Quote:
Originally Posted by coon74
I've read a blog entry 'How much variance comes from AI situations?' and my hair stood on end. Dear InsuredPlay owners, either you're deliberately hiding your knowledge of math, or you urgently need to learn the definition of variance (wiki) - the variance is equal to the mean of the square minus the square of the mean - and hire a mathematician, preferably a qualified risk manager. It's sad that Quad, a respected blogger, gave a wrong calculation too.
Also, your bubble insurance has a whopping collusion breach. If one of a gang of six players insures a series of 6-max SnGs but the other don't and flip stacks between each other to ensure that the first player finishes exactly in the third place all the time, and then the winnings are divided evenly, it's clear that IP loses lots of money (the team gets 5/6~0.83 BI from IP every time, out of which only 3/5=0.6 BI is spent on rake, in fact much less due to rakeback). The poker room will most likely deem it chip dumping and close the accounts, so I can't do this trick and am honestly sharing it with you, but think twice whether you should insure tourneys at all, especially if you don't have a security team.
Even an honest player can theoretically gain advantage because bubble insurance creates a new, flatter, prize structure for him and he can apply a (tighter) unexploitable strategy tailored to the new structure, while his opponents unaware of IP will use strategies that are close to unexploitable in terms of the original uninsured structure, but exploitable in terms of the new insured structure. All the players will win because they'll de facto play different tourneys, hence only IP will lose.
Such loopholes of course exist in almost any insurance policies but I want to stress that you can't do without checking reliability of signups thoroughly like poker backers and usual insurance companies do.
Hi c00n74,
Since my name + blog has been quoted in this thread, I wanted to have the chance to respond. If you read closely, you will see that I state,
"If my interpretation is correct, you are asserting that these pots amount to 34% of the money won/lost during a given session and by extension that is the degree of variance."
The correction I make in his
calculating the amount of money won/lost in a session. That does
not by extension mean that I agree that that number tells you how much variance is due to those all-in pots.
All my calculation tells you is the answer to the questions, "What percentage of money wagered in this sample is due to all-in spots?"
Your criticism of the methodology stands:
The simplest way to calculate the variance of the sample would of course be to include the 'mean of the squares' of all the individual pot sizes, just the all-in pots, and just the non-all-in played pots. You could then make a legitimate variance calculation of the sample. However this would still not achieve the aim detailed in the title of that blog post. Your other posts in this thread represent a sterling attempt to do so, although I am still not sure how to resolve the variance1 v variance 2 debate (see below).
Please note that I said in my reply:
"I am little troubled by your methodology for calculating variance."
The intention at the time I posted was to correct an obvious scaling error that I saw when first scanning the article out of curiosity. I did not mean for my post to be interpreted as an endorsement of the overall method. As it stands, I am pretty confident that he has significantly underestimated the amount of 'variance' due to all-in pots through this approach.
The implication in your post was that I did not understand variance. That is simply not the case.
If that was not your intention, and you merely meant that it was a shame I did not go so far as to correct the methodology also, then I apologize in advance for my failure in interpretation.
The actual debate
The problem is that this throws you in the deep end with regards what poker players mean when they read the word 'variance'. The calculated variance of the samples is not the same as the 'luck factor' which poker players mean when they talk about "variance". We can consider two categories, I will term variance 1 and variance 2.
Variance 1
The mathematical definition of variance, which can be calculated from a sample, given sufficient data, or otherwise estimated given assumptions about the distribution of that data.
Variance 2
What poker players/readers/potential customers of the insuredplay site mean. This includes relatively simple coolers like getting set-over-set and more subtle 'running into the top-of-his-range' all month. It is not immediately clear to me how to measure this concept.
I hope you will understand my difficulty when I initially responded. If I had said outright, "here, just do this calculation and you will find how much variance is due to all-in spots" that would have been misleading. I would have been correcting the 'variance 1' calculation, but would have been understood by many poker players to have stated outright, "This is how much of variance 2 is due to all-in luck". Rather than open that particular can of worms, I chose to make the minor correction under his own terms.
Regards to all,
Quad
Last edited by Quadrophobia; 11-23-2012 at 09:31 AM.