inaccuracy of all-in EV

[Rapala]

The Problem With All-In EV (All-In Equity) | PokerTracker 4 Preview

I've just read the above article with some dismay as I've been basing how I'm doing on this line for quite some time so to find out this is overestimating my ev profit really sucks.

To try and get a more accurate picture, I've made a filter for only pots that are hu on the flop as well as only sites that show mucked cards. I know I'm now ignoring multiway spots but would I now at least be getting the right EV for hu pots?

[netsrak]

I'll forward it to my colleagues for a more qualified response.

[Patvs]

Both EV methods are flawed.
The article also states [it results in a] "smaller bias in the other direction (i.e., the user should theoretically appear slightly lucky rather than very unlucky), but a bias nonetheless".

We don't think the bias is smaller. They're very similar.
Using filters to filter for only pots that are hu on the flop will lead to even more biased results. Because it includes hands where you commit half your stack with the BEST hand preflop, then commit the rest of your stack with worst hand (being OUTFLOPPED) postflop.

This is one of the big flaws of "All-EV" HoldemManager uses when you focus on individual hands. But it really shouldn't matter much in the longterm.

Read: http://forums.holdemmanager.com/mana...ituations.html

[Patvs]

Udbrky also found this interesting thread:
CoTW: Why all-in-EV is a horrible measure of overall luck - Micro Stakes Full Ring Games - Micro Stakes Poker Strategy Forum



2+2 discussion on your question:
inaccuracy of all-in EV - Gambling and Poker Software Forum

[Rapala]

Thanks for the informative reply. I understand it shouldn't be used as a big guideline to playing strength but it's nice to know it is more or less accurate.

As they say, proof of the pudding is in the eating and I wonder if you've seen examples of players with million+ hand databases and whether the bb won and 'bb ev won' start converging? If there were enough examples of this it would show that it is by and large accurate, whereas if most large sample db's are skewed in one direction, we would know what the tendency is.

[Patvs]

All cashgame and STT databases: longterm the EV line and winnings line will converge... followed by some ridiculously long stretches where you're running above (or under EV)... and they will converge again, (repeat).





There are three exceptions: (when the lines will never converge)
-1 corrupt database (solution: reimport your hands into a new database)
-2 EV BUG in HoldemManager (when this is fixed, reimport your hands into a new database)
-3 If you play NL400 and are stacked being a 98% favorite. Your opponent hits his one outer. You lose 95% of your bankroll. And you play NL2 for the rest of your life. Actually you can filter this out in the graph by showing the graph in number of big blinds instead.

[Rapala]

lol number 3. Thanks very much for all the info, that graph def goes a long way to proving Mr Hwang wrong!