I thought I might try something new by dumping thoughts on a book I just finished. Hopefully this will help me gather my thoughts in a more organized manner and if I’m lucky someone else might also find this post useful or interesting.
The Undoing Project tells the story of a pair of Israeli psychologists as they work to overturn conventional thought about human behavior. Their lives, friendship, and research are on their own a compelling story, but I was specifically drawn to this book due to its opening chapter. Lewis sets the stage for his larger story by first introducing another figure known for repudiating conventional wisdom – Houston Rