PDA

View Full Version : Negative $EV tourney numbers



AbogadoLoco
04-30-2009, 04:11 AM
I just tried out the tourney function of 1.08b by playing 20 of the 7.50 superturbo SNGs on Fulltilt.

It appears that HM is coming up with some very odd numbers. The buyins are $7.50 each, and I noticed this when i played a single SNG and my EV$ adjusted profit/loss decreased by *more* than the $7.50 buyin. So, I should have been required to reimburse the other players more than my buyin because I played it so poorly:eek:

So, I clicked on the Data View tab and compared the "$ Won" column and the "$EV Won" column. Most of them look like they should, i.e., i won $0, but the EV adjusted as something like $5, or $9, etc.

But, for 3 of the 20 tourneys, I have negative $EV Won numbers. Now, how can this be? I lost a $7.50 SNG, and my E$EV Won is showing up as $-6.68, for a total loss of over $13 for one $7.50 SNG. This makes me worried about the calculations not just in these 3 SNGs, but also the other 17.

morny
04-30-2009, 07:12 PM
You dont add -6.68 to your buyin, its just showing how much you would have won/lost with average luck

AbogadoLoco
04-30-2009, 07:16 PM
You dont add -6.68 to your buyin, its just showing how much you would have won/lost with average luck

It still doesn't make sense though.


Let's assume I have zero SNGs played. I play in a $7.50 SNG. I lose it, but the EV number is showing a negative $-2.50 number, and my overall expected EV$ as negative $10. This is not me adding this up in my head, this is what HEM is telling me.

So, I played a $7.50 SNG, but I should be down $10? Huh.

Another example. I just played in another $7.50 SNG, which I won, giving me $31.50 for actual dollars won. But, my EV number is negative, so my overall EV adjusted number went down by more then $7.50 for just this one SNG.

Again, this makes no sense.

morny
04-30-2009, 07:20 PM
The problem is with your overall expected EV, there is no such thing, if it say -$2.50 it means if you play the tourney 1000 times youll lose approx 2.50 on average not 7.50+2.50

AbogadoLoco
04-30-2009, 07:27 PM
The problem is with your overall expected EV, there is no such thing, if it say -$2.50 it means if you play the tourney 1000 times youll lose approx 2.50 on average not 7.50+2.50

I think you are missing the point.

I just filtered so just my last SNG is displayed. It was a $7.50 SNG, which I came in first for $31.50. My $EV Won number is -$2.32.

Now, in both the filtered data section, as well as the graph section, it says:

1 Tourney.
Winnings: -$9.82
Average Buyin: $7.50

So, how is it possible to have a *negative* $9.82 "EV$ adjusted value" when playing a $7.50 SNG. It can't possibly be less than -$7.50.

morny
04-30-2009, 08:18 PM
Email me the sng and a link to this thread morny@holdemmanager.net

AbogadoLoco
05-01-2009, 10:25 PM
Not sure if this will work or not, but this is supposed to be the insert code from photobucket for the screenshot that illustrates the issue.

It has one $7.50 SNG, with an expected EV adjusted loss on that one SNG equal to $14.18, or an ROI of -189% :eek:

On the the dataview tab, if you actually go through and add up the numbers and see how these are reflected in the total win/loss numbers, you will see that a $0 in the $EV Won column actually represents a loss of the full SNG buyin, or $7.50 in this case. a $7.50 in this column reflects a breakeven expected value, and a $15 in this column reflects a $7.50 expected win.

Furthermore, if the number is negative, then the SNG expected loss is showing up as an amount equal to the the SNG buyin plus this negative amount. A -$2.50 in the $EV Won column is impacting both the graph and the overall numbers by showing a $10 expected loss for a $7.50 SNG.

<a href="http://s620.photobucket.com/albums/tt289/MensaWashout/?action=view&current=screenshot2.jpg" target="_blank"><img src="http://i620.photobucket.com/albums/tt289/MensaWashout/screenshot2.jpg" border="0" alt="HEM window"></a>

<a href="http://s620.photobucket.com/albums/tt289/MensaWashout/?action=view&current=screenshot1.jpg" target="_blank"><img src="http://i620.photobucket.com/albums/tt289/MensaWashout/screenshot1.jpg" border="0" alt="Photobucket"></a>

crd
08-14-2009, 01:15 PM
Hello,

Is this problem has been solved?
I have exactly same problem with a negative "$ev won" and also with some over positive "$ev won" ( "$ev won" > max winnings)
I sent a HH for overpositive case to support but have no response for the moment.

thanks.

fozzy71
08-14-2009, 03:37 PM
Please www.7zip.org the hands and email them again, to fabio@holdemmanager.net, with a link to this thread, and your forum name.

Rvg72
08-16-2009, 12:55 AM
When we worked on EV implementation we made a decision that allow for "impossible" EV values - this was after consulting with many stat experts. Forcing EV to be no lower than $0 and no higher than the win amount would cause a bias issue over larger samples so we did not include any caps.

To imagine a case where you would have a negative EV imagine a tourney where you won an all-in initially where you were an underdog. You are now "lucky" as far as the overall tourney. On the next hand you call an all-in bet on the river and lose. You won $0 in the tourney and you were still lucky so you have a negative EV.

I know it might not make a lot of sense when you look at an individual tourney but over larger samples (which is the main purpose of the EV line) capping the values would make the results incorrect.

Roy

evertonroar
02-26-2011, 08:53 PM
has this ever been sorted out? im not a stats expert or anything but this justnt feel right. When the EV calcs are done, at no time can you possible lose more than the Equity than you have. so it just cant go below zero at any time. on the reply from rcg72 - to me this doesnt make sense as the $EV is not related to the final result. so it doesnt matter if you 'were lucky but still lost' The EV cant go below 0

Patvs
02-27-2011, 07:29 PM
When we worked on EV implementation we made a decision that allow for "impossible" EV values - this was after consulting with many stat experts. Forcing EV to be no lower than $0 and no higher than the win amount would cause a bias issue over larger samples so we did not include any caps.

To imagine a case where you would have a negative EV imagine a tourney where you won an all-in initially where you were an underdog. You are now "lucky" as far as the overall tourney. On the next hand you call an all-in bet on the river and lose. You won $0 in the tourney and you were still lucky so you have a negative EV.

I know it might not make a lot of sense when you look at an individual tourney but over larger samples (which is the main purpose of the EV line) capping the values would make the results incorrect.






Capping the EV that it cannot be less than the buyin amount, and cannot be more than the price for first place, will lead to biased longterm results.

Don't focus on the EV for just one game, but focus on the longterm results.

mechapoitier
05-03-2011, 01:34 AM
can I play your stat experts hu all day please

negative$ev is absurd, just think dont copy paste.
say you win all in AA vs KK then lose all with second nuts on river. Should you expect to lose more than your buy in? Your graph will say so. The whole system youre using for the calculations is a logic nightmare. YOU MUST RETHINK IT ENTIRELY.

And tons of other problems there: just looking at the small sample I imported I can see many all in hands not "rated" for $EV Diff.

I was pondering a switch from other tracker because of all the keeeeewl features in HEM.
But the level of this... And how you replied to the previous without real consideration...
FAIL.

Patvs
05-03-2011, 01:53 PM
say you win all in AA vs KK then lose all with second nuts on river. Should you expect to lose more than your buy in? Your graph will say so. The whole system youre using for the calculations is a logic nightmare. YOU MUST RETHINK IT ENTIRELY.


Only in some very rare situations will the EV actually result in a negative EV.
For most games the EV will just be between 0 - and the maximum amount you can win.

You must understand the independent chip model (ICM) resulting in the Luck Adjusted Winnings for STTs is completely different than "EV" in a cashgame.

Example:
You're in the bubble... 4 players left. Top 3 is in the money. Blinds are high at 300/600.
Player 1 1500 chips (=YOU the HERO)
Player 2 1500 chips AK
Player 3 5000 chips QQ
Player 4 4500 chips

Player 2 and player 3 go preflop allin with AK vs QQ.
You (player 1) fold preflop... but this hand will still affect your $EV Diff.

(If QQ wins you're automatically in the money.... if AK wins you will be the shortstack most likely to not survive the bubble)

----

As stated in my previous post, not capping the Luck Adjusted Winnings, results in more accurate LONG TERM results.

