Serious Bug in Poker Tracker w/ BB/100 or BB/hr

[SamJack]

I discovered a serious bug in Poker Tracker with calculating BB/hr and BB/100 calculations for multiple level played. (e.g. $2/$4 and $3/$6).

See the thread of e-mail I've had with Poker Tracker support. Let me know if I'm confused or if you agree with me.

My first e-mail.


Hello, I've found the product very useful. so Thanks,
However, I've ran into an issue which I think is a bug.
Look at the attached picture. You will note that the 'Toal Amount Won is -$26.00 (loss), however, the BB/100 Hands shows a positive +$0.10. Having a loss and having a positive BB/100 does not make sense.

I'm not sure how you come up with the weighted average BB for the total row, but it needs to be the following.

(#hand for level 1/# toal hands) * BB for level 1 + (#hand for level 2/#total hands) * BB for level 2 + ...

Using that formula give weighted aveage BB of 4.272205 and gives BB/100 of -0.11 (loss) which makes sense.

I hope this can be included in you next patch.
Thanks for such an excellent product.

Sam


My follow up

Addition to my earlier e-mail.

It looks like PokerTracker is taking the weighted average of the BB/100 which gives wrong result.

As noted in the earlier email. It needs to take the weighted average of the BB itself then divided by #hands/100

thanks,

Sam


Their response

Hi Sam,

The reason for the positive BB/100 or BB/Hr. is because there is no weighting on that field at all, it's simply a the (BB won from each level / total hands) / 100. I had a weighted number in version 1 and received so many emails questioning it (especially the BB/Hr. because the number didn't always match on each tab because of how the data is displayed) that I removed the weighting in v2 and modeled it after the way PokerStat did the calculations.


My response to their reponse

Hi, looking at the numbers, it doesn't seems like "(BB won from each level / total hands) / 100" as you say. The number seems to be weighted average of the BB/100 from each of the level.

This however is not the right way to take the weighted aveage. You need to get the weighted average of the BB itself then divide by totalhands/100.

I'm not sure what the complain from V1 was, but right now, the tool is showing Positive BB/100 when my winning is Negative. This cannot be right by any account.

I would appreciate if you could look into this further, especially since the answer you provided does not match up with the behavior of the program.
Thanks,

Sam


This is a bug! Can you guys help me convince Poker Tracker?
If you aren't certain what I mean, see the screen shot below. I know I've been sucking big time.

http://www.geocities.com/pakman_69/player-summary.JPG

Sam Jack

[NoChance]

I'm not quite sure what you are trying to say.

If you are saying the BB/HR should be more for a $2/$4 table then a $1/$2 table then I disagree.

Example: If you have 1000 hands at $2/$4 and 1000 hands at $1/$2 then the BB/HR for the two should simply be averaged together because the same number of hands were played at each level. It should not be weighted higher for the $2/$4 table simply because it is a higher limit. BB/HR is BB/HR regardless of the stakes.

If you are saying that you have 1000 hands at $2/$4 and 3000 hands at $1/$2 and the program simply averages the two BB/HR numbers together, then I agree with you.

Example: In this case, the correct weight would be 25% of the $2/$4 BB per HR and 75% of the $1/$2 BB per hour because 75% of you hands were played on the smaller table.

I hope this makes sense.

[SamJack]

I think you miss understand. I'm talking about being weighted based on the number of hand played @ each level.

In short, When I have a total loss, my rolled up BB/100 should not be a positive #.

SamJack

[tiltboy]

Since this was intentionally done by the programmer wouldn't it more correctly be called a feature? /images/graemlins/laugh.gif

Seriously, is it really that big of a problem? You get an exact measure of your BB/hr or per 100 hands FOR EACH LEVEL in PT. Isn't your average BB/hr or per 100 hands pretty meaningless?

[J.R.]

Why couldn't you have a positive bb/100 rate at a low limt, and have played at a higher limit and lost, which brings you overall results into negative teritory but still yields a positive bb/100 result.

[SamJack]

I am a software engineer myself. And yes, sometimes we claim that the defects in our products are 'working as designed'. In this case it is simply wrong.

consider the following

I play 100 hands at $1/$2 and lose $100 giving -50BB/100.
I play 1000 hands at $10/$20 and and win $1000 giving +5BB/100
Note that I have a net winning of $900

If you simply take the average as you say and as poker tracker support incorrectly claims their software does, you get.
-45BB/100 (that's negative) while having $900 in winning.

If you take the weighted average of the BB/100's which is still wrong but is what the software is actually doing, you get.
(100/1100)*-50 + (1000/1100)*5 = 0.0 BB/100 while having $900 in winnings.

