Interesting topic. I play HUSNGs, and we also have (expected) bb/100.
Now since we only have SB and BU the std dev for them is around 54. Is this true? Does this mean HU is more "accurate" than 6max or something which is on avg 95? How is this std dev being calculated?
Originally posted by VorpalF2F
To put that simply, If variance is low, must 100k runs will have a mean close to the overall mean.
If variance is high, then some 100K runs will be far better than normal and some will be far worse.So "variance" is not synonymous with "luck", but is better read as "the range of possible results that should be considered normal"
Cheers,
--VS
Yes that's correct. But there are other finer points. HEM or PT4 can never tell you what your "true variance" is. They can only model it.
What they do is simply take every single sample of 100 hands from your database and see where your winrate for that 100 sample lies.
If the number of hands is high enough, then those samples (1k samples if you have 100k hands) will be shaped like a bell curve (there will be very few samples around -130 bb/100 and +130 bb/100, but very many samples closer to 0 bb/100).
The worst 0.15% of winrates are the -3σ winrates (or worse)
0.15% - 2.2% -> -2σ winrates
2.2% - 15.8% -> -1σ winrates
15.8% - 84% -> within standard deviation
etc
And so, in essence the software takes all the winrates from all the samples, lines them up, takes the winrate that lies 84% down the line, subtracts it to the winrate that is 15.8% down the line, and calls that the "standard deviation" of your winrate.
That's the best estimation that can be done, but it is not you true standard deviation. Over the course of your database the stakes change, your skills change, the field changes, and so on.
Originally posted by legand73
I'm thinking though that maybe i should be thinking about what part of villains range i'm running into when i play a certain line cause that is a part of variance is it not? say if i run into the weaker part of his range and win a big pot then that's an example of good variance?
Yes that is certainly a part of variance that's not part of the EV line. Another factor could be how lucky you got with the flop, turn and river runout (for instance, you're very lucky if an opponent "has" to call your bet/bet/bets with a marginal hand because of the board runout while you have a value hand).
That kind of variance can be a factor if you're a NL100 grinder and you got staked to play NL1K in a fishy field and suddenly you run bad. Otherwise, if you always grind the same limits, that too will even out.
The problem with spotting this kind of variance is that even when you get good at recognizing it, you can't really be objective and add everything up at the end of the session to figure out exactly how lucky or unlucky you got the way you can do with all-in EV. You can do it only to some limited degree. So you have to be balanced and careful not to use it as an excuse for losing while you ascribe all your successes to skill alone 
Originally posted by ragney
Interesting topic. I play HUSNGs, and we also have (expected) bb/100.Now since we only have SB and BU the std dev for them is around 54. Is this true? Does this mean HU is more "accurate" than 6max or something which is on avg 95? How is this std dev being calculated?
The standard deviation is often higher in heads-up then 6-max because you play more hands so there are more of those big pots that differ from your winrate.
However, your standard deviation could be lower is HUSNG than cash with 100bb stacks, especially if you play turbos or hypers, because the stacks are so shallow. Standard deviation describes how much the possible outcomes differ from the expected value at average - if the stacks are very small, then there's no possibility for the outcomes to be as far from the expected value as there would with big stacks.
Example: You pick randomly either 2 or 4. The average would then be 3, but you can't choose that. Both 2 and 4 differ from 3 by 1, so the standard deviation would be 1.
Example 2: You pick randomly either 0 or 6. Again, the average is 3, but now the difference of the possible outcomes to the average is much bigger - both differ by 3 this time, and therefore the standard deviation is 3.
Example 3: This doesn't have to be symmetric and there can be more than 2 outcomes. You could have a 25 % chance to pick 1 and 75 % chance to pick 5. The average is then 0.25 * 1 + 0.75 * 5 = 4. This time the outcome differs by 3 with a 25 % probability and by 1 with a 75 % probability. Therefore, the standard deviation would be 1.5.
Oh and I forgot to add - there's also standard deviaton per game in tournaments. In HUSNG it's close to 1, but depends on your ROI (more likely on the probability that you win the game).
Originally posted by HuhtalaJ
Famous Finnish people you might know:
Isn't Shevtshenko Ukrainian? 
Originally posted by outspan
Originally posted by HuhtalaJ
Famous Finnish people you might know:Isn't Shevtshenko Ukrainian?
The football player yes, but the poker coach is a Finn
Originally posted by HuhtalaJ
Originally posted by ragney
Interesting topic. I play HUSNGs, and we also have (expected) bb/100.Now since we only have SB and BU the std dev for them is around 54. Is this true? Does this mean HU is more "accurate" than 6max or something which is on avg 95? How is this std dev being calculated?
The standard deviation is often higher in heads-up then 6-max because you play more hands so there are more of those big pots that differ from your winrate.
However, your standard deviation could be lower is HUSNG than cash with 100bb stacks, especially if you play turbos or hypers, because the stacks are so shallow. Standard deviation describes how much the possible outcomes differ from the expected value at average - if the stacks are very small, then there's no possibility for the outcomes to be as far from the expected value as there would with big stacks.
Example: You pick randomly either 2 or 4. The average would then be 3, but you can't choose that. Both 2 and 4 differ from 3 by 1, so the standard deviation would be 1.
Example 2: You pick randomly either 0 or 6. Again, the average is 3, but now the difference of the possible outcomes to the average is much bigger - both differ by 3 this time, and therefore the standard deviation is 3.
Example 3: This doesn't have to be symmetric and there can be more than 2 outcomes. You could have a 25 % chance to pick 1 and 75 % chance to pick 5. The average is then 0.25 * 1 + 0.75 * 5 = 4. This time the outcome differs by 3 with a 25 % probability and by 1 with a 75 % probability. Therefore, the standard deviation would be 1.5.
I understand abit now, thanks.
The std dev I wrote was the one I filtered from PT4, sample size about 60k hands. Was surprised that its lower then the one from 6max since the variance in Hyper HUSNGs are pretty big. My avg win per game (ITM) is ~53.5% which equals to 4-5% ROI I think.
On this topic.
I was trying to find out some more concrete stuff on the variance of an MTT grinder. Hint, it sucks. You can be doing pretty well, ROI of 20% and still expect to lose money after 5k tourments 13% of the time.
Here's a great blog explaining variance in large-field MTTs, smaller-field MTTs, NLHE 6-max cash, and 6-man and 9-man STTs.
Shame he hasn't done PLO... uuhhgggg PLO mtt's....
