Fail often?

Or did you just make a special effort today?

OP, the red line debate has been done many times over, and views differ.

Red line discussion thread

Feel free to tell me how I am wrong.

I think they have said they don't calculate 18 mans and up correctly and that it should be disregarded.

(If you think about it, it makes sense because HEM would never have the stack size info from the other tables so it would be impossible to calculate ICM correctly.)

Isn't it obvious?

This couldn't be more wrong.

Yeah but it could figure out when you're at a final table and just use cEV otherwise. (If Stars would ever help out a little bit and put the number of entrants somewhere in the HH.)

I explained why I posted that. Do you have any better explaination for why I am wrong?

You failed at this one too.

Meh, it's hard to just "disregard" 18-mans because it looks like alot of the 18-mans for which I lose at the 1st table show up as 9-mans in HEM, while I think most of the final table ones have more players in HEM. If they just use chipev for first table, that would be fine. But not sure if they do this.

That is exactly what happens. Your losses go down as 9-man losses - for 18, 45, 90 - any mutli-table SNGs where Stars doesn't put the number of entrants anywhere in the HH. But by the time you get to a FT, HEM figures out it's some kind of MTT and ignores it. So the red line incorporates a lot of the bad early bustouts (as 9-man bustouts) and none of the big $EV FT wins. Unless you have some way of filtering out your 18-man (and up) HHs altogether, your red line will always be too high if you play a lot of them.

Whenever I want to see if I'm really on a lifetime cooler I just look at 6-max, and see my red line is generally where it should be for those.

I think HEM could make assumptions based on the exact # of chips at the table:
27k = 18man
67.5k = 45man
270k = 180man
and apply $AIEV just at final tables?

This might not be the right place for this question; but my only alternative is STTBBV and getting a serious answer there isnt very easy.

This is a question about 9mans. So your varaince, size of up and downswings, is mainly a function of your roi. The higher your ROI, the less severe your downswings are gonna be. Is there the same relationship between your ROI and the worst under EV stretches you will experience. Say someone with a 5% roi has played 50k games, and the worst 1k game stretch theyve had is 75bis under ev. Would someone with a 10% ROI over the same stretch expect there worst 1k stretch to be more, equal or less BIs under EV than the 5% guy?

I think the answer here is no. IMO how well your redline correlates to your actual winnings is purely a function of how lucky u got in all in situations of the tournament. Having a lower or higher ROI should not affect the correlation.

So I guess the only thing that would affect how swingy in comparison to your greenline the redline is, is the format you play. The more severe the bubble factors are in the format you play, the more severely you can get ****ed over equitywise when you run bad. Guess thats why they say the formats with the harshest bubble factors (DONs and hypers) arent well represented by the redline.

Thx for the response wayneking, knew I came to the right place.

Yeah, alot of people are under the impression that the redline for games with flat payout structures such as DON's and hypers is not very accurate because of the "check it down" strategy and in these situations no ICM calulations can be made. But i think this can go both ways and there is no reason why this wouldnt even out.

However where i think there may be a problem in the hyper format is in the fact that ICM grossly over-values short stacks which of course is how u spend most of ur time in these.

Variance/stdev is a function of finish distribution. At 16s player A is 15/13/13 for 18.1% and stdev 1.57, B is 13/13/13 for 9.7% and stdev 1.52. A will have bigger swongs vs his true winrate but my guess is that he has smaller drops vs red line cos of how it works (his green is steadily going above red). Playing style will also probably effect how much you swing vs red line cos if you never go to showdown red won't be an adjustment towards true roi, but I'd guess that due to the correlation between % of 1st and stdev it will basically mean that players with higher 1sts/roi will deviate most from the red line but due to high roi resulting in green>red that probably wouldn't mean running more below red than a lower roi player (assuming pretty similar finish distros)

My immediate reaction to this was "everything you just said is wrong." There are certainly limitations to ICM, but your post is just a broad overarching statement that doesn't have any real substance. Part of what being a good sng player is about is knowing when and how to deviate from ICM. However, even that deviation still uses ICM as a basis point. Knowing its limitations allows you to bend the rules of ICM for a specific situation, but that does not mean that it is not vital to a winning player's arsenal. You definitely need to provide an actual hand before what you say can be given any credit.

