * 100

1. If this is calculated using PFR instead of RFI. Why is it mixed with cnt_p_4bet_after_raising here (the definition of that stat says: "after making the FIRST raise") and opp_when_open_raised (self explanatory).
So what is this stat based on now or what is it supposed to do? Doesn't make logic sense to me.

2. If this includes 5bets+, why is it devided by 3bet_def_opp... makes no sense at all but I already said that it the post before so yea..
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Re: stats: inaccuracy problems

Postby kraada » Thu Apr 12, 2012 4:45 pm

We created the stat to mirror the existing 4Bet Range statistic used elsewhere in the poker world. We changed the name (as it's not really a range, it's just a ratio), but we did not change (or invent) this particular formula. Since it is such a popular formula, we did not want to alter it.

However, the reason we have the custom statistics interface is so that, if you want to use a version with RFI, you have the freedom and opportunity to do so.
kraada
Moderator
 
Posts: 54430
Joined: Wed Mar 05, 2008 2:32 am
Location: NY

Re: stats: inaccuracy problems

Postby Fox128 » Sat Apr 14, 2012 7:09 am

Well... Personally I don't think it's a good policy to include such inaccurate stats like 4Bet+ Ratio just because others, be it HM or whatever, use it. Why not use the 4Bet Range stat, which was invented by a PT3 user from this community (maybe polish it a little to make) as a built in default instead of that. But well, I guess I made my point clear and it's up to you.

I'd appreciate if you could also answer my other questions though:

1. Open 4Bet+
-Why do I get MP results for 6max here?
-Why is this number even higher than my results for 4bet range/ratio after raise? (I'm pretty sure I'm not cold 4betting (which is open-4betting right?) more than I do 4bet in general...)

2. Some of the column definitions:
(cnt_p_4bet), (cnt_p_4bet_opp), (cnt_p_4bet_after_raising),...
all of these are used for 4Bet+ STAT definitions so I'd assume they refer to 4bet PLUS themself. Is that the case?
If so, will you write that into the description of those in one of the next updates please?
It's pretty hard to build our own accurate custom stats, using existing columns, if the descriptions aren't complete / we don't know what they are actually doing.
Fox128
 
Posts: 56
Joined: Sun Mar 09, 2008 1:27 pm

Next

Return to PT4 Feature Requests

Who is online

Users browsing this forum: No registered users and 0 guests