I did read your post but I thought that quoting you and going point by point would be slower than trying to explain my stance in a different way.

Do not fabricate blatantly false statements. Every owner of a Holdem Manager license wants to know the EV using the algorithm I described. Unlike PokerTracker, Holdem Manager uses the correct algorithm.

If you actually bothered to read the topic on the PokerTracker forums it will become quickly apparant that the PokerTracker developers don't even understand their own algorithm, let alone the one they should be using.

It is of use to you. It's just a more accurate version of EV. Imagine if PokerTracker changed their algorithm so that it would no longer calculate EV if both the cards you held were spades (for whatever reason). This change could not create a bias in the overall EV calculations, it'd just be less accurate by excluding a situation that should have had an EV calculation done. In that way same, I believe they are ignoring situations in multi-way pots which shouldn't be ignored.

Let's realise that in the exception of closing action pt does more than this it calculates ev at point beyond player going all in as long as all subsequent action finishes on that street and results in a heads up all in.

You've made a messed up scenario where the opponent simultaneously has a range of TT+, AK+ yet also has a range of exactly AA and you're picking and choosing between those two to make EV calculations look bad. You can't use the range TT+, AK+ to do the EV calculation then use the range AA to show how EV can be inaccurate. It's one or the other.

But in any case, people use EV because it is more accurate than actual winnings - nobody claimed that EV was perfectly accurate. In your example, EV clearly is shown to not take into account some information. But what you have NOT done is shown why using EV would be worse than actual winnings use for determining how good you are in that situation.

Wait a second - where did you get that idea? Can you explain to me, preferably using citations of wolfram mathworld, how being all in on the flop therefore means that the EV should be calculated as it was on the flop?

I did not think I would ever say this but that algorithm is WORSE than the one currently implemented in PT4. Their algorithm simply ignores cases that it should be calculated. The one you propose here would create biases and favor some playing styles over others.

In that scenario your EV would converge in the same way that actual winnings do, but faster. If you look at the long post I made in this topic you would understand that the average of the EV calculated would be exactly the same as actual winnings.

Do you believe that there is any case (a strategy tree) where my EV algorithm would converge to a different value than actual winnings? If so, show me it.

No but the present algorithm does as well. And your algorithm will necessarily always be at least equal or closer than present because it includes more information when it calculates. But if that was a measure then an algorithm that always calced ev on turn regardless of whether action closed on flop would be better again.

I own a Holdem Manager licence, and do not want to see the EV calculated using the algorithm you described. So your fabricated assertion about the HEM user base is blatantly false. Join the club I guess?


The page you linked doesn't mention flop, turn or river. As a poker player, I want to see EV calculated with undealt cards being unknown. I don't want to see it calculated for specific turn cards, when I made my last decision before the turn card was dealt. It's of no interest to me to know that my expectation increased because a particular turn card actually came.

I think people want to know the expectation that directly results from their decisions. I certainly do.

In your example, both the turn and river cards were unknown at the point I care about . Can you explain why you're happy for the turn card to become known, but the river card to stay unknown?

I don't know what you mean by this. It's almost as if you think the EV value is supposed to be as close as possible to actual winnings in the long run.

I don't think I can say anything clearer than my earlier example. If I get all in with 99% equity, I want my EV line to say the expectation was that I win 99% of the (main) pot. That reflects that I played well to get all in with 99% equity. If I lose the pot, my graph will show that I got extremely unlucky. Not that I played badly because the turn was the 1% card and there was a side pot.

By the way, your attitude in your posts is basically "I am right, everyone else in the world is wrong , and people who disagree with me are stupid". Whether or not you actually think that, that is how your posts read. And whether or not you are actually right, it makes you look like an ass, and antagonizes anybody who reads your responses.

I am happy with my calculations and have no interest in what you think any more (luckily the PT developers don't either) - so if I don't reply again it's because I had something better to do.

To be clear, this isn't an ad hominem attack, it's just honest feedback and explaining that you could find more people sympathetic to your cause if you spoke in a more humble fashion.

Suppose we have two AI players who are going to play heads up. We calculate the bb/100 of one of the AIs by looking at the entire strategy tree and it is 4.3bb/100. If they were to play each other, Both the actual winnings and the EV would converge to 4.3bb/100. (for this example I use AI because their strategies would not change (do not confuse this with having a strategy which has no element of randomness in it))

EV is supposed to converge to your true win rate. If EV and actual winnings converged to different amounts, then it wouldn't be your expected value would it? I expect my value of playing to be 4.3bb/100. If my EV converged to something other than 4.3bb/100 something would be wrong.

heehaw earlier in the topic thought he had caught me out with my statement that All-In EV should be calculated when no more decisions are to be made. He stated "If we apply that theory to the extreme, that means in hands where no players are all-in, we should wait til the showdown to calculate EV (since that's when no future decisions are to be made)." But this actually works perfectly. If we ignored EV in showdown situations on the river when no players were all in, our EV certainly wouldn't converge to our true win rate.

I avoided saying that EV does not converge to actual winnings because I don't think that's a very accurate statement. They both converge to your true win rate, but EV converges faster than actual winnings. Thus making it a more accurate estimation of your true win rate.

I know exactly what you were saying and it is wrong. The PokerTracker developers actually agree with me on this. Calculating your EV on the flop in the case of the AA on an ATT board, and villains had T9 and A4 situation would be wrong. They explain this here. But unfortunately the PT developers go on to jump to the incorrect conclusion that EV should therefore not be calculated in that hand.

I've been following this for awhile and here is my take.

A problem with both PT's way and yours is that the calculation they make ignore card removal effects. Your way will increase this. If you can live with that, then both are okay as long as both are always accumulating a result for an allin even when the allin is not called. If either does not, there will be selection bias. The intent of such calculations as you suggest are purely as a variance reducing measure. Ignoring card removal effects, your way will converge more quickly to an EV for going allin with the increase of sample points. Neither your way or PT's are useful in evaluating EV of a specific decision, since neither incorporate information of hand ranges for the allin bet and allin call decision. This can only be approximated with a very large sample.

Note that the use of databases can be very misleading. For example, after the completion of about 80% of the Dwan/Antonius match, Dwan was well ahead by about $1.6M. Somewhere on the FT site, they published (no longer there) the results of the 10 biggest pots. These went something like 7 to 3 in Dwan's favor, I noted that if they had gone 7 to 3 in favor of Patrick instead, PA would have been leading the match rather than TD. Despite playing nearly 40,000 hands, the determination of the match outcome was seemingly due to a measly 10 hands! IOW, which is the better player is anybody's guess.