Comments on: The Flaw of Averages http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/ A slice of lime in the soda Sat, 18 May 2013 23:43:55 +0000 hourly 1 http://wordpress.org/?v=3.4.2 By: mediocristanite http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-8307 mediocristanite Mon, 02 Nov 2009 01:33:47 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-8307 Sam Savage's book is great. I work with a bunch of very intelligent, highly educated people in a very quantitative field (investment management) - and yet, for all the sophisticated analysis we apply - the flaw of averages lurks everywhere and crops up in the most seemingly benign fashion and impacts decision-making all the time. Fixating on 'black swans' is all the rage these days, but let's not forget to tie our shoe-laces. Sam Savage’s book is great. I work with a bunch of very intelligent, highly educated people in a very quantitative field (investment management) – and yet, for all the sophisticated analysis we apply – the flaw of averages lurks everywhere and crops up in the most seemingly benign fashion and impacts decision-making all the time. Fixating on ‘black swans’ is all the rage these days, but let’s not forget to tie our shoe-laces.

]]> By: Stan Lucas http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-6841 Stan Lucas Wed, 16 Sep 2009 07:06:47 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-6841 Well, isn't the hedging of site A with site B just something a good oil executive could tell you? Since site A is so potentially unstable, just put a minimal amount of cash into it. If it goes great you still benefit, if it goes badly, site B compensates. Ross Perot keeps most of his investments in safe investments, any risky investment he wants to get involved in, he uses cash from the dividends of his safe investments for. Well, isn’t the hedging of site A with site B just something a good oil executive could tell you? Since site A is so potentially unstable, just put a minimal amount of cash into it. If it goes great you still benefit, if it goes badly, site B compensates.

Ross Perot keeps most of his investments in safe investments, any risky investment he wants to get involved in, he uses cash from the dividends of his safe investments for.

]]> By: drewbie http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5937 drewbie Tue, 25 Aug 2009 20:18:54 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5937 I think an area outside of business decisions that could benifit from a distribution analysis instead of a mean would be life expectancy. It's often pointed out that the US ranks 50th in life expectancy, but the CIA factbook this number comes from doesn't list a variance, and it's accurate to a hundreth of a year (which is ~3 days). While the algebra behind this average is certainly calculated accurate, using such a precise number for so broad a group is rediculous. Comparing the distributions (along with statements like "78% of the population will survive past 50") would be much more educational. I think an area outside of business decisions that could benifit from a distribution analysis instead of a mean would be life expectancy. It’s often pointed out that the US ranks 50th in life expectancy, but the CIA factbook this number comes from doesn’t list a variance, and it’s accurate to a hundreth of a year (which is ~3 days). While the algebra behind this average is certainly calculated accurate, using such a precise number for so broad a group is rediculous. Comparing the distributions (along with statements like “78% of the population will survive past 50″) would be much more educational.

]]> By: Sam Savage http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5905 Sam Savage Mon, 24 Aug 2009 20:47:30 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5905 As to angels on the head of a pin, the example I use is discussions of how to ride a bicycle. This is absurd. You don't learn to ride a bike by talking about it but by riding one. That's why we need models; to express things that are otherwise meaningless, and connect the seats of our intellects to the seats of our pants. As to angels on the head of a pin, the example I use is discussions of how to ride a bicycle. This is absurd. You don’t learn to ride a bike by talking about it but by riding one. That’s why we need models; to express things that are otherwise meaningless, and connect the seats of our intellects to the seats of our pants.

]]> By: Francine McKenna http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5885 Francine McKenna Mon, 24 Aug 2009 14:12:04 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5885 I find these types of discussions extremely frustrating because I am a manager. They are a sort of quant porn, if you ask me. (P.S. Felix, I don't read management books either.) The probability (but not the value, and don't get me started on expected value) of events that are the result of human behavior are impossible, in my mind, to pin down with mathematical precision, Freakonomics not withstanding. I read this and think, "angels dancing on the head of a pin..." Which reminds me of one of my favorite jokes: How do you make God laugh? Tell him your plans. I find these types of discussions extremely frustrating because I am a manager. They are a sort of quant porn, if you ask me.

(P.S. Felix, I don’t read management books either.)

The probability (but not the value, and don’t get me started on expected value) of events that are the result of human behavior are impossible, in my mind, to pin down with mathematical precision, Freakonomics not withstanding. I read this and think, “angels dancing on the head of a pin…” Which reminds me of one of my favorite jokes:
How do you make God laugh?
Tell him your plans.