Can you post an example of a "all in hands not "rated" for $EV Diff"?

mechapoitier
05-03-2011, 08:52 PM
A TON of hands are not rated for $EV Diff
I play heads up shootouts only. I saw another thread pointing out problem in those. But from seeing the diversity of glitches and bugs in your EV stuff, I doubt it is linked to a particular kind of game. IT IS SO VASTLY AND INTENSELY WRONG GUYS. I mean come on, did you ever test your software??

Also dont try and teach me poker concepts. This is disrespectful and if you just knew how unappropriate

I have lost enough time with you and your amateur software already.

I guess Ill have to make my own HUD positional stats on PT.

ADIOS.

mechapoitier
05-03-2011, 08:57 PM
Only in some very rare situations will the EV actually result in a negative EV.

I LAUGHED HARD. Go play heads up shootouts and see what happens in HEM.

Patvs
05-03-2011, 11:23 PM
Since you only play HU shootouts, this is a completely different matter.
(the main discussion in this thread was whether or not negative EV should be even made possible for regular SNGs)

As pointed out in http://forums.holdemmanager.com/manager-bugs/74331-4man-ev-lol.html#post361191 email us a handhistory of a game:
-resulting in a negative Luck Adjusted Winnings, where it doesn't make any sense
-or containing "tons or hands where the EV Diff is not calculated"
It seems there only is a bug affecting the shootouts.


A TON of hands are not rated for $EV Diff
I play heads up shootouts only. I saw another thread pointing out problem in those. But from seeing the diversity of glitches and bugs in your EV stuff, I doubt it is linked to a particular kind of game. IT IS SO VASTLY AND INTENSELY WRONG GUYS. I mean come on, did you ever test your software??
ADIOS.


Can you be more specific of the other diversity of glitches in "our EV stuff"?

mechapoitier
05-04-2011, 11:20 PM
I have no time for emails.
just play shootouts and see for yourself the TON of hands. its like 50% of them in all shootouts so I really dont know why I should send you an EXAMPLE.

Patvs
05-04-2011, 11:40 PM
Fiammaz from the thread: http://forums.holdemmanager.com/manager-bugs/74331-4man-ev-lol.html#post361191
sent us some handhistories.

He plays: €92.80+€7.20 Heads Up SHOOTOUT (Two Rounds)


Four EV scenario's: (what it should be... if the buyin = $100)
1. You win round 1 (no coinflips)---> you win round 2 (with a coinflip)---> You win $400---> $EV Won = $200 (400*50%)
2. You win round 1 (no coinflips)---> you lose round 2 (with a coinflip)---> You win $0---> $EV Won = $200 (400*50%)
3. You lose round 1 (with a coinflip)---> You win $0---> $EV Won = $100 ($200*50%)
4. You win round 1 (with a coinflip)---> you win round 2 (with a coinflip)---> You win $400---> $EV Won = $100 (400-100-200)
5. You win round 1 (with a coinflip)---> you lose round 2 (with a coinflip)---> You win $0---> $EV Won = $100 (-100+200)


Current EV calculation: (for scenario's 4 & 5)
4. You win round 1 (with a coinflip)---> you win round 2 (with a coinflip)---> You win $400---> $EV Won = $300 (it wrongly just does 400-100 ignoring the round 2 all-in hands)
5. You win round 1 (with a coinflip)---> you lose round 2 (with a coinflip)---> You win $0---> $EV Won = -$100
It completely ignores all the all-in hands for the second round.

Handhistory file clearly writes:
"Match Round I" or
"Match Round II" to every hand.

HoldemManager should know that with a buyin of $92.80:
It should treat round 1 hands as a regular $92.80 buyin HU SNG.
And it should treat round 2 hands as a regular $185.60 (2x $92.80) HU SNG and continue to calculate the $EV Diff for every hand.

I've sent this as a bug report to a programmer.

mechapoitier
05-05-2011, 12:00 AM
its true what you said.
I had this feeling that most of the ignored where in the second round but Im so unfamiliar with the UI for now that I could not point the finger on it. also I was very upset about this which probly prevented me to think efficiently.

I mean "ME A LUCKBOX?????????"


I hope update soon so I can see the beauty of tournament ev.
And when its done, do you care for a discount?