HEM VPIP oddity I just discovered

dmorris68

dmorris68

League Staff Geek
Can someone confirm if PT3 uses the same logic?

I sat down at a Stars SnG last night. After the 4th hand was over, I looked down at my hero stats and noticed my VPIP was 33%. Unpossible out of 4 hands dealt, amirite? I mean, my own awesomeness aside, even I can't play 1.33 hands, lol.

So I posted the below screenshot crop on the HEM forum:

weird_hud_vpip.png


The reply I got back from HEM support was:

If you played a hand in the BB and got a walk this hand is counted but not used for VPIP
Which explains it, as I did walk my first hand after sitting down, and just called another. But is this right? How is it different than checking my option in the BB, in that either case I didn't VPIP anything. Shouldn't VPIP always be Hands VPIPed / Total Hands?

For instance, if in the first 4 hands of a HU match I:

1. Walk in the BB
2. Raise in the SB
3. Check option in BB
4. Fold in the SB

Shouldn't my VPIP be 25% there, since I only VPIPed 1/4 hands? Rather than the 33% (or 50% if they treat the check same as the walk, and if not, why not?) that HEM says it is?

I never noticed this before, probably because I never looked right at the stats after such an obvious hand count like 4 where the odd percent jumped out at me
 
absoluthamm

absoluthamm

<==Poker Face
I have noticed that before too, but never gave much thought to it because of the small number of hands and figured it would correct itself out with more hands.

With the example you listen above, I would actually assume your VPIP would be 66% if I'm reading it right.

1. Walk in the BB - Calling raise = +VPIP1
2. Raise in the SB = Raising = +VPIP2
3. Check option in BB = No Count Cause of Check
4. Fold in the SB = Had Option to Call But Didn't = -VPIP3

Did I read it wrong?
 
dmorris68

dmorris68

League Staff Geek
With a walk you didn't put anything extra in, so 0 VPIP
A raise you put more in, so +VPIP
Check option, nothing extra so 0 VPIP
Fold, nothing extra 0 VPIP

Essentially the BB by itself never counts, nor the SB if you fold it. Any time you put in anything beyond that, it's +VPIP.

I get what they're saying that a walk is 0 VPIP as I expect, but the hand should not be removed from the VPIP equation -- it's still counted as a hand you played without VPIP, and should be part of the divisor IMO.
 
absoluthamm

absoluthamm

<==Poker Face
My bad, when you differentiated your first BB from your second, I thought you meant walk as a cold call or something. Actually, now that I see hwo you put it, I would rather call it more of 50% VPIP. As the two BB's are just washes, but the SB you had to call or raise to continue the hand, so the one that you folded would still be counted as -VPIP, therefore, you put money in the one SB, but folded the other SB(so it is counted just like you would fold any other position except the BB), therefore you are playing 50% of hands where you were forced to put in money to continue playing. The last part being the most important. I am still a little confused though, maybe it's all the snow outside right now... getting me a little too excited.
 
dmorris68

dmorris68

League Staff Geek
A folded SB doesn't count towards VPIP. Nor a folded BB or a checked BB. In those cases you haven't put any money in that you weren't forced to. They count towards total hands though (i.e. towards the 4 total in my example). In my example, there was only one hand where I "voluntarily" put money in, and that was the raise. Therefore VPIP should be 1/4 or 25%.

What HEM support seems to be saying is that the walk hand doesn't count towards the 4, meaning that they only counted 3 hands in the VPIP equation, resulting in 1/3 or 33% VPIP. That's where (I think) I disagree with them.

But as I re-read his response, he actually seems to be saying what I'm saying all along, that the walk is counted but does not contribute to VPIP. In that case, it doesn't explain how 4 hands can yield a 33% VPIP.

EDIT: A better example that I just posted back on the HEM forum. The HU example was contrived off the top of my head to make it easier to eliminate other positions and account for different blind actions that might or might not contribute to VPIP. This one actually more accurately demonstrates what happened in my game where I saw this.
  1. I sit down and get a walk in the BB on the first hand. My VPIP equation is 0/1 or 0% VPIP.
  2. I call or raise from the SB on the second hand. My VPIP equation is now 1/2 or 50%.
  3. I fold the 3rd hand. My VPIP equation is now 1/3 or 33%.
  4. I fold the 4th hand. My VPIP equation is now 1/4 or 25%.
Thus, there is no possible way that I can imagine 33% being possible out of 4 hands. I can only imagine that netsrak at HEM support misunderstood my question, because his response actually confirms that a walk is counted as a dealt hand but not towards VPIP.
 
Last edited:
Stu_Ungar

Stu_Ungar

Legend
ITs interesting but TBH it dosent have much, if any, practical effect.