]]> By: Sam Savage http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5871 Sam Savage Sun, 23 Aug 2009 23:07:18 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5871 I have not only heard of variance, skewness, and kurtosis, but refer to them as RED WORDS. I define these as terms left over from the steam era that cannot be uttered in a singles bar. They are often used by statisticians to blow smoke. Instead I recommend using the entire distribution, stored in something called a Distribution String or DIST. It consists of 1000 Monte Carlo trials stuffed, like a genie in a bottle, into the single cell of a spreadsheet. Every time you make a change something, all thousand trials rip through your model like bullets through a machine gun. Check out the demo models at ProbabilityManagement.org. I have not only heard of variance, skewness, and kurtosis, but refer to them as RED WORDS. I define these as terms left over from the steam era that cannot be uttered in a singles bar. They are often used by statisticians to blow smoke. Instead I recommend using the entire distribution, stored in something called a Distribution String or DIST. It consists of 1000 Monte Carlo trials stuffed, like a genie in a bottle, into the single cell of a spreadsheet. Every time you make a change something, all thousand trials rip through your model like bullets through a machine gun. Check out the demo models at ProbabilityManagement.org.

]]> By: Ash http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5863 Ash Sun, 23 Aug 2009 19:10:09 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5863 Average is a measure of just the central tendency. There are other measures which summarize a distribution. Variance, skewness, and kurtosis are the measures of other characteristics..... have you heard of them Savage? Who looks at only the averages and takes decisions btw? Average is a measure of just the central tendency. There are other measures which summarize a distribution. Variance, skewness, and kurtosis are the measures of other characteristics….. have you heard of them Savage? Who looks at only the averages and takes decisions btw?

]]> By: Eric Dewey http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5860 Eric Dewey Sun, 23 Aug 2009 17:33:25 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5860 Yes, and yes. This post demonstrates the power of being able to think about mathematics rather than arithmetic. I am not a mathematician, and do not fully understand the technicalities of the arguments. Does this make me more or less like the standard board member? Because of the cogent thinking and writing, I am able to understand and appreciate both Felix' critique of Savage's book, and Savage's well-done response. Were I a board member faced with the decision outlined in Savage's example, I would be more comfortable making the decision to hedge site A with site B, simply because I would feel far more informed about the actual risks presented, as well as infinitely more confident in the management team's ability to manage those risks... Yes, and yes. This post demonstrates the power of being able to think about mathematics rather than arithmetic.

I am not a mathematician, and do not fully understand the technicalities of the arguments. Does this make me more or less like the standard board member? Because of the cogent thinking and writing, I am able to understand and appreciate both Felix’ critique of Savage’s book, and Savage’s well-done response.

Were I a board member faced with the decision outlined in Savage’s example, I would be more comfortable making the decision to hedge site A with site B, simply because I would feel far more informed about the actual risks presented, as well as infinitely more confident in the management team’s ability to manage those risks…

]]> By: MattF http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5856 MattF Sun, 23 Aug 2009 12:23:12 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5856 As far as I can tell, the mean has one useful property-- given a set of measurements of some quantity, the mean <i>of that set of measurements</i> is the unbiased estimate of the quantity's actual value. For anything else, <i>caveat computor</i>. As far as I can tell, the mean has one useful property– given a set of measurements of some quantity, the mean of that set of measurements is the unbiased estimate of the quantity’s actual value. For anything else, caveat computor.

]]> By: ColdMountain http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5853 ColdMountain Sun, 23 Aug 2009 11:52:20 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5853 Savage would be the first to tell you that his tool is primarily for closed systems dice, cards, etc…, not open systems like the market. It’s a very clever tool for closed systems. But it can be adapted for the fat tails most of us have known about for around half a century. Taleb seems to have just discovered it. Another feature of open systems as many including Tavakoli pointed out in her credit derivatives book, is that none of the models capture risks that only management can deal with. Information asymmetry is just one example. Savage would be the first to tell you that his tool is primarily for closed systems dice, cards, etc…, not open systems like the market. It’s a very clever tool for closed systems. But it can be adapted for the fat tails most of us have known about for around half a century. Taleb seems to have just discovered it. Another feature of open systems as many including Tavakoli pointed out in her credit derivatives book, is that none of the models capture risks that only management can deal with. Information asymmetry is just one example.

]]> By: coverpunch http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5852 coverpunch Sun, 23 Aug 2009 07:27:05 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5852 http://en.wikipedia.org/wiki/Variance "Whereas the mean is a way to describe the location of a distribution, the variance is a way to capture its scale or degree of being spread out." http://en.wikipedia.org/wiki/Variance

“Whereas the mean is a way to describe the location of a distribution, the variance is a way to capture its scale or degree of being spread out.”

]]> By: Sprizouse http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/comment-page-1/#comment-5850 Sprizouse Sun, 23 Aug 2009 04:45:28 +0000 http://blogs.reuters.com/felix-salmon/2009/08/23/the-flaw-of-averages/#comment-5850 The Halo Effect by Rosenzweig is a great management book that's not really a management book either. If you haven't already read it, give it a go. If you like Taleb's stuff and this book you'll probably like The Halo Effect. The Halo Effect by Rosenzweig is a great management book that’s not really a management book either. If you haven’t already read it, give it a go. If you like Taleb’s stuff and this book you’ll probably like The Halo Effect.

]]>