Complex solutions are always not needed for complex problems

Takeaways:

  1. We do NOT need complex systems/models for complex problems. (Aruguably they are only complex because we have not found simple models).

For example, a model for how to position an outfielder can go two ways:

COMPLEX: Calculate speed and angle off the bat. Factor in wind, spin, so on and so on.

SIMPLE: Run in the linear direction of where the ball is heading. When it appears to be at it’s peak, keep the angle of view between you and the ball the same over time.

COMPLEX: Bunch of checklists if you will be able to land a sailing airplane at an airport

SIMPLE: Look at the air traffic tower through your cockpit window, if it rises upwards – you will not make it.

If you believe that the human mind does some combination of the above… just ask yourself how a dog catches a frisbee when you throw it! They don’t know how they do it. But they do it well.

In a world of risk, hueristics are always second class. In uncertainty this is often not so! Risk meaning well understood probabilities (blackjack) and uncertainty meaning unknowable futures (predict the stock market)

Measurement error is much more common that computational error. The issue with many academically created tools is that they are not practical. They introduce too much measurement error or computational effort! At times it will be be better to be off by 10% everytime (bias) than to be unsure how off you are all the time (variance)

For example, don’t do combinatorics for card counting. Just +1,0,-1 for certain cards and bet more heavily when you are +5 or more.

Fit your models to predictive performance rather than past fit. You do not want to “fit” your models to the data. Who cares how well you can predict the past!

As a general rule try to use heuristics that a) have single inputs (e.g. if a customer hasn’t purchased for 9mo they are inactive b) look for things that have been in use by practitioners c) how would a dog solve this complex problem?

When you see a system where participants do not “die out” you have a fragile system. For example restaurants die out. Hedge funds die out. Banks do not die out (but arguably have shifted risks to hedge funds).

You must have equal upside and downside exposure. Otherwise you will be put into a defensive decision mode (don’t get sued) if you don’t have enough upside – or – excessive risk taking (traders shooting for salaries, double down on risky moves when under).

Excellent discussion of Kelly Criterion as follows. Imagine 30 gamblers at the casino. One of them goes bust (loses it all). 29 survive. Now imagine instead of 30 gamblers, there are 30 days and one gambler. If you go bust on day 7, there is no days 8-30. Probability in space is NOT the same as probability in time.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s