a good introduction to forecasting (and how it usually doesn’t work very well), though it’s a little bit too basic if you’ve already read the kind of stuff i usually read. the first chapter is probably like the sixth or seventh explanation of the 2008 financial crisis i’ve read, and of course i was already well aware of the twist at the end of the good ol’ deep blue vs. garry kasparov chess anecdote. some chapters did bring in some fresh material, especially the one on poker (the author is a part-time pro poker player). on the surface it might seem like an easy way to make a living if you’re good at it but it’s actually quite brutal, apparently the vast majority of winnings come from just the worst 20% of players or so, and if there isn’t a constant stream of fresh “fish” (suckers) coming in and losing lots of money then it can turn into a grueling grind, especially stressful because you also have to constantly manage your psychological state and avoid going “on tilt” (playing bad because you’re mad). plus, even if you play perfectly, there can be long dry streaks simply from having bad luck.

overall, my big takeaway is that forecasting sucks in almost every field besides weather forecasting, and it’s such a difficult problem that there doesn’t seem to be much hope for improvement outside of some deus ex machina like AI suddenly solving everything. i also learned one especially freaky thing: it probably isn’t that shocking to hear that there is an inverse relationship between earthquake frequency and scale, e.g. tiny earthquakes are extremely common whereas megaquakes only come along once every century. by plotting known frequency/scale data on a log plot, it’s possible to extrapolate and estimate the approximate magnitude and frequency of a hypothetical future gigaquake. but, it’s also possible to do this with terrorist attacks, and plotting then extrapolating from 40 years of terrorist attacks up to the year 2000 predicts that an outlier attack the size of 9/11 is due…