The Severed Cortex

Link: http://www.amazon.com/Fooled-Randomness-Hidden-Chance-Markets/dp/0812975219/ref=pd_bbs_sr_1?ie=UTF8&s=books&qid=1203862480&sr=8-1

I really, really enjoyed this book. I'd even say it changed the way I think about a lot of things. Perhaps it's just luck that I like it, as the author might suggest, but it probably has more to do with two things: math and philosophy. I won't say I'm a big reader of philosophical texts, but I have read a few philosophy books in my day and I've always had an predilection to gravitate towards them. I'm probably one of only a handful of people who have read Nitzsche while working on a towboat. Who knows, I might be the only one ever to read it whilst chugging the hours away on some seemingly endless waterway. I worked with a pilot who turned me on to Vonnegut and was on Jeopardy so there's a chance he might he read a little philosophy too. Math tomes are a little harder to come by and usually they come in one size. Boring. While this book can get a little boring with the math, or at least boring to those who don't like math, I certainly didn't find it to be that way. There aren't tons of formulas to dissect and interpret so trust me when I say that it's not like your math book back in high school.

The basic premise in the book, as you could have guessed from the title, is that most of what you believe to be cause and effect is probably more due to the factors of luck than anything else. I know it's hard to believe, especially to those who might have been history majors, but it just might be true. History majors like to say that history repeats itself and we can learn from it. Taleb believes that a lot of those historical outcomes were the result of luck and that relying on them to teach you a lesson is a fool's game. How does he come to that conclusion? Basically he uses a mathematical construct beloved by financial analysts called Monte Carlo analysis, or sometimes referred to as Alternate History.
I first learned about Monte Carlo analysis in the middle of a job interview for Goldman Sachs, which is probably not the opportune time to learn about it, especially when interviewing for a job where the primary focus was supporting the Monte Carlo analysis 'black box'. Later it would be explained to me by my good friend and financial guru Jay Watz. While infinitely more complicated than this, Monte Carlo analysis basically takes a set of parameters as input. Those parameters are put through the 'black box', which contains an algorithm of what you might be trying to measure, and then spits out all the likely outcomes, including outliers. If you don't understand standard deviations and the bell curve model, you won't understand how to measure the outputs, but generally a large number of outcomes will be within a certain, what I'll call normal range. Some outcomes will be outside that normal range, maybe a small amount, while others will be wildly out of that range. Those outside of the range are what Taleb might consider to be luck. Think of it like this. When the U.S. got into the Iraq War, there were a certain number of known parameters. (No I'm not about to get all Rumsfeldian on you). If we were able to quantify them in some way, we could have constructed a Monte Carlo 'black box' to determine what the outcome of the war might be. In some cases the Monte Carlo analysis might show we would be greeted as liberators a la Dick Cheney, and in others the analysis might show we'd be stuck in a 100 year quagmire a la John McCain. And still others might produce an outcome somewhere in between those two. In fact, most outputs would probably show something in between those two extremes. So, out of the myriad of paths the war could have taken, it has taken a path much closer to the 100 year outcome. A little luck could have pushed the outcome down the Dick Cheney path. There are hundreds of paths the war could have taken, and some are more likely than others, but Taleb might argue that the path the war did take was purely luck. Yes, better management at the top of the civilian defense arena might have increased the probability of an outcome more to Rumsfeld's liking, but there is no guarantee that it would have happened that way.
Here's an example that might be easier to understand. How many times did this happen to you at school? If you take a test, there are several outcomes that might occur. You may do poorly even though you studied. Meanwhile your friend, probably some jack ass named Drew, stayed out and drank beer instead of studying. He took the test and managed a passing grade. Surely you wouldn't say Drew had superior intelligence (even though he does ), because drinking beer instead of studying is not a smart decision. However, through some bit of luck he managed to answer the questions well enough to receive a passing grade. You on the other hand, through some bit of luck, managed to not answer the questions well enough. Monte Carlo analysis might have used the parameters of class attendance, class participation, and the amount of time studying for the material. In your case the likely outcomes would have been in a general range of 70 - 90. In other words if you took the test an infinite number of times, most of the time you would score in the 70 - 90 range. Drew on the other hand, given the same parameters as input, would most likely have outcomes that would be bunched up in the 30 - 60 range. However, both outcomes would contain 100 as an outcome, a much more unlikely scenario for Drew, and zero as an outcome, a much more unlikely scenario for you. They both could still happen and I know if Drew received a 100 on the test, you would definitely categorize that as pure luck.
Now comes the important part. If you look back at the history of the test that you and the fictitious Drew took, there are several statements you could make. You might look at Drew's score and say studying does not help when taking a test. You might also say that drinking beer before a test improves test taking skills. You could also say that your class attendance, participation, and studying has no effect on test scores. None of these are true, of course. In much the same way you cannot say that more troops in Iraq would have produced an outcome more to the liking of the neo-cons. The probability of that happening may have been more likely, but it's possible that luck would have borne out something more like we see today. More importantly, you cannot say that low troop levels brought us to the place we are today in Iraq. In a Monte Carlo analysis, low troop levels may have had some outcomes that were favorable to the U.S., although I'm sure those would be outcomes outside of one standard deviation and definitely more prone to luck.

