The other day I played a hand like this, I saw on the hud it shows EV Diff = 8.65. Meanwhile, my equity is 60%, which means EV Diff must be greater than 46 (the amount I lost). I think the software confused EV with Diff in this hand, so I emailed the Drivehud team about this issue.





As a result, they also calculate EV with a formula that is unlike anything I've learned about GTOs. Their certainty when responding through multiple emails made me doubt what I had learned, so I decided to bring this issue to the group and ask for everyone's opinion. Is my EV in the above case 14.377 or 9.18 like they said?
Besides, I also don't understand the formulas in the last email they replied to me, can anyone explain it to me more clearly?
Thank you!

Hi Duong,

EV calculations, $EV, $EVdiff and the like are at this point very well tried and tested calculations and you can expect any of the major software that have been out more than a few versions to get them correct in every circumstance. Your default then, should be that you are misunderstanding what a stat means, not that the developers have made a mistake. By all means bring it to their attention if you think there is a bug, but please accept what you're being told.

Rather than delete this thread I'll first try and explain.

$EV diff is the difference (hence "diff") between what happened (the hand result), and what will happen over the long run ($EV, pot equity).

In this hand, you play a $93 pot and get all-in on the flop with 60.05% equity vs villain's 39.95%. This adds up to 100%. It is correct, and I verified using a different equity calculator.

So, your $EV of this hand is $93 * 0.6005, which is $55.8465

What actually happened, is that you lost $46. Your $EVdiff, then is the difference between what did happen and what "should". You "should" win $55.8465 averaged for every time you invest $46 in to the pot in this situation (that is, win a $93 pot 60.05% of the time), so the difference is:

$55.8465 -$46 = +$9.84.


If you'd have actually won this hand you would have gained $93. Your $EVdiff would then be showing a large, negative number! Which, you'd probably say "that can't be right, since the hand is plus EV!". But no, in this case $EVdiff would be correct to show negative -$35.1535, for the very same hand up until the flop (with a different runout). Because you "should" not have won the whole $93, only 60.05% of it. And $EVdiff is the difference between what happened and what "should".


Now then, with regards to your graph question. Yes, the purple line is going down. BUT, it is going down less than the real money line. $9.84 less in fact. That there is the $EVdiff in action. You expect the graph to go up. It did! but you need the other scenario too, where you win, the real money line shoots up by $93 yet the EV line goes up $35 less than that to "balance it out". Then, your real money and EV lines would be both at +$19.69, but the swings would be lesser (but still there!) on the EV line.

You keep coming back to "EV diff should be >46 because I am more than a 50% fav". That's not $EVdiff, that's just plain $EV. $EVdiff is the difference, and over an infinite sample the sum of all $EVdiffs will be zero, no matter if the player is a strong winner always playing spots where they are the fav, or a massive fish donking off huge stacks.


Argh, it's been years since I've done these calcs so maybe there are errors myself. But I think that's right!

Hope that helps,

Dave.

Hi Dave,
Thank you for changing the title for me. DH2 responded to the old headline that offended them and I also found it inappropriate to write.
I've been using DH2 for 5 months now, and it's just that hand that makes me wonder.
Thank you for your extremely detailed explanation, I understand what you mean.
I applied your formula in another hand, with similar conditions:


As you said: $EV = 120.5*0.595 = 71.7
I actually lost 59.7
so the difference is: 71.7 - 59.7 = 12
EV Diff according to you and Brian explained would be $12 while the hud shows EV Diff = 70.85
How do you explain this?
Besides, if calculated according to the formula I presented in the email to Brian, it will give accurate results with other hands that I have played, except for the AsKs hand that I have questions about.
The purple line in the graph is the EV line, it cannot go down when equity > 50%
Thank you!
Duong

Hi,

Indeed I would expect $EVdiff to show a value of $11.99 here as you have also calculated. $70.85 seems too large a, $EVdiff value here for any kind of hand at these stakes, much closer to raw $EV but it's also not quite that? ($71.68).

It would appear there is indeed a possibility of inconsistency for this stat between these two hands!

I am not a DriveHUD user, are you then saying "normal" $EVdiff is similar to this one (large values), with AsKs being the outlier?