If you do the correct thing and take the weighted average of the BB first, you get.
Weighted BB = (100/1100)*2 + (1000/1100)*20 = 18.36363636
Winning/bb/100hands = 900/18.363636/11 = 4.45BB/100
Which is the right and meaningful figure.

This is important for anyone who plays multiple levels.


I hope this makes sense.

Sam

[tiltboy]

[ QUOTE ]
This is important for anyone who plays multiple levels.


[/ QUOTE ]

I'll say this and let it drop. What you have written above is presumptuous--I play levels from $2/4 - $5/10 at Party and my average BB/100 or BB/hr for all these levels is meaningless TO ME. Perhaps you gain some insight into your play by knowing this (I frankly can't imagine what though) but it does nothing for me. If Pat polled me as to whether he should "fix" this or not, I would vote NO.

[ptrack pat]

As I replied to you in a follow up...

Sorry, I had didn't use the parenthesis correctly. It's (Sum of BB won at each level) / (total hands / 100). In your example you sent me, you had won 44.5 BB at 2/4, lost 31.5 at 3/6 and lost 7.5 at 1/2. This comes to a total of +5.5 BB (44.5 - 31.5 - 7.5). Divide this number by 53.12 (5,312 hands / 100) and you get the .10 value that you see.

[Wayne]

[ QUOTE ]

In short, When I have a total loss, my rolled up BB/100 should not be a positive #.

SamJack

[/ QUOTE ]

You are absolutely wrong. It is quite possible to lose money overall, yet have a positive BB/100 ratio.

Example:
$1/$2 99 hands +$180, +90BB
$100/$200 1 hand -$200, -1BB

I have a total loss of $20, yet I have won 89BB, or +89BB / 100hands.

The entire purpose of using BigBets in the calculations is to eliminate the difference in limits. It is used to measure how well you are playing poker across all limits.

[SamJack]

The most value I get out of the BB/100 is Expected return on my investment. If you want this to be reflected in the rolled bu BB/100 across levels, BB must be weight averaged.

Sam

[SamJack]

Hi Pat, thanks for answering my questions.

In this thread some have indicated that the way it's done in Poker Tracker is the way they expect.

I guess I mostly use the BB/100 as an indicator of expected return on investment. In that regard, it seems that taking the weighted average of the BB's would give a number I would like to look at.

Anyhow, thanks for responding to my inquiries.

Sam.

[Wayne]

[ QUOTE ]
The most value I get out of the BB/100 is Expected return on my investment. If you want this to be reflected in the rolled bu BB/100 across levels, BB must be weight averaged.

[/ QUOTE ]

Can't you just calculate winnings / hands played?
or winnings per hour?

It sounds like you want to take this number and convert it to the average BB.

Repeat Example:
$1/$2 99 hands +$180, +90BB
$100/$200 1 hand -$200, -1BB

Which numbers would you like to see:
a. +89BB/100
b. -5BB/100
c. -$20/100

[SamJack]

Repeat Example:
$1/$2 99 hands +$180, +90BB
$100/$200 1 hand -$200, -1BB

Which numbers would you like to see:
a. +89BB/100
b. -5BB/100
c. -$20/100


I'd like to see
B. -5.03BB/100 = (-20/((99/100)*2 + (1/100)*200))/1

[BamaGambler]

[ QUOTE ]

Repeat Example:
$1/$2 99 hands +$180, +90BB
$100/$200 1 hand -$200, -1BB

Which numbers would you like to see:
a. +89BB/100
b. -5BB/100
c. -$20/100


I'd like to see
B. -5.03BB/100 = (-20/((99/100)*2 + (1/100)*200))/1


[/ QUOTE ]Earlier you said you want to reflect how well you are playing across all limits. How does this number reflect that? With the given data you have played very well yet your answer is negative.

[Grivan]

[ QUOTE ]
If you take the weighted average of the BB/100's which is still wrong but is what the software is actually doing, you get.
(100/1100)*-50 + (1000/1100)*5 = 0.0 BB/100 while having $900 in winnings.


[/ QUOTE ]

This is the right way to calculate BB/100 hands. It seems that the number you want to look like is (Money Won)/(100 hands), which is is a different number then BB/100 hands. In the BB/100 hands calculation a BB of $2 and a BB of $20 are considered the same by convention.

[rigoletto]

I fail to see the use of your 'correct' calculation.

Let's say you have the followint results

limit hands result BB/100 hands
5/10 200000 $40000 2
10/20 100000 -$15000 -0.75

Your weighted average would be 0.625 BB/100 hands.

You say you want to know your expected return over multiple limits, but all that happens in this example is that you decieve yourself into thinking that you can beat the 10/20 game!

[Bokonon]

Well, I gather that they do A., and that's exactly what I want, so I ain't bugging them at all!