PDA

View Full Version : Fold (Total) vs. Steal calculated wrong?



SuhtPaska
05-10-2012, 04:18 AM
I have version 1.12.10 (16 Apr). Problem appears in HUD & popup, real time & in reviewer, Cash & tourney.

Fold (Total) vs. Steal is calculated incorrectly (or just with strange logic) when sample size gets bigger. Under 100 hands it still shows right but after that it starts to weight towards Fold BB vs. Steal. Should it be this way?

In my mind:
SB (Fold+Call+3bet) Vs. Steal = 100%
BB (Fold+Call+3bet) Vs. Steal = 100%
TOT (Fold+Call+3bet) Vs. Steal = 100%

Where TOT is SB +BB.

For example 1 player with 3k sample size has stats Vs. Steal:

Pos. Fold Call 3bet
SB 88% 6% 6%
BB 81% 16% 3%
TOT 81% 12% 5%

So
SB line is 88% + 6% + 6% = 100%
and also BB = 100%

But
TOT line is 81% + 12% + 5% = 98%

Where did the 2% go? Should not Fold_(Total)% be somewhere between Fold_(SB)% and Fold_(BB)%? (Now it is same as Fold_(BB)%)

I have checked that the Call_TOT is calculated right from the Call_(SB+BB) and 3bet_TOT is also calculated right from the 3bet_(SB+BB). So the problem is only in the value of Fold_(Total)_vs._Steal.

Short story told long. Sorry.

And thank you for the fast replay for my last question.

Patvs
05-11-2012, 12:05 AM
Update to: http://www.holdemmanager.com/Downloads/HmUpdate_Release_1.12.10-04May2012.exe

The assumption that
SB (Fold+Call+3bet) Vs. Steal = 100%
BB (Fold+Call+3bet) Vs. Steal = 100%
TOT (Fold+Call+3bet) Vs. Steal = 100%

is false.

With a larger sample size, scenario's will occur where for example the BB doesn't have the option to call/3bet.
BTN steals. SB 3bets.
Now the BB can only: Fold, and 4bet. He can also call, but it's not the same as 'calling a steal'. He'll be calling a 3bet, which is something different.

As a result, he'll be always able to fold... but he won't always be able to 'call a steal' and 3bet, thus the stats are NOT based on the same sample size.

SuhtPaska
05-11-2012, 02:31 AM
That is very true. Did not think those scenarios. But then... :p

with my logic those scenario's (4bet or Cold_Call_3bet) should also corrupt SB_ & BB_vs_Steal stats, or?

But (SB & BB) are always 100%! (as Total_vs_Steal is not). So it seems that Total vs. Steal is calculated taken into account those extra scenario's but SB or BB not?

For example player stats again vs. Steal:

------ FOLD ------- CALL ------- 3BET
SB _ 91% (23) _ 00% (23) _ 09% (23)
BB _ 59% (39) _ 21% (39) _ 21% (39)
TO _ 68% (62) _ 13% (62) _ 16% (62)

So 3% is "missing" from Fold_Total_vs_Steal stat. If these were translated into pcs it would show that 2 (cases/pcs) is missing.

Ok, there could be an extra column after 3Bet for those special scenario's but it would show in this case:

------ 4bet/etc.
SB _ 00% (23)
BB _ 00% (39)
TO _ 03% (62)

So my question is still: 0%+0%=3%?

How come it does not show anything in SB or BB rows. In my mind it should be something like 0%+3%=3% and the 3% would be taken from Fold_BB_vs_Steal stats.

I am probably totally wrong again, but as said there are no stupid question

- only stupid questiners :rolleyes:

Patvs
05-11-2012, 10:38 PM
Do you have a screenshot of a popup that shows all the sample sizes of the number of hands are the name, something that looks like:

SB _ 91% (23) _ 00% (23) _ 09% (23)
BB _ 59% (39) _ 21% (39) _ 21% (39)
TO _ 68% (62) _ 13% (62) _ 16% (62)


Another scenario:
Hands exist where there was no SB.

