Regression to the Mean
Before we get started, I’d like to give a quick shout out to our sponsor’s today modelthinkers.com. They have a thriving community and knowledge base built around mental models and rational thinking.
I’d highly recommend you check them out.
They also have a special 25% off for readers of this newsletter.
There was a popular study that was doing rounds among the Israeli fighter pilots which stated that pilots tend to perform better with negative feedback rather than positive feedback.
This was in stark contrast to multiple controlled research studies that advocated the exact opposite. So which one of these is accurate?
What if someone told you that the pilots’ performance had little to do with positive or negative feedback?
What if they told you that it is not the feedback but math of randomness that contributed to the pilots’ performance?
To understand this better, we need to explore the concept called regression to the mean.
Definition: Regression to the mean is the idea that outcomes will get closer to the mean as they increase.
Hitting the Bull’s Eye
If you started throwing darts at a board and hit the Bull’s Eye with your first 3 attempts, the chances of you hitting the Bull’s Eye on your fourth attempt will be low.
It is more likely that you will hit somewhere on the board that is not very close to the bullseye.
This is the natural behaviour of any random quantity.
Hitting the bull’s eye is a rare event, just like hitting something 10 metres away from the board. And as you keep adding the number of attempts, chances are that you’ll start hitting close to where people usually end up on average.
Think of this like the class bell curve. As more and more attempts start accumulating, most attempts will end up among the average ones.
The good attempts that hit the target will become rare and the bad attempts that hit far from the board will also become rare.
This is widely prevalent especially when there is a random factor involved.
Grab 25% off on ModelThinkers PRO membership exclusive to my readers.
Tall, Short, Average
If you walk into a room in a city where the average height is 6 ft, then chances are that you might see a few people who are 6.5 ft or more and some who are 5.5 ft or less. But the majority of the people will likely fall very close to the 6 ft mark.
And as you keep adding/counting more people, the majority will start accumulating around the 5.7 - 6.3 ft range.
The same applies to any randomised numbers like IQ, wealth, stamina etc.
Applications of the Model
Regression to mean is a mental model from the domain of statistics but the applications can be widespread.
Misusing the model is also very rampant where we falsely attribute results to an arbitrary factor when in fact it was just simple statistics.
Think of people who believe in superstitions when it comes to sports. Their team might score a goal when they do a ritual but as they keep doing their ritual the team’s performance will keep moving towards the average.
But owing to confirmation bias, it might seem to them that their team is doing well when they perform a ritual.
Avoiding Bad Decisions
There are 3 ways to avoid bad decision making by using the regression to the mean model.
Work with longer time frames and wider data ranges
Focus on process and not just the results
Don’t count on extremes while planning & forecasting
Longer Time-Frames: When you use longer time frames and wider data ranges, you will get a clearer picture of the different factors of influence. Essentially, you’ll be able to remove false attributions.
Process Over Results: When you focus on processes that improve your chances of winning rather than one-time windfall results, you are more likely to succeed in the long run. You are better off making 30% a year with a airtight process for 30 years instead of making 300% one year with no plan on how to repeat it again.
Avoiding Counting on Extremes: Really good results and really bad results are not common. So when you are strategising, planning or forecasting, do it based on the average performance.
So that is it! What are your thoughts on this model?
Let me know by replying to this email.