Over 20, 30, 40 hands sats are meaningless anyway so looking at VPIP over 4 hands is fairly pointless.

Lets say you have 1million hands on a player. What adjustments do you make given his VPIP is 30% vs 33% ?

If the answer is "I make many adjustments, here is a list".. then yeah fine Ok you have a point.

If the answer is "I cant really put the adjustments into words.. but I do make adjustments" .. then in reality you dont make any adjustments you just dont want to admit it (If you cant say what they are then they dont exist).

If the answer is "I dont really do much if anything differently" then why worry about it?
 
dmorris68

dmorris68

League Staff Geek
Thanks Stu.

I agree that 4 hands is pointless, but it's an excellent sample size to mentally pick up on the details of the calculations. It's what made it stand out to me and be noticed.

I also agree that calculating VPIP this way isn't likely to make enough statistical difference to matter because walks are not that common in the grand scheme of things and over large samples. However it has been pointed out that in STTs walks are a lot more common than in cash games, so it could skew a bit more there.

It's more of a math principle to me. VPIP is such a fundamental poker stat that it should be universally defined and accepted, and not varying in interpretations like it seems to be now.

I can accept either definition, as long as I have a sense it's the "right" one and that we're all the same page with it. It's just the nit in me -- I have a compulsive need to just know the right answer, whatever it is, and what I've believed to be the right answer up to this point (as apparently others do as well) is now in question.
 
Stu_Ungar

Stu_Ungar

Legend
This is where PT3 has the edge over HEM.. If you dont like the stat or think it should be slightly different... redefine it to whatever you like!

HEM.. you get what you are given!
 
absoluthamm

absoluthamm

<==Poker Face
You can create your own stats in HEM as well, it just isn't as easy as it is with PT3. It is all done outside of the program.
 
bonflizubi

bonflizubi

Visionary
there is a long debate over on 2+2 about this discrepancy. PT3 set the definition over 10 years ago and it has become the defacto standard. HEM has decided to use a different definition. (For reasons unknown).

I think the debate was in the official PT3 thread over there but I don't remember if that was where it was or not. I do remember that th HEM guys responded by saying they'd re-look at what they were doing.
 
dmorris68

dmorris68

League Staff Geek
there is a long debate over on 2+2 about this discrepancy. PT3 set the definition over 10 years ago and it has become the defacto standard. HEM has decided to use a different definition. (For reasons unknown).

I've heard some say PT3 calculates the same as HEM, but I've seen others (like The Minder on the HEM forums) do a side-by-side test after importing the same hands, and there was a discrepancy in VPIP so obviously it isn't calculated the same between them.

I think the debate was in the official PT3 thread over there but I don't remember if that was where it was or not. I do remember that th HEM guys responded by saying they'd re-look at what they were doing.
Well, they haven't responded again in the thread I created on their forums, so I guess either they're tired of hearing about it, or they're circling the wagons to come up with either a mea culpa or a gtfo, lol.
 
Deco

Deco

Legend
I'm making a HUD atm. I ran into this problem. Just saying thanks for this thread it's saved me a lot of time not having to find out HEM and PTs stance on this.

[*]I sit down and get a walk in the BB on the first hand. My VPIP equation is 0/1 or 0% VPIP.
[*]I call or raise from the SB on the second hand. My VPIP equation is now 1/2 or 50%.
[*]I fold the 3rd hand. My VPIP equation is now 1/3 or 33%.
[*]I fold the 4th hand. My VPIP equation is now 1/4 or 25%.[/LIST]Thus, there

PT3s stance would be this right?
* 0/0 = 0% :: 1 hand
* 1/1 = 100% :: 2hands
* 1/2 = 50% :: 3hands
* 1/3 = 33% :: 4hands

Did you find out exactly what rule HEM follows?


Personally I believe PT's way is best. I googled this after testing my HUD against my Brother and Dad. They play 100% of pots, I was running card dead for the first dozen hands, I didn't raise once. Despite playing every single hand both of them had vpips no larger than 66% because the BB checks were counting as folds. I think seeing as the player has not made a decision on whether to contribute to the pot or not it should not count towards the players vpip as a fold or call.

My HUD is for live poker, sample sizes will be much smaller and the whole table limping to the BB much more common so it is fairly important I get this right.
 
dmorris68

dmorris68

League Staff Geek
No, I never did figure out who follows what rules. I had forgotten about this actually. Since I now run HM2 and am beta testing PT4, I suppose I could gin up some test hands and compare.
 
dmorris68

dmorris68

League Staff Geek
I'm making a HUD atm. I ran into this problem. Just saying thanks for this thread it's saved me a lot of time not having to find out HEM and PTs stance on this.