The point of my post here wasn't saying that your hot cold hand equity wasn't useful. It wasn't saying that the icm value of your stack when calculating the value of the hot cold equity vs result is less useful than the chip equity value. The point of the post was just saying that in many situations there is a clear answer that is different than the conclusion an icm solution finds. If this is a situation where the correct play is to call, hem will show this as a losing decision in your expectation graph.

I remember reading (a while ago) how the red line should be as if you were watching your green line on a 4 times larger sample.
Do I rememeber that wrong, and if not, is that still considered a good rough approximation?

i'd like to know that also. thx

IIRC someone calculated the amount of 'luck' attributed to all-ins and it came to about 70%. So you could say 1000 games of EV line is equivalent to a regular sample of 3334.

But be careful, the red line has some very specific tendencies. I believe Juk touched on this at some point in this thread. In my experience for example in super turbos if you are losing all ins in the very early stages your green will plummet and your red will be mostly flat. Conversely, if you have a really good run over a few hundred tournaments your red line will shoot upwards with your green typically albeit a slightly less slope. Again this is just personal experience, but if you are way under EV over a sample, your 'true' ROI should be a little above your EV ROI. And if you run way above expectation over a sample, your 'true' ROI should be a little lower than your EV ROI. Again this is my only my opinion, but it is based off of an extremely large number of games played, and by someone who puts a lot of stock into the red line.

Drew, what you think about the possible inaccuracies of the redline in superturbos. One of the shortcomings of ICM is that the overvalues your ICM when youre in early position and about to be hit by the blinds. I know that you should be making a lot of from EP that wiz is gonna say is bad. I dont know how and if that would affect AIEV.

Im just curious cause I've played 22k 6 and 9 man sup turbos and have never had my greenlind above my red more than a few BIs, and am currently about 240 BIs under EV. Do you think this is just superrunbad, or does it show a flaw in the reddline at STs. I dont see how after a 20k game sample, being that far under true EV could be possible.

Well think about this: the standard deviaton in STT roi is 160/sqrt(n). For 22k that comes out to 160/148= 1.08. So there is a 95% chance that you are within 2.16% of your true ROI either way. To think that after 22k you could be 2.16% above or below your true ROI is pretty sickening (This assumes identical buy-in obviously).

Yes it would effect your real equity in a tournament differently than ICM says. But again ICM is just a model, and an imperfect one at that. But it is the best model we have to go with. These shortcomings should even out over the long haul, i.e. sometimes it will say I should have lost more equity than I did, sometimes less. But you will be put in that situation close enough to equal times over the long run that it will even out. Which is why I cautioned the poster above about using the redline after such a small sample. I know I have been 200 BI's over and 300 BI's under EV at different times in my redline. I actually went from roughly 125 over to 100 under in a span of about 750 games, it really was sickening. FWIW 240 BI's over that sample is run bad IMO, but not horrible. I believe UConn posted one worse at higher stakes recently in a much smaller time frame.

Plus like the equation shows you is that in terms of being within 1 or 2 standard deviations of real ROI does not increase all that much as you increase the sample past a certain point. Simply because a massive swing one way or the other before hand may never actually even out. If you were to bump it to 40k games your still only within 1.6% 95% of the time. Which is why playing with a real low ROI can be a real test sometimes and I applaud the guys that do it.

I forget where he posted it but Juk posted his massive sample an although it swung back and forth IIRC it ended up right around his true ROI. There certainly could be some slight flaws like it overvalues short and your finish may be like .14 3rds, .13 1sts and .12 2nds. but not to the tune of 240 BI's.

I'm terrible at staying on point and sounding clear, but did that make sense?

Will it show redline filtered by position?

Dont think so. On the profit/AIEV graph, I think the only filters that exist are for date and player. It can only show redline alongside profitlines, which only come from complete games.