Monte Carlo analysis is just the first step though. Next Taleb describes survivorship bias, or the perception that one finds when one believes to be in a certain group. This is a much tougher concept to grasp than Monte Carlo analysis. The example that finally clicked for me was one that Taleb described in the book. Suppose you live in Ladue or Town and Country or Wildwood, however you can barely afford to live there. Your neighbors, on average, make somewhere in the neighborhood of 500K per year while you make 125K. You drive a Mazda 3 and your neighbor drives a 7 series BMW. They also have a 35 foot cabin cruiser on the lake while you have a jon boat you fish with. They have 60 inch big screen HDTV's and you still have an old CRT TV. On the surface, it would appear that you are very poor when compared to your neighbor(s), if you only compared yourself to your neighbor(s). However, if you compared yourself to the general population you would find that you are quite wealthy. Understanding survivorship bias is understanding the 'pool' you are placing yourself in, and understanding how being in that 'pool' creates an appearance that may or may not be true. Taleb uses the booming market of the nineties to illustrate this point for traders. In the nineties, a lot of people did well in the market. Was it because they had some superior strategy for investing in the market? Not likely. A Monte Carlo analysis might show that multiple investment strategies might have done well during this time period. Some may have been sound, some maybe not. Some of those were not sound may have still produced investment returns in the double digits, while some sound investment strategies may have not done well. (Similar to taking a test) Those that did do well may not have realized that their good fortune was due to luck and most likely the luck that everyone in the market was doing well. Those who applied those same strategies in a down market like today will find themselves on Monster looking for a new job.

Taleb has multiple illustrations of Monte Carlo analysis and survivorship bias and how many things in life can be attributed by luck. Oftentimes he's able to illustrate these points in clear simple ways, and other times he does so by being a pompous ass. Hey, I've never met a philosopher that didn't think his ideas were above all others. I have no doubt that Taleb is right on a lot of counts, and asshole or not you should pay attention. If nothing else, this book changed how I thought about things. I'll never be able to hear about some medical study that links some cause and effect and actually be able to say I believe it to be true with a straight face. C'mon. How many times have we heard one thing, only for it to be refuted by a dozen studies years later? I'll also never be able to stand idly by when someone says, we should take a lesson from history. What we shouldn't do is stand in a bucket of water while working on an electrical circuit, rub your bare genitalia on a public toilet seat, or say the words Hey, y'all, check this out before attempting something that might be viewed later as being incredibly stupid. Those are definite bad ideas. Read this book and everything else you encounter you will definitely become skeptical about.