PT3s stance would be this right?
* 0/0 = 0% :: 1 hand
* 1/1 = 100% :: 2hands
* 1/2 = 50% :: 3hands
* 1/3 = 33% :: 4hands

Did you find out exactly what rule HEM follows?

I just re-read your post on this. No, what you describe is what HEM was doing, not PT3.

According to what bonzi posted above, the 25% VPIP that I assumed was correct was PT's algorithm, and for some reason HEM decided not to factor walked hands into the VPIP denominator at all.

I've not checked to see if either HM2 or PT4 changed their calculation since then (doubtful). It would be easy to check -- create a new database in each and import the same sample of hands in both, making sure some hand(s) were walks. The VPIP's will differ if they're still using different algorithms. If the difference is that HM2 still doesn't consider walked hands in the VPIP calc, then the total VPIP% for all hands imported should be higher for HM2 since the denominator would be smaller. E.g., if you imported 100 hands with 2 walks, PT's VPIP formula would be x/100 where x is # of hands VPIPed, while HM2's formula would be x/98.

As I said, I preferred the PT result (25% in my earlier example) over HM's result (33%) just because it seemed more logical to me and is what I expected, but at least if everyone could just agree on one or the other it would make me happier. I hated having two different concepts of VPIP.

vpip = handsPlayed/vpipChances
Hows this look? Come up with this little bit of code from this thread. I think it follows the BB vpip rules.

http://pastebin.com/cWmpmUra

Well again, whose rules and who is right? But either way, no, that doesn't look right. :)

First, what do you mean by handsPlayed in the formula vpip = handsPlayed/vpipChances? Do you mean hands dealt? Hands not folded? To be clear, it should be called something like handsVPIPed, since the numerator in a VPIP fraction is always the number of hands where you voluntarily put money into the pot. If you check your option in the BB, you've still played the hand but you've not VPIPed. So either you've poorly named the variable or you've misunderstood the VPIP calculation completely. :)

Second, you don't account for the BB checking his option, which both PT and HM do count towards the VPIP denominator (i.e. for one checked BB, VPIP = 0/1 = 0%). So using your logic it should be more like:

Code:
// If we're not in the bb OR someone has raised or called, increase vpipChances
if ( position != 2 || raised > 0 || called > 0 )
  vpipChances++;
THAT would basically mimic the way HM1 was calculating VPIP.

But my argument was, and still is, that the PT approach seems more logical in that it increments vpipChances whether walked or not. In other words, take the if() condition out entirely and just always increment vpipChances for every single hand dealt (or, more logically, just use the hand count and don't have a separate vpipChances).

To illustrate by example one more way:
  • After 4 hands my VPIP is 25% (1/4).
  • On the 5th hand I get a walk.
  • PT3 would have my VPIP at 20% (1/5), i.e. the walk hand was added to the denominator.
  • HM1 would still show 25% (1/4) because it never counted the walk at all. It's like the hand was never dealt.
 
Deco

Deco

Legend
handsPlayed hasn't confused me so far as up until today my calculations were:
vpip = handsPlayed/hands
pfr = handsRaised/hands

Now I'm no longer using hands as my denominator (for vpip) I can see why it would be confusing. I'll rename it when I next get the chance.


To illustrate by example one more way:
  • After 4 hands my VPIP is 25% (1/4).
  • On the 5th hand I get a walk.
  • PT3 would have my VPIP at 20% (1/5), i.e. the walk hand was added to the denominator.
  • HM1 would still show 25% (1/4) because it never counted the walk at all. It's like the hand was never dealt.

If it folds to our bb we did not get an opportunity to voluntarily put money in the pot so why should it count? Same applies to having it limp round to us, rather than having the option to call/raise to add to our VPIP we only have the choice to raise.

This first started bugging me when playing a 3way game with my brother and Dad. They are both pretty close to 100/0s (I've never bothered to teach them how to play :p). Yet due to myself having a cold run of hands so the BB was never being raised their 100%vpips turned into 66% as they were having the hand fold to them or limped to them every time.

Second, you don't account for the BB checking his option, which both PT and HM do count towards the VPIP denominator (i.e. for one checked BB, VPIP = 0/1 = 0%).

Crap are you sure? I thought checking your option didn't count towards the denominator. I best get looking this up again.
 