That's right, I think there was an error with the AsKs hand, so I tried to talk to HD2 to fix it, because I bought a 1-year package of this service.
As you can see in the emails I sent to DH2, I intentionally took pictures of these two hands, with similar characteristics but the EV Diff displayed is very different. But they didn't think it was unusual.
I will present it again for you to understand as follows:
First, I agree with what HD2 confirmed in the second email:
EV diff. = "how much you win on average" - "how much you actually win"
I do not agree with the EV calculation formula that you and DH2 calculated, as far as I know and have checked with google on reputable sources:
EV = (%W * $W) – (%L * $L)
I will apply the above formulas to hand Kh9h:
EV = (0.595*62.2) - (0.405*58.3) = 13.4
how much you actually won = -58.3 (due to the calculation done on the flop)
EV Diff = 13.4 - (-58.3) = 71.7
The reason for the difference with the number 70.85 is because hud subtracted the rake before performing the calculation.
What if I win this hand, then?
how much you actually won = 62.2
Then EV Diff = 13.4 - 62.2 = -48.8
Why the value seems larger than you think, is because of these hands I lost. According to equity, my ev should be a positive number, so the gap between the amount I lost and "should have won" would be larger than the amount I lost.
To make it easier to visualize, on the graph, EV Diff. is the distance between the purple line (EV) and the green line (Net won). This is true for all hud software, including DH2 or PT4.
The above is my opinion. I hope to soon receive feedback from you and others about the calculations I have made.
Thank you!

Hi,

I think one point which is causing issues is that you are using an EV calc that is suited for the purposes of calculating the EV of a play - like, "what is the EV of making this bet" where "amount won" and "amount lost" are (/can be) different values. You demonstrate this with:

EV = (51.3*0.6005) - (41.7*0.3995) = 14.37

This is correct for how to calculate the EV of this call, taking in to account the dead money from previous streets.


All-In EV calcs (AIEV) does not do this, it simply applies equity to the whole pot, as if all the money went in on the street the all-in happened. Probably some applications call it "Equity Adjusted" and we poker players call it incorrectly EV or something like that.

They have to, otherwise the graphs would be biased and always trend in one direction as you're only ever doing EV calcs on half the dead money where you are ahead. You can make always profitable EV plays by betting large (but not all-in) on early streets, and then "should I call $10 to win $150" is nearly always a yes, and an EV-adjusted graph would always show you being very unlucky as you should clearly be a winner making these +EV bets all the time. Obviously people don't go to that extreme, but just the constant blinds would bias the graph in that direction! So AIEV is always just whole pot * equity. I think that's why they do that anyway! Hopefully that makes sense.

So, the above calc would be the $93 * 0.6005 = $55.85 ($9.84 EVdiff), which I guess now the $93 is a pre-rake number so all these numbers are a little wrong but I don't think we show the post-rake anywhere right? But that's where the $8.65 comes in probably?)


Here's an old Pokertracker article which covers mostly different problems with AIEV, but does show this method used: https://www.pokertracker.com/blog/20...-all-in-equity


Still, I could def be making errors here myself - it's been a long time and it's very easy to get the order of things back to front somewhere, or mix up the definition of two terms and such! And this doesn't address really in any way why there is seemingly different methods between the AKs and K9s hands. Still looks like perhaps something unexpected is going on somewhere!

Thanks for your post. I read through your ticket in support. The reason the calculation is wrong is because you're playing on a capture site (WPT) that doesn't provide hand histories, and the rake was captured incorrectly. There were some issues w/ rake in the initial release, and you likely just need to update to the latest version of DH2. It's possible there's still some issues w/ the latest, but there were rake fixes in the latest build.

You were told by support to wait a day so the developer that handles these calculations could check for you, because he was ill. One of the other developers who sent the formula, doesn't handle the graphing, and doesn't know this part of the product. If you had just waited, we could have had you send an advanced log and figured this all out.

Any ways, I'd still advise sending a log from the support icon in dH2 and checking the option, send adv log. We'll have someone verify if it's still an issue in your version.

Thank you so much Dave.
I have looked at the link you sent. Now I know the difference between EV and Allin EV.
Most of the articles I've read about poker knowledge only use Allin EV, and many times they just write "EV". So I thought the two terms had the same meaning.
I would like to confirm that every time I have mentioned EV before, it has meant Allin-EV. And the term used in PT4 and HD2 is also Allin-EV. The purple line on the graph is Allin-EV. The calculations I did were also Allin-EV.
However, I agree with you that there was an error that led to too much difference in EV Diff results between AKs and K9s hands. I will send the data requested by FreakDaddy for further processing by DH2.
Sincerely thank you

Thank you, I have sent advanced log