PDA

View Full Version : The EV being incorrect myth



RunBad4Life
09-07-2012, 01:43 PM
Why do so many people say the comparison of EVbb/100 v bb/100 for AIPF coin flips in tournaments is wrong? Really fail to see how such a simple sum could be wrong?

fozzy71
09-07-2012, 06:53 PM
Most of the recent noise about Tourney EV on the forums is because there are some problems with the tourney detection of some pokerstars hypers which is causing the EV to be calculated incorrectly.

RunBad4Life
09-07-2012, 07:52 PM
yeh im not takling about STT's or redlines. I'm talking about bb/100 in the reports....MTT SnG's included

fozzy71
09-07-2012, 10:16 PM
yeh im not takling about STT's or redlines. I'm talking about bb/100 in the reports..

I guess I don't understand what you are saying/asking. Can you give a specific example or attach a hand.



..MTT SnG's included

EV is never calculated in MTT's.

http://forums.holdemmanager.com/manager-general/24059-ev-not-all-situations.html

RunBad4Life
09-10-2012, 05:49 AM
What I am saying is extremely simple. reports, set filter to all in on the flop, bb/100, EVbb/100...the difference between the two. Correct or not, if not, why not?

fozzy71
09-10-2012, 04:07 PM
It should be accurate for SNGs, unless you are playing on Pokerstars. There are some known bugs we have been working on with the results and EV for stars tourneys (Hypers in particular). The thread about the stars issue is here - http://forums.holdemmanager.com/manager-general/247931-request-fix-evline-60-200-hypers-6max.html

Here is a link to the latest HM1 internal build - - http://forums.holdemmanager.com/general-support/announcements.html

Please download it and create a new database, then import a sample of recent hands. If you find any EV discrepancies for specific hands you want us to investigate please zip and email the problem hand histories, with a link to this thread and your forum name, to support@holdemmanager.com

RunBad4Life
09-10-2012, 08:28 PM
Pokerstars yes. Well I've had that version for months so switching databases didn't make any difference on figures.

fozzy71
09-10-2012, 09:51 PM
Pokerstars yes. Well I've had that version for months so switching databases didn't make any difference on figures.

That version in that announcement was just uploaded today.

RunBad4Life
09-14-2012, 08:57 AM
oh well it says 1.12.11 which was out months ago? Regardless, there was no change to numbers anyway so I'm guessing they're correct. Not sure how you'd know if they were or weren't anyway...

fozzy71
09-14-2012, 12:53 PM
oh well it says 1.12.11 which was out months ago? Regardless, there was no change to numbers anyway so I'm guessing they're correct. Not sure how you'd know if they were or weren't anyway...

1.12.10b was released in June. Sometime in July or August we changed the version number of our internal builds to '1.12.11 Internal' to help us identify if a customer reporting problems was still using the official release or a newer internal release.

To check the actual version number you need to right-click the HoldemManager.exe > Properties > Details:

The 10Sept build is .7322 I believe.

83342


If you think the numbers are incorrect please send us some screen shots and hand history files for testing, along with a detailed description of what you think is wrong. Without the actual hand histories we can't really help determine if there might be a problem and there are no screen shots or hands attached to this thread yet.

Please zip and email the problem hand histories, with a link to this thread and your forum name, to support@holdemmanager.com

RunBad4Life
09-30-2012, 07:55 PM
Where is that 1.12.11 gone now? link isn't valid and latest release now seems to be the 1.12.10 from june. I'm installing on new machine...

fozzy71
09-30-2012, 09:01 PM
The announcement might have got hidden because it was set to expire on the 30th of September. I just extended the announcement expiration to the end of October so it should be visible again, but here is a direct download link also.

Make sure you install the full HM1 setup first, before you update:

1.12.10b installer - http://www.holdemmanager.com/downloads/Holdem_Manager_Setup.exe

1.12.11 10Sept update - http://www.holdemmanager.com/Downloads/HmUpdate_Release_1.12.11_10Sep2012.exe

RunBad4Life
10-01-2012, 09:55 AM
thanks

RunBad4Life
10-18-2012, 04:22 PM
Am gonna open this topic up again cause I am constantly being told EV for 18man is completely wrong by just about every 18m grinder out there. No one will say how they even know this , just that it's been a taken for granted fact for eons. How do I calculate things to see if they ring true?

An exmple here, we've taken just 1 hand and filter is set to allin=true so AIPF EV in other words.

http://imageshack.us/a/img829/9383/evexample.jpg (http://imageshack.us/photo/my-images/829/evexample.jpg/)

http://imageshack.us/a/img831/8871/evhand.jpg (http://imageshack.us/photo/my-images/831/evhand.jpg/)



How do I figure out if this is correct?

