At least somebody has their priorities straight.

AIEV line in HEM is your winnings line, but with equity adjusted values for all-ins instead of actual values.

So it consists of 3 parts:
-Showdown winnings from hands that weren't all-in before the river
-Equity adjusted winnings for hands all-in before the river
-Non-showdown winnings
(in other words: Winnings + EV $Diff)

I agree that the term All-in EV they've decided to use is a bit misleading, "Winnings with equity adjusted all-ins" or something like that would be more exact.

However if the line literally showed just the EV of all-in situations like the name misleadingly suggests, it would be completely useless because not all showdown winnings are from all-in situations. You'd have to filter your actual winnings with "all-in == true" filter and then compare your All-in EV to that number just to get anything useful out of it.

So a question. This seems to happen quite a bit I think or I used to think so. US and not playing...
Anyway, we have top set, pot pot and on the river there is 5bbs left. The obvious flush comes in and villain donks. We would be stupid not to call, but when we do we see that we are beat. Our Allin ev shows we had 0 ev but obv it would have been ridiculous to fold.

Is this addressed some way??

That spot has nothing to do with AIEV, it's only supposed to remove variance from all-in situations. That is not an all-in situation. So, compared to winnings AIEV does nothing in that situation. Obviously there can't be any bias, when nothing is done. Questions like that are a prime example of how the purpose/capabilities of AIEV are often misunderstood. It DOESN'T tell you the EV of plays you make or "how much you should've won" - it simply removes a (small) part of variance from your results, which is still better than no adjustment at all.


85% of the time he misses that draw and it's gonna look like you won the pot with 100% equity, which balanced that out in the long run. That's why in the long run variance will even out even from those situations.

OK, clearly I don't understand what defines an allin situation. I could definitely use some clarification.

Going back to our example. The river bricks, we put in our last little bit and since he can't beat anything, he folds. If HEM doesn't see his hole cards, doesn't that go into nonsd winnings and not into an allin ev???

So two questions.
Thanks

As I said in my previous post, non-showdown winnings are also included in the EV line, so it goes into both lines. So in that case the whole pot will go into your EV line. The line in graph called all-in EV is simply all your winnings, just with all-in results replaced by equity adjusted values.

You should always think of that EV line just as winnings line where variance have been removed from SOME situations (all-ins), but not from all situations.

Sorry for my denseness here, but thanks for explaining. So you said my example was not an allin situation. What makes a hand Allin?

I do understand that when we win a pot , we almost never had 100% equity and it adjusts for that.

it's pretty simple,

AIEV is exactly the same as the winnings line, except pots where there was an all in before the river, the winnings line is adjusted to reflect hand vs hand equity rather than the result. Even if its 1c in a million dollar pot.

also your all in ev on the river is 100:0 or 50:50 always

both lines are complete random bull**** relative to the true ev line, which is all that is important and hem has no way of showing this whatsoever. The AIEV is on avg more accurate than the winnings line but you can never know if it is more or less accurate than the winnings line in your sample.

'sick graphs' are often very exaggerated because of this.

You could imrove the AIEV if you could stipulate or limit it to a single raised preflop pot and the AI bet is a 3 or 4 bet (psbs).

The thing that remains counter-intuitive to me is how non-SD winnings are included in the EV line. How does HEM calculate that exactly? there must be a fairly simple equation involved.

what's variance?

It's similar to the rakeback line.. it simply adjusts your cumulative winnings figure at each hand:

AIEV line = [winnings WHERE nonsd = true] + [winnings WHERE nonsd = false] + [winnings AIEV adjusted]

edit - [winnings AIEV adjusted] = the column in HEM that says Ev diff or whatever is noted below V. It probably comes across differently the way I wrote it.

aiev only recalculates the results for pots that got ai before the river. It does nothing with pots that didnt go allin before river, thus nonsd winnings have no part to play here.

They are simply added to that line. AIEV line is simply the winnings (= green line = both nonsd and sd) corrected with the amount of "EV Diff"
(EV Diff = the difference between actual results IN ALL-IN SITUATIONS and equity IN ALL-IN SITUATIONS)

Just like the "rakeback" line doesn't actually tell you how much rakeback you've got, it tells how much your rakeback + winnings are total.

I admit to being someone who agonizes over the aiev line a lot, and in the absence of having a coach to critique my game, tend to use it as a barometer of how I'm doing over medium sized samples.

I'm posting this graph for lulz, but mainly because I'm interested in whether people think it is reasonable to conclude anything based on the aiev line (instead of actual results).

FWIW, this is all the hands I've played on stars since I switched to PLO post-black friday.
Stakes are about 70% PLO50, 30% PLO100.

Comparing the first 125k hands to the last 125k hands (and ignoring the horrible spew in the middle), in actual results I'm up roughly 60 bi over each 125k hand stretch, but in terms of aiev I'm up about 20bi in the first 125k hands and about 90bi in the last 125k hands.

I'd like to think that the improvement in my aiev-adjusted winrate (where the actual winrate is about the same) is suggestive of improved play. Feel free to tell me I'm wrong... I'm ok with just running good as long as it stays that way permanently

I wish pt3/hem had an "effective EV" analysis tool:

Pre-Flop Bet*Pre Flop Equity + Flop Bet*Flop Equity + Turn Bet*Turn Equity + River Bet*River Equity = Effective EV

^^ spot on

aiev is just an indication so are actual winnings. Also note that aiev is calculated when at least 2 players are allin. So If you're sitting with 100bb and the other 2 players are playing deeper the aiev is calculated when another/both player are allin. When for instance they check it down and you lose at showdown your aiev would be 0, when you did have equity.

who's commentary are u watching?

+1. Effective EV would be a good equalizer to let everyone have a cool head like Roope references (even running @EV and winning big =/ playing great necessarily).

srsly... ?

the AIEV line is exactly the same as the winnings line, but if there is an all in and a card to come, it adjusts the winnings line by the equity instead of the result. non SD winnings have no consideration whatsoever. It only adjusts for the equity in all in pots.

hunterSC, force, sc2strategy, lifesaglitchtv, HDstarcraft

: / srsly I haven't even ****ing played this game or the original but for some reason I keep watching the replays. Use to play a tonne of red alert online competitivelly so I think thats why

Jeff W and provotrout,
unfortunately street-by-street EV is a biased way of calculating EV, it doesn't tell you anything real. That's why it never got popular although there was program able to calculate it.

If two players get 70% of their 100bb stacks in as 85% favorite but then villain sucks out and shoves the river. Let's look at the situation for both players:

a) Player 1 makes a super disciplined fold for for that last 30bb shove, as he figures there's nothing he can beat, since every draw got there and villain's line is super consistent with a draw. He folds, which means there will be no showdown -> SbS EV would assume he got all his money (70% of stack) in dead, so his SbS EV for the hand is -70bb.

b) Player 2 gets stubbord and doesn't want to when villain shoves the last 30bb into a 140bb pot, even when he can't really beat anything reasonable at all. There will be a showdown that shows he got his money in with great equity on earlier streets. So his SbS EV for the hand would be +89bb


This leads to a massive bias in street-by-street calculated EV. The player who actually played better and made the correct fold after getting a lot of money in as a favorite will get -70bb EV, while the calling player who gets frustrated and makes the stupid crying call gets rewarded for +89bb in EV.

This is clearly a bias and not variance, since you can cheat EV by calling down more eg. in situations where you had nuts (or any hand with great equity) in earlier streets and there was betting. You could even do this with missed draws. If you had a massive compo draw with wrap+fd and turned another fd on the turn, you could call a small river bet with Q high to get better EV for the hand than you would get by folding... as long as equity*bets on earlier streets is more than you are calling on the river.

I'm not 100% sure, but I don't think you can make unbiased EV adjustments in any situations where there is still play left in the hand (I mean if players can still act on later streets)

Seriously, I love this.

This is true, but it's also true for non-showdown winnings. Every time you fold, your red line goes down and every time you make bad bets/raises etc you're freerolling your red line at the expense of your overall winnings. Yet non-SD winnings are still a very useful tool.

Nope, not really. It's true that you are able to increase your nonshowdown by sacrificing SD winnings, but even in those cases the lines represent what really happened. However, in SbS EV's case that is simply statistical bias caused by faulty EV adjusting method.

They are completely different things.


fwiw I think that showdown and nonshowdown lines are by far the most overused stats in HEM. Everytime someone (paging crashwhips) posts a graph with positive redline and 5bb/100 winrate people go crazy "ZOMG WHAT A SICK REDLINE YO" and fail to realize that his winrate is pretty usual for a good regular at those stakes. Then you see people posting >10bb/100 graphs with a losing redline and wondering "what am I doing wrong?" when in fact they are winning with a higher rate than the redline guy is. Just today saw this (for the 1000th time) in the newest redline thread at HSPLO.



(edit: crashwhips is a really good player, not trying to say anything else. just find it weird how his graphs draw way more attention than most graphs with actually higher winrate)

Like Nikachu said: Redline down ? fk that!

It also depends on the Stakes you play. At donkfest tables where people never fold (yeah especially omaha at micro and midstakes) good luck keeping your redline up.

As long as your winrate is high why would you even care ?

But back to topic: I have problems with variance too, im basicly running absurdly UnderEV over a pretty big sample size. Its not very easy to deal with especially when you know you have an edge over most of the competition.

But its Omaha right ? Where a monkey with a clownhat can luckbox 10 buyins on one single table and a good player looses it all to him by just pushing his 85% 65% 73% hands against an aggro monkey not winning a single all in. And then to get barrated by the same guy being a "loooooozzerrr".

I made a thread about variance too and people gave good advices, most of which i allready knew but its allways nice when people back you up.

- evaluate your play - 20 buyins over 1k hands is a leak, you said yourself due to bad play

- take short breaks if you get coolered or feel tilted
- ever considered a stoploss ? 5 buyins
- step down a level and get confidence up again

i complain about variance too... i think you allways do if you are affected by bad variance and feel helpless even tho you know you dont play bad. if you play bad, just stop, review hands, get advice, improve your game and then continue.

variance is a deviation from the norm its a natural phenomenon and will allways be there. you dont complain about gravity do you?

its all easier said than done, i almost gave up playing yesterday after losing like 6 huge all in pots where i was a 95% fav in the hand against a maniacish donkey, but this game is still fun for me.
and getting my money in good most of the time, makes me feel confident. bad variance fks me up at the moment, but what can i do, im a stuburn idiot!

I'm a big fan of nikachu and I really respect him for openly being the biggest nit ever (and for the most part, justifying it really well)

Even nowadays so many people don't seem to understand that adjusting basically means is going from balanced strategy to exploitable strategy. The more you adjust, the more exploitable your strategy becomes. For example never making big bluffs is a really exploitable strategy, but it's definitely the correct adjustment against 90/55 players who'd never fold a pair.

It's NL but whatever:




PS. Who do I have to bribe for "Exploitable? LOL" undertitle?