7126 - some stats are not correct

[Singum]

Hi,

River Fold after Bet-Call Flop & Turn Bet can be 200% for me, I am not sure of other "after" stats. Please verify, do I need to send any log?

[random]

Hello Singum,


Thank you for bringing this to our attention.

I tested stat on few different hands and was unable to reproduce this issue.

Looks like this is only a problem with a specific situation. Would you happen to have handhistory where this situation occurred so we can have a look at it?


Thanks you!

[Singum]

Here it is: https://docs.google.com/open?id=0B5Nid_abAM0TZUJQQWZRR19pclE

Check that stat of ilikesashimi

[Veteran68]

Don't know if it's related, but for the first time ever, since the 7126 update, I'm seeing PFR exceed VPIP. I first saw it on a villain on 1 table, who had a VPIP/PFR of like 18/20. Then later I noticed my own hero stats on 2 other tables were reading like 17/18 and 18/20. Pretty sure this is unpossible. :)

[Singum]

This is a tracking program. Any wrong stat is making us lose money. I think the test model of HM2 is very unreliable.

I only can notice some wrong stat if it's completely impossible. I don't know if there are other wrong stats with smaller margins.

[Not Excessive]

Don't know if it's related, but for the first time ever, since the 7126 update, I'm seeing PFR exceed VPIP. I first saw it on a villain on 1 table, who had a VPIP/PFR of like 18/20. Then later I noticed my own hero stats on 2 other tables were reading like 17/18 and 18/20. Pretty sure this is unpossible. :)
If the stats are calculated from "opportunities", then I think PFR can be > VPIP.

[Singum]

Can someone from HM team confirm this is a bug from my HH?

[Sarek]

Situation: Hero plays SNG (32 hands)
he had 32 opportunities to do VPIP - correct?
he DID vpip 3 times
and all 3 times he did this in form of PFR
so VPIP = 3/32 - correct?

Continue...
when he played, he encountered allin from shortstacks 2 times.
In new HM logic, in these 2 hands he has NO opportunity to do PFR.
This means: total # of oppotruniries for PFR is (32 - 2) = 30
so PFR = 3 / 30
correct?

now even without running Calc.exe :
is (3/30) bigger than (3/32) ???
YES

[random]

Hello Singum,


Thank you for posting the hand histories above.

I was waiting for developers to include the fix into build before posting, but since you are requesting a reply I would like to inform you that we have already released the fix for this issue internally in built 7134.

Fix will be included in any builds following 7134 that will be posted in our forum: http://forums.holdemmanager.com/hm2-releases/

Please be assured we take stat issues very seriously and they take high priority over any other issues.


If you have any more questions or concerns, please feel free to ask.

[Veteran68]

In new HM logic, in these 2 hands he has NO opportunity to do PFR.
This means: total # of oppotruniries for PFR is (32 - 2) = 30
so PFR = 3 / 30
correct?
The way you explain it here, yes. But according to the bolded part, you're saying this is a new change -- which I assumed it had to be (or a bug), because I'd never noticed it before, then after this last update it was happening on nearly every table. So I guess my question would be, why are you changing a universally understood stat? To my knowledge, and I'll poll some fellow online players to be sure, but historically PFR has always been calculated based on action and NOT opportunity. In all my years of using trackers, and not just HM, this has been the case as far as I remember. Stats like VPIP and PFR are almost universally understood to mean something very specific, so why would you want to deviate from the norm here?

It brings to mind the same debate we had a couple years ago over VPIP calculation, another formula that everyone takes for granted yet PT3 and HEM calculated them slightly differently. And I happened to agree that PT was correct there as well. It's a bit off-putting to see common stats like this that are calculated differently for no apparent reason. I'd sure like to hear the reasoning behind these decisions.

[udbrky]

Our goal is, has always, and will always continue to be, to have the most accurate stats possible. So to answer your question of why we are changing a universally understood stat? to make it more accurate.

You said that historically PFR has always been calculated based on action, NOT opportunity, but this is incorrect. Before you can act, you have to have the opportunity to act. And PFR = times you acted by raising preflop divided by times you could act by raising preflop.

If an action is not possible, it shouldn't be counted as an opportunity.

