I mean obviously you could be playing bad, but a 3k sample is pretty small by today's standards if your ROI is 5% or less. If it's still that way after 7-8k games, then I'd start to worry.

HM1 redline is flawed, it is even stated on the HM website.

They are in a debate on how to fix this, from fozzy in the HM support thread:

"There are some bugs in the EV calculation in both HM1 and HM2 currently depending on which site/games you are playing. Part of the holdup on the HM1 EV issues is related to the fact that we use a lookup table for EV which isn’t as accurate as full enumeration. Full enumeration seems to hurt import speeds pretty significantly, so were looking into a couple of options. We are considering a more accurate lookup table but this is a lot of work due to the combinations involved and it will be a while before the developers can implement it. Another option is giving the user the option of full enumeration or lookup tables but changing this would require a reimport and choosing the full enumeration, while it would be the most accurate, could slow import speeds significantly which would affect HM1 and HM2 both. Before we can implement this we first want to explore how accurate the new lookup table is."

holy cow, this thread has 109k views

Hey guys, does anybody remember who the poster was who posted a graph for the hypers last year where they ran insanely below EV (it may have been in one of the PG&C threads or in another STTF thread I saw it - was towards the end of last year IIRC)? If so how many buyins was he below EV (and over how many games)?

Also has anybody else you know of posted a graph showing 200+ buyins below EV?

Juk

I think Wayneking7 has the worst I've seen; like 1000+ BI < redline. Pretty much anyone that has played a lot of hypers/supers will have a 200+ in there somewhere.

Here's my 6-max 10bb super-turbos (Merge).

Thanks!

On your graph between 13k and 14k games you ran about 200 buy-in below by the look of it?

That makes me feel a lot better then as I'm down about 250 buy-ins over the last 3k games (10-man 50/30/20) and I doubt the standard deviation is that different for the 6max games.

Juk

EDIT: I've just written some code to work out the percentiles, etc via monte-carlo simulation (very slow but will post when it's finished executing), but after seeing your graph I doubt it's gonna be much worse than a 1% event.

Yep, here's that area pulled out (starting from the greenline peak). At the time, I would have sworn it was an act of unholy demons.




Looking forward to the sim results.

Here's the results after 1M simulations:

ActualPrizeEquityDifferenceSum : -2638.87 (-263.887 buyins)
ActualChipEquityDifferenceSum : -736272 (-368.136 buyins)

Your $EV percentile : 0.042%
Your cEV percentile : 0.762%

This is over 2987 games so assuming the standard deviations are nearly the same for both our games types, it looks like running ~250 buy-ins below EV over 3k games is about a 1/2000 event.

(just for interest I'm re-running it on my luckiest month ever to see how that compares)

Juk

Over 3353 games:

ActualPrizeEquityDifferenceSum : 1785.58 (178.558 buyins)
ActualChipEquityDifferenceSum : 610750 (305.375 buyins)

Your $EV percentile : 98.501%
Your cEV percentile : 97.055%

Juk

Thanks. Nice to see something more accurate than the ROI Simulator method.

Still not sure what to make of ActualPrizeEquityDifferenceSum converted to buyins, if anyting.

For the record, the numbers for graph I posted are:
ActualPrizeEquityDifferenceSum: -261 buyins
ActualChipEquityDifferenceSum: -405 buyins

Some similarity to yours.


Looks like you'll have to keep waiting for that 1/2000 upswing. I did manage to run about 200 BI > EV in 1600 games recently, which felt pretty special.

Sorry if this has been asked before but which red line is more accurate HEM2 or PT4. Only just started playing Hypers but over the last 125 tournaments HEM2 red line is -$11 and for the same tournaments PT4 is +$1 (micro $1.50s). The green line is identical +$24.

PT4 uses a different formula to calculate EV (they call it All In Equity I believe). There are some flaws in the current HM2 calculations that we hope to address in a future update but I don't have an ETA for when those changes will be made live. Once the changes are made live in HM2 they should also transfer over to HM1 since they share the same parsing code and EV is calculated when the hands are parsed.

There are some bugs in the EV calculation in both HM1 and HM2 currently depending on which site/games you are playing. Part of the holdup on the HM1 EV issues is related to the fact that we use a lookup table for EV which isn’t as accurate as full enumeration. Full enumeration seems to hurt import speeds pretty significantly, so were looking into a couple of options. We are considering a more accurate lookup table but this is a lot of work due to the combinations involved and it will be a while before the developers can implement it. Another option is giving the user the option of full enumeration or lookup tables but changing this would require a reimport and choosing the full enumeration, while it would be the most accurate, could slow import speeds significantly which would affect HM1 and HM2 both. Before we can implement this we first want to explore how accurate the new lookup table is.

