Intuition vs. formulas: simple algorithms are likely better

In “Thinking, Fast and Slow” (TFS, $2.99 for Kindle edition) Chapter 21 “Intuition vs. Formulas” author Daniel Kahneman cautions us against relying solely on intuition.  The chapter notes several examples of where simple formulas based on scoring of a handful of factors outperforms predictions by experts and their intuition:

1. Prediction of future price of Bordeaux wines.  A simple formula based on temperature, rain at harvest, and rain the prior winter was much more accurate than the intuition of wine experts.

2. Predicted grades of incoming freshmen at the end of the year.  A simple formula based on high school grades and one aptitude test was more accurate than the intuition of trained counselors who conducted 45 minute interviews and had access to more information.

3. Predicted marital stability.  Dawes’ research found stability is well predicted by a simple formula, frequency of lovemaking – frequency of quarreling.

4. Medical school admission and subsequent success, susceptibility to sudden infant death syndrome, and other examples are noted where simple formulas produce more accurate predictions that expert intuition.

Low-validity environments

In sum the research suggests “final decisions should be left to formulas, especially in low-validity environments.”  What exactly is a low-validity environment?  A low-validity environment is an environment where the outcome is not predicted correctly often.  In these environments human intuition is given too much weight over simple objective criteria with a known correlation to the outcome.

So what about intuition?

The TFS book and research referenced therein is not suggesting intuition should be totally disregarded.  There appears to be a couple scenarios where intuition matters:

1. When there is rare and decisive information not in the formula.  This is the “broken-leg” case.  When trying to predict whether or not someone goes to the movie a simple formula likely won’t have an input for the rare case of a broken leg.  However, a human using their intuition with that piece of information will make a better prediction.

2. When there is a “moderate validity” or “high validity” environment.  That is, when predictions based on intuition are known to correlate well with the outcome.

Scenario #1 offers a scenario where intuition should be used for the final decision: when there is new and decisive information.  Scenario #2 suggests that intuition should not be used as the final decision.  Rather, intuition should be used as one factor in the simple formula with all factors receiving equal weight.

Practical application: stock selection

Simple formulas based on scores of a handful of factors known to be correlated with the outcome can provide better predictive power.  Where can we apply this approach?  The Student Investment Fund at Sacramento State already partially implements the approach.  To screen stocks scores are assigned to each of seven factors and summed together[1].  This assigns equal weight to each factor.  From there students perform ratio analysis, Graham-Buffet analysis, and discounted cash flow analysis.  We currently do not look at technical indicators but should.  There should be a score for that as well.  The results of all analyses can be combined into an overall score.

Allow me to summarize one way to implement the simple formula approach to stock selection, the Sacramento State Student Investment Fund way:

  1. Quant score: this is simply the final score from the JO screen result.
  2. Fundamental score: after completing the ratio, Graham-Buffet, and DCF analyses assign a score on the same scale as the Quant score.
  3. Technical score: after looking at a number of technical indicators assign a score on the same scale as the Quant score.
  4. TOTAL = Quant + Fundamental + Technical
  5. Ignore your intuition and choose the alternative with the highest TOTAL [2]

Conclusion

I hope you found this post an interesting read.  I also hope you consider whether or not you overweight your own “expert” intuition in low validity environments.  🙂

 

Endnotes

[1] If you agree with the TFS argument, 7 is probably too many.

[2]  I don’t think we are at the “high validity” point where our intuition can accurately predict the outcome of superior risk adjusted future returns.  This is why I suggest ignoring intuition.  At best, you could include “Intuition score” in the TOTAL.

Advertisements

Leave a Comment

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 )

Google+ photo

You are commenting using your Google+ 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 )

w

Connecting to %s