And this is what the clarification is regarding, in the times when there is no possible action to raise, the opportunity to raise is not counted, thus it is not a PFR opportunity:

Player 1 (5000) raises All-in
Player 2 (10,000) has the opportunity to raise, call, or fold.
Player 3 (7000) has the opportunity to raise, call, or fold.
Player 4 (3000) is last to act and can only call or fold. In this spot, there will be no raise button on his screen - only call all-in or fold.

Thus, he can't PFR - which means that he is the preflop aggressor. So if you were coaching this player, would you look at him and ask why he didn't raise preflop? The stats would suggest maybe he didn't raise as big of a % of his hands that he played, if you count that as an opportunity. But it's not an opportunity.

[Veteran68]

I don't disagree with your logic as outlined. It makes logical sense. However I'm also of the opinion that once a stat becomes universally understood and accepted, changes should be made extremely carefully and probably coordinated across the domain/industry. I've never seen a PFR stat exceed a VPIP stat before, and I'd argue that most poker players that are aware of these stats haven't either -- I'm a mod or admin on a couple of different poker forums so I'll poll others to confirm my assumption that PFR has always been understood to be a subset of VPIP.

My concern is that if you and your competition aren't coordinated on changes like this, now you have PT users talking about a certain stat that means something different to a HM user and vice-versa. And when ESPN announces players VPIP/PFR, which version are they using, HM's, PT's, or maybe their own? So stats are no longer considered "sacred." As a contrived example, it's akin to someone deciding in a vacuum to change the Farenheit to Celsius formula slightly because they believe it's more accurate, despite the fact their results won't match everyone elses.

I'm just a consistency nit I guess. In the grand scheme of things and over a significant sample size, just like with the VPIP differences HM and PT had (or still have, I haven't checked lately), these changes aren't earth shattering and will be hardly noticeable if at all. But it's still perverting what I believe is a widely held assumption that PFR never exceeds VPIP.

I'm not making a huge issue of it, I just like to understand the reasoning behind decisions like this and to debate the merits of them.

[random]

Few points I would like to make:

1) Our competitors have also switched to more accurate versions of VPIP/PFR, rather than using stats as we know them from previous versions.

2) You are correct that over big sample in cash games, actual difference in values of these stats is negligible, and you will not see PFR being higher than VPIP. So this change makes no difference at all to you analyzing players' games.

3) Where you will see PFR higher than VPIP is in very aggressive games where players are facing a lot of all ins. Especially in tournaments and very short stack games. Best thing to do is to take advantage of new information that this stat provides. When you see someone with higher PFR than VPIP (say you see someone with VPIP/PFR 20/25, rather then 20/20 after 10 hands), you instantly know that he was faced with all in bets in some of the previous hands. How you use use that information is up to you, but now you have more information that was previously missing.

4) Many players already replaced their PFR stats in HUD because they didn't feel these stats were accurate enough since they show all raises combined(RFI, Raise over Limpers, 3Bet, 4Bet...). Instead they use RFI (Raise First In, which is not affected by this change in calculation logic, since we dont face all-ins when we are first in the pot) and 3bet stats, which are more accurate when we trying to determine players range in certain spots.

Hope this helps clear up some things.

[Veteran68]

Good points all around. Thanks for the clarification. If you and the competition are somewhat coordinated in this effort to make stats more accurate, then I support it. I just have a problem with the biggest stat trackers in the industry having different definitions of stats between them.

As to PFR vs RFI, I keep PFR around because I do prefer to lump in RFI + raise over limpers. If you break everything out into its own stat, then you risk stat overload -- sometimes it's good coalesce related data together. But I agree that 3bet/4bet stats should stand on their own and I do use these.

Sometimes change is good, it just takes time and adjustment to accept them. :)

[udbrky]

I disagree that we all need to get together and have the same stat. For years, we've disagreed over facing 3bet spots - and we both feel we're right (iirc, PT4 will adjust and be like us). But if we see our stat as more accurate, we're not going to go with the "industry standard" just because everyone else does it.

this is how things don't get changed when they should.

[Sarek]

I don not want to post direct links, but there ARE threads in competitor's forum with the same question "why PFR is bigger than VPIP" - and almost with the same answers: "with the updated and improved VPIP and PFR definitions in PT4 that can happen"

[Veteran68]

Interesting discussion to be sure, but unfortunately a hijack from the OP's issue apparently. I though they might be related HUD/calc bugs in the 7126 update.