Deco

Deco

Legend
Damn looking at your HM forums post and the thread they linked you to and they don't mention checking round.
I did spot pfr isn't counted when it folds to the BB. It's less significant but probably something else I should add as you do indeed not get an oppurtunity to raise in that spot.
 
dmorris68

dmorris68

League Staff Geek
If it folds to our bb we did not get an opportunity to voluntarily put money in the pot so why should it count? Same applies to having it limp round to us, rather than having the option to call/raise to add to our VPIP we only have the choice to raise.
I'm not saying it adds to your VPIP (which would imply VPIP% goes up, i.e. numerator increments). I'm saying in PT, which is where I first learned poker stats, it is counted as a hand you were dealt in (i.e. increments the denominator). Therefore, because it changes *something* in the fraction, the VPIP number changes. With HM1, for a walk they didn't change either value in the fraction and so VPIP stayed the same.

Neither a walk nor a check adds voluntary money to the pot preflop (VPIP is a preflop only stat) obviously. PT counted them both the same (increment denominator), which seems consistent. But HM1 counted only the check and ignored the walk, which seems inconsistent. As bonzi pointed out, this discrepancy had apparently been debated at length at 2+2 before I posted this, so obviously others felt strongly about it too.

Don't get me wrong, I can accept the school of thought that says with a walk you had zero choices to make, therefore it shouldn't count anywhere -- consider it like a misdeal or something, it never existed. Ok, fine. Let's just do it the same everywhere. My problem was that when something like VPIP is defined as a "statistic" there is a reasonable expectation of it always being a mathematical certainty and not subject to different values or interpretations. If it's true that PT invented the stat, then HM1 shouldn't have perverted it with their own calculation and called it the same thing. That was my overriding point.

I'm over it now, as we discussed in the grand scheme of things over large sample size, it's a pretty minuscule statistical difference because walks are infrequent enough as to not skew total VPIP too much between PT and HM. Over thousands of hands it might not be more than a couple of percent maybe (speculating). I'm just a perfectionist and seize upon these kinds of things because the inconsistency annoys me. :)
 
Deco

Deco

Legend
Got a pretty good response from HM on what's what.
http://forums.holdemmanager.com/hud-bugs/42142-hud-vpip-bug-2.html#post1057891


I'm over it now, as we discussed in the grand scheme of things over large sample size, it's a pretty minuscule statistical difference because walks are infrequent enough as to not skew total VPIP too much between PT and HM.

Completely agree. With my HUD I'll have smaller samples lower numbers of player and more common BB walks (Live passiveness) so the effect is amplified somewhat.
That and coding stats has been my favourite part of making this App so any excuse to improve them I seize upon.
 
Reptar7

Reptar7

Visionary
tl,dr most of it. Does it matter? Because a sample of 4 hands is worthless. Over hundreds of hands you won't be able to tell as it will have no significance. It's kind of interesting, but even if it is wrong I don't need it to be fixed.
 
Deco

Deco

Legend
tl,dr most of it. Does it matter? Because a sample of 4 hands is worthless. Over hundreds of hands you won't be able to tell as it will have no significance. It's kind of interesting, but even if it is wrong I don't need it to be fixed.

An example of were it heavily affected my stats. sample was about 20hands yet vpips were about 30% off the mark.

This first started bugging me when playing a 3way game with my brother and Dad. They are both pretty close to 100/0s (I've never bothered to teach them how to play :p). Yet due to myself having a cold run of hands so the BB was never being raised their 100%vpips turned into 66% as they were having the hand fold to them or limped to them every time.
 
dmorris68

dmorris68

League Staff Geek

It's been a long time (over a year, which is like ancient history at my age as far as memory goes) since I posted in that thread and looked into this, but I'm 94.3% certain :p that checking in the BB affected VPIP% with HM1 as it did with PT3. Only walks were treated differently. But I can't say with 100% confidence now without revisiting and researching. And I'm not really motivated to do so. :)

tl,dr most of it. Does it matter? Because a sample of 4 hands is worthless. Over hundreds of hands you won't be able to tell as it will have no significance. It's kind of interesting, but even if it is wrong I don't need it to be fixed.
By not reading the thread you miss the point. ;)

It's not about 4 hands being significant -- of course it isn't, as was explained repeatedly throughout the thread.

The point was only that 4 hands made the discrepancy stand out and be easily observable. VPIP is considered to be a mathematical statistic -- which by definition is concrete and not open to interpretation or subtle differences in calculation. My point was if HM was going to change the basis of a statistic, they should have called it something else.

In practical terms we all agreed it was miniscule over a large sample, but it's a difference nonetheless. And Deco has a point that it will be of greater issue with his live tracker because the sample size will pretty much always be an order of magnitude lower than online (of course, one can argue that some poker stats just don't apply to live for that reason, as they really aren't valid without a large sample, but typically VPIP is useful even over small samples).
 
Last edited:
Poker Odds - Pot & Implied Odds - Odds Calculator
Top