How exactly is your variance calculator implemented, because leaving the input and clicking 'update' a few times can generate VERY different results.Hypothetically inputting for a ~$40 HU SnG over 1000 matches @ 57% winrate, I see upto $4.5K winnings and down to $750 loss! - Some graphs are quite steadily increasing without many swings, some have huge swings... Some swing up-down, others down-up, some swing both ways to end almost even.Its almost like it is randomly (but weighted by winrate) generating 1000 results, and tallying/graphing them?For say 57% winrate, obv over the 1000 games 570 will be wins and 430 loses, but the order is purely random ?? - which doesnt seem to accurate (or logical) to me, as a programmer as well as poker player. (Note: not that I have thought about this problem to figure how it *should* be done - maybe base streaks on winrate also?)
Obviously something wrong with it, can't believe nobody realised before :)
Hello AJG,It is a variance simulator, so each time it generates a given outcome given the winrate. It does not give you the standard deviation or some sort of statistical number for you to reference. Its purely to get a feel for how the variance is for a given winrate and number of games.
DonNew: Yes, I know that, but thats my point... It doesnt!The variance implied by the graphs can be hugely different given the exact same input.Try it... especially with a marginal winrate. Try like 55% and keep the same values, just click 'Update' and notice the next dozen or so graphs generated. As I said, 1 will show a huge profit and rather steady growth, another will have extremely wild swings and end up near 0, even show a loss. And everything in-between.
well thats why they call it variance :)when I said a given outcome i did not mean the same outcome each time you update. but if you have a 55% winrate it translates roughly to 5% roi, so if you do the simulation over lets say a 1000 games or more it will most of the time show a nice profit. you really need to put in a lot of volume with a marginal winrate before it approaches your expected long-term roi.. and thats even assuming a constant roi over the entire simulation.
Yeah, but the results are wildly different given the same input paramters.And I understand variance, I was using it to investigate the profitability of SO of 4/8 players when i noticed this first (although it still happens for STTs)If I plug in some numbers and get a nice looking graph 1st up then that is misleading, when clicking update again might show me breaking even, or even down, over 1000 games... Perhaps the page should read "Click update a few times to get a fell for how variance can effect your results" ??Also, if I plug in 10,000+ games the 1st 1000 NEVER look as wildly different as only inputing 1000 games... Scale considered...Just seems somewhat misleading to me, as the 1st graph is most often on of the 'better' ones ;-)
To be honest, it's not even as variant as it should be.Why? Well the calculator assumes you have whatever winrate you input over each game I believe. So if you put in a 55% winrate, it will assume you have a 55% chance to win each game and give you a sample of what will happen.In reality, you might have a 60% edge against one player and a 50% edge against another. For 100 games you might have a 53% winrate, for another 100 games a 57% winrate. Etc. etc. A few other things:- Longer samples will look more similar as variance tends to be less variant over the long term.- The 1st graph shouldn't be a "better one" by any stretch. Make sure you hit "update" and then count that as your first graph.- You'll get some sick nasty different graphs. Variance is insane, most people don't understand that, even massive winning professionals.- Outliers make sense, ever wonder why the guy in a super turbo with a steady up and down graph is so good? He's certainly good, but he's most likely running like god as well, as the edges are too small for anybody to have a straight up and down graph for a truly long sample (10k+ games).- It is calculated just like flipping a coin. Say somebody came to you and put in "50%." Then you flipped a coin every time they put hit "update" and you tallied down that result. You might have 3x in a row where it's tails, or a back and forth for 10 updates. You certainly wouldn't expect it to rotate nor to have the same results over samples of 5, 10 or even 100. So each input just gives a sample result, it could be the very same result it just gave or something totally different. Put something out of reason like an 80% winrate in there and you'll see much more similar results from time to time, as the room for variance is much smaller I believe.