Remodeling the Economic Future

[peristaltor]

Decades ago I dated a chess player, a very good chess player, one who trained with chess masters and knew first hand many of the names in competition at that time. One day in the smokey basement pub where chess players meet to play, she came back from a game downright pissed off

Comments

[albionwood]

Great post! Ties in with a lot of what I've been reading and pondering lately.

Failure to recognize assumptions makes us especially vulnerable to heuristic traps. We love an elegant story and want to believe it - so we often don't question the basis of the story. Quantitative economists are pethologically susceptible to this, because there are so few mathematical tools available for analysis of non-Gaussian data, or of small-probability events. Yet most real-world number sets are non-normally distributed, and small-probability events are usually the most important. So we blindly apply our lovely parametric statistics and produce elegant predictions, which are right 99.9% of the time, and then an event occurs which according to our models has a 0.0001% probability. How did that happen?

Guess I need to get back in the habit of reading Greer's blog. I got bored with it when he spent too much space making predictions about the future. Such predictions have an almost irresistable fascination, but are generally worthless.

Love the "playing like a fish" story and analogy. One thing about high-level chess, much of it is about getting into your opponent's head. Kasparov was the world's best at that aspect, which is partly why he lost the second Deep Blue match: no head to get into.

[peristaltor]

"One thing about high-level chess, much of it is about getting into your opponent's head."

I just flashed on those lessons. Her instructor was indeed not only relating historic games, he was suggesting that certain strategies might lead to certain outcomes, and then suggested certain historic counter-strategies to foil those offensive strategies. He was indeed teaching her how to get into a match-level opponent's head in order to anticipate the probable next moves, not (as I assumed) teaching her the pros and cons of various known strategies.

But yeah, with enough memory a computer need only mentally conjure an entire history of games based on playing with itself, or simply following a trail of probable moves too complex for the human opponent. Great point!