I will say that I've posed the PFR question on two other poker forums I frequent. I posted it as a teaser without explaining why at first, so as not to coach or prejudice the feedback. On the one forum where I've gotten replies, the unanimous response was PFR could never exceed VPIP. Not a single one could conceive of such a thing. Small sample size and all, but it would seem to indicate (as I thought) the vast majority of players who use stats share the same belief in how such a fundamental stat is calculated, and were incredulous to hear that PFR could ever exceed VPIP. After I explained your position and acknowledged that over a decent sample it wouldn't matter they tended to agree, but it doesn't change the fact that such a significant change in a fundamental stat is jarring to a lot of people.

It was also mentioned that cash players would rarely if ever see this, but those who played primarily super turbos or all-in-or-fold games would see a much higher occurrence of it. So I predict you're going to get a few more concerned reports of it before it becomes widely known and accepted. :)

[Patvs]

All stat formula's constantly change over time, even VPIP and PFR.

VPIP was already changed, if it's folded to you in the big blind and you win the hand, that hand counts as a "hand played", but it doesn't count towards your VPIP. PokerTracker recently also had their VPIP use the same tweaked formula so it matches ours.

We basically -in turn- now use their updated PFR formula. (my initial response was... it made no sense)
So the good news is both parties strive for a (coordinated) universal formatting of stats. Even when the stats definition is different than what everybody grew up on... reading Super System, and The Theory of Poker.


Anothe example of a change:
4 bet range in HM2 is different than HM1.
HM1 uses: PFR * 4bet
HM2 uses: RFI * 4bet.

[Veteran68]

So the good news is both parties strive for a (coordinated) universal formatting of stats.
This is what helps me live with the changes. :) I'm okay with change, as long as it's logical/rational and everyone is consistent.

[Singum]

ok, i am waiting for next release

[goldseraph]

ok i have a player i have 44k hands on, and he is now a 15/52 vpip/pfr, I have no idea how to read this now. He is also showing as a 37% 3bet, surely he isn't 3betting 37% of opportunities. He has a 56.6% 4bet now - that means he 4bets that often when his open raise is 3bet? I find this whole process very confusing and there needs to be some guide to adjusting to you guys completely changing how stats are calculated/interpreted.

[random]

Hi goldseraph,

we discovered an issue when using observed/datamined hands, so far only reported on Cake poker. Could you please confirm if this is the issue in your case as well? If it is, what site do you play on and if you can submit few handhistories to hm2support@holdemmanager.com, together with a link to this thread.

We are looking into and appreciate any feedback you can provide.

[wlodi]

Hi,
the same problem on entraction, some player has sqz 33% 3b 40% in HUD, after a lot of hands, which isnt true, that sad.

[random]

Hi wlodi,

Could you please export hands for this player and send them to hm2support@holdemmanager.com ? We would like to take a look at this asap and fix any potential issues.

[wlodi]

Dont know how and which of them, but I am after database update, play in this moment 4handed game on entraction and my oponents has stats in hud:
VPIP,PFR,3b,sqz

23/68/56/47.3 after 22k hands
40/90/95/75.0 after 7.3k
38/63/25/ - after 121

[Veteran68]

Dont know how and which of them, but I am after database update, play in this moment 4handed game on entraction and my oponents has stats in hud:
VPIP,PFR,3b,sqz

23/68/56/47.3 after 22k hands
40/90/95/75.0 after 7.3k
38/63/25/ - after 121

Do you have datamined hands? There was a bug in an earlier release that caused large samples of datamined hands to miscalculate the new stat changes. I have no datamined hands and do not see this with any of my large-sample players.

[fozzy71]

Dont know how and which of them, but I am after database update, play in this moment 4handed game on entraction and my oponents has stats in hud:
VPIP,PFR,3b,sqz

23/68/56/47.3 after 22k hands
40/90/95/75.0 after 7.3k
38/63/25/ - after 121


Do you have datamined hands? There was a bug in an earlier release that caused large samples of datamined hands to miscalculate the new stat changes. I have no datamined hands and do not see this with any of my large-sample players.

This should be fixed in 7243+. Unfortunately, you will need to update to the latest internal build and import the hands to a new database to fix this.

Here is a link to our latest internal build 7283 - http://www.holdemmanager.com/Downloads/7283_Hm2AutoUpdate.exe

*This is a brand new internal build with limited internal testing at this point. There may be undiscovered bugs in this builds so please use with caution and report any bugs you discover so we can get them fixed before official release.