So the number of opportunities is the same, but it's based on a different number of hands.
For example

SB _ 91% (23) _ 00% (23) _ 09% (23) (23 opportunities out of 100 hands)
BB _ 59% (39) _ 21% (39) _ 21% (39) (39 opportunities out of 102 hands)
TO _ 68% (62) _ 13% (62) _ 16% (62) --> thus this affects the TOT. which will use 62 (23 +39) opportunities out of 102 hands, but this doesn't take into account the 23 opportunities for the small blind weren't actually based on 102 hands.

SuhtPaska
05-12-2012, 06:12 AM
I almost get what you are saying but it might be just a bit too early morning. I get the idea of missing SB but I think it is already in the numbers.

Those numbers (23) (39) (62) are actual times/pcs (already including missing SB, or?) and the percentages in front of them are calculated based on those times/pcs (original attachment at the end). So I am not sure if it is anymore mathematically "legal" to transform/unify (dividing per 100/102 hands) pieces into percentages (23times=>23%) because then there would be calculations as percentages out of percentages?

But even if those original percentages were unified with 100 (SB) or 102 (BB) and Total row would be counted based on out of 102 hands (why BB?). Should it not also affect other Total columns than only Fold_Total_vs_steal? So Call and 3bet would also be smaller because they are unified by 102?

But the problem always appears only in the Fold_Total_vs_Steal. It would only make sense it Fold was divided by 102 (based on bb hands) and call/3bet was divided by 100 (SB hands). Not very logical.

This is the original table again. Every box has (times or pcs)/(opportunities)=(times%). So:

----|----- FOLD -------|---- CALL ------|---- 3BET -------|
SB | (21)/(23)=91% | (0)/(23)=00% | (2)/(23)=09% |
BB | (23)/(39)=59% | (8)/(39)=21% | (8)/(39)=21% | => 21% is actually 20.5%
TO | (42)/(62)=68% | (8)/(62)=13% | (10)/(62)=16%|

My opinion the problem is the 2 times/pcs missing in FOLD_Total box (should be 44/62=71% which would be logical because SB is 21pcs and BB is 23pcs).

For some reason 2 times/pcs are not counted as folds in Total. 4bet/Cold_call_3bet theorem would make sense. But then again they (4bet/Cold_call_3bet) are counted as folds in FOLD_SB or FOLD_BB. That does not make sense!

I am SO confused :rolleyes:

Original from HEM1. Just in different order.

Patvs
05-13-2012, 12:00 AM
I've written this up as a bugticket: HMONE-78, after noticing in HM2 the numbers do always add up to 100%.

In HM1 the TOT numbers usually only add up to 97%.

Zabagad
01-05-2013, 08:31 AM
I've written this up as a bugticket: HMONE-78

And when will somebody start to look for this bug?

Patvs
01-05-2013, 08:31 PM
This ticket is already closed. Issue was fixed July 10, 2012.
Update http://www.holdemmanager.com/Downloads/HmUpdate_Release_1.13.01_28Dec2012.exe (and try reimporting all your hands into a new database)

Zabagad
01-06-2013, 11:04 AM
This ticket is already closed. Issue was fixed July 10, 2012.
Update ... Release_1.13.01_28Dec2012.exe

Thank you Patvs,
but if I follow the official links on this side to the latest HM1 version - I reach HM1 Version 1.12.10b 2012-06-09
and not the 1.13.01 you mentioned.
Can you give me the link to the "real" latest Update section?
Not only the link you already gave me (this was working) - I mean the link to the folder where all the Updates are and will be...

TY again and in advance
Bye
ZAbagad

Patvs
01-06-2013, 08:42 PM
Last HM1 build is: http://www.holdemmanager.com/Downloads/HmUpdate_Release_1.13.01_28Dec2012.exe
There is no real HM1 thread with the latest HM1 internal/beta builds anymore.

Sometimes they're post here: http://forums.holdemmanager.com/releases/