The pot I have 65.2% chance of winning is 5826, how is that cEV difference of 3801 arrived at first of all?

Secondly, I understand how the bb/100 for this hand is -354.75 because I lost 2838 which devided by bb800 is 3.5475 which 100 times over makes an easy 354.75/100. But how is the ev number 120.34? I SHOULD have won 5826 so that devided by bb800 comes to 728.25 bb/100, so how is the 120.34 arrived at?

Next there's the actual cEV figure over on the right stating that adjusted EV means I should've won 962.68. How does that work also?!!

Have a feeling I'm totally not gonna understand the answer so can you pretend you're talking to a 2yr old? Thanks

Patvs
10-18-2012, 07:20 PM
The pot I have 65.2% chance of winning is 5826, how is that cEV difference of 3801 arrived at first of all?




cEV Diff = what you should* have won - what you actually win.
*You should have won 65.24% of 5826 chips: 3801 chips. That's your chip equity.
That's what you mathematicaly should win.

You ended up losing the hand. So you ended up with 0 chips. 3801 - 0 = an cEV Diff of +3801.
If you had WON the hand. The hand would have resulted in 3801 - 5826 = a cEV Diff of -2025.




Secondly, I understand how the bb/100 for this hand is -354.75 because I lost 2838 which devided by bb800 is 3.5475 which 100 times over makes an easy 354.75/100. But how is the ev number 120.34? I SHOULD have won 5826 so that devided by bb800 comes to 728.25 bb/100, so how is the 120.34 arrived at?


You should have won 3801 chips. (3801 / 800 * 100 = 475) The "120" is the Diff between 475 and 355.
475 + -355.





Next there's the actual cEV figure over on the right stating that adjusted EV means I should've won 962.68. How does that work also?!!


962 = -2838 (chips you lose) +3801 (chips you should have won).

Back to the "120". 120 is (962 / 800) * 100.

RunBad4Life
10-18-2012, 09:37 PM
Thanks for that. So it sounds correct then so why does everyone insist it's not? Or is it that when a whole bunch of hands is equated that things start going wrong?

Patvs
10-18-2012, 10:02 PM
There is no Luck Adjusted Winnings in 18man SNGs. (which only works for STTs)

Chip EV works fine, but it's not very usefull. (because it doesn't take the blind levels into account in the overall report, so winning an early coinflip will not compensate for losing a lot more chips in a coinflip late in the game)





So a plausible criticism would be

RunBad4Life
10-19-2012, 09:33 AM
Did you not finish a sentence there?

YEs I know $ev doesn't work for 18's and chip ev is going to be skewed but you can just filter that to blind level anyway if desired. But is the bb/100 and EVbb/100 still correct? If I start a session and after half hour in and 30 AIPF's I am 400 bb/100 below EV it's correct yes?

Patvs
10-19-2012, 07:37 PM
Yes the bb/100 and EV bb/100 for chip EV is accurate.

Though... a plausible criticism would be... there are many hands where All-In EV in itself just doesn't make any sense.
If you encounter such a hand or situation it can completely screw up the results. Read: http://forums.holdemmanager.com/manager-general/24059-ev-not-all-situations.html

RunBad4Life
01-28-2019, 06:20 PM
So here we are once again. Here is a 5 century (feels like it) long graph. Can you confirm this is in no way faulty or inaccurate? I don't think it is faulty, as I know for a fact I run like death 365 days of every single year, but it just seems too insane. I also know of no one person who can produce anything similar, which kinda confirms it isn't faulty also. It covers 25,000 tournaments worth of all-in pre flop hands.

163193

fozzy71
01-29-2019, 11:53 AM
It is impossible for us to know if that is accurate just by looking at a screen shot. If you think the EV is calculated wrong on any specific hands you can export those hands using the right-click menu in the reports and zip/send them to us with a description or screen shots of what results you are seeing so we can have one of the EV testers/developers take a look at them on a hand by hand basis.

udbrky
01-30-2019, 10:45 PM
I also reached out to a HS SNG grinder who plays a lot of hyper turbos to see if he has any ideas.

RunBad4Life
10-05-2022, 11:10 PM
Now double the amount of AIPF hands from SnG's. This translates to 2,000,000 hands overall and fully confirms my account is rigged. I fail to see why this data is not enough evidence to prove online poker RNG's absolutely do NOT randomise. I should be compensated, it's a disgrace!

BB/100
166055

CHIP EV
166056