Yeah, sadly the ROI simulator only displays to the nearest 0.1% so extremely rare events won't show up.

As for using my results then it depends on a few things: how similar the standard deviations are for the payout structure (eg: for DoNs/F50 it would be quite different, but for 10max and 6max it's likely not that different), how similar the number of all-ins per game is (Party's turbos play a lot like hypers after the first 3 levels) and to a lesser degree the "structure" of the all-in meet-ups (eg: if you go nearly all-in a lot and then push the flop and/or have a lot more very wide ranges clashing because of low effective stacks, etc). I can't say for sure but I doubt any of these will be all that different though.

I can be pretty certain of my own results though because the monte-carlo algorithm is just a slight adaptation of the algorithm used to calculate all-in EV, eg: for each all-in in each monte-carlo rollout, rather than subtracting the actual outcome's equity that occurred from the expected equity, you choose an outcome probabilistically and subtract that instead and so on. Then you can just count how many times you simulated a run worse than your own actual run. Somebody good with stats could prolly work it out numerically instead (ie: something like the binomial distribution but with a variable p, etc).

Well it's not happening tonight, another session, another ass-whoooooping:

I'm trying just look at it like seeing something rare and unusual like an eclipse or a comet, but it is getting old fast now... A few more days of this and I'll be sitting in a faraday cage wearing a tinfoil hat!

Juk

Hello guys!
I started grinding the 180s this year and have been running pretty bad. Can someone please explain to me how the redline is calculated?

Number of Tourneys: 2135
Net Winnings: -1795$
$EV: +10198$
ROI: -14.89

Whats the difference between luck adjusted Winnings in the graph and $EV in stats?

[IMG] Uploaded with ImageShack.us[/IMG]

Thank You

The red line is calculated based on ICM, which is impossible to do so in case the tournament have more than one tables, for 180s see cEV.

Can anyone explain why the $EV diff does not match up exactly with the luck according to the ICM chart in HEM?

EV Result Luck Hand Player

$133.20 $5.30 ($127.90) AsAh Villain
$147.44 $215.00 $67.56
$154.66 $215.00 $60.34 6d8d Hero

hand order AsAh > 6d8d probability 78.0259%

0 stack 1310, icm ev $168.77
1 stack 805, icm ev $128.73
2 stack 885, icm ev $137.80

hand order 6d8d > AsAh probability 21.6552%

0 stack 0, icm ev $5.30
1 stack 805, icm ev $215.00
2 stack 2195, icm ev $215.00

hand order AsAh = 6d8d probability 0.3189%

0 stack 655, icm ev $116.17
1 stack 805, icm ev $136.64
2 stack 1540, icm ev $182.49


Basically everything is accurate, and if you sum up the 3 probabilities / EV's you indeed get a luck adjustment of $60.34 for me for the hand. However in my HEM reports page, the hand is listed as having a $EVdiff of -$60.81

What is the cause of this being off by $0.47?

HEM doesn't factor in the specific suits, but uses some generic AA v. 68s equity instead.

In your hand there's no Ad redraw, so your actual $EV is slightly better.

Are you sure that's the explanation? As you can see, HEM shows what it thinks my equity is in the ICM readout, which actually lines up with factoring in suits.

equity win tie pots won pots tied
Hand 0: 21.815% 21.66% 00.16% 370802 2730.50 { 8d6d }
Hand 1: 78.185% 78.03% 00.16% 1336041 2730.50 { AhAs }

So why would the ICM readout give me the results with suits factored in, but then the final result be without suits factored?

This is a known issue.

HMONE-127
Luck Adjusted Winnings wrong on PS

We hope to have it fixed in one of the next 2 HM1 releases but I don't have an ETA for how long that will take.

Oh ok, thanks. Well out of curiosity, do you know if it is wrong in any specific direction? I ask because I have consistently been running above EV and am curious as to whether I'm running hot or the red line is just plain broken.

I really can't say for sure. EV is not my strong suit and it honestly makes my head hurt.

If you are playing Hypers by chance and have some time to kill you can read the posts in the following thread by Patvs on our forums as he is our resident EV expert on the tech support team.

Yeah that was me unfortunately

Can you run a sim on my results? But that about half 6m hypers and half F50, so std dev obv diff for the 2 but im sure we can get an approx. figure

This is just sick ^^^^

Yeah its been a fun 16months, remember thinking before this that running 200BIs under EV was ridiculous

Is it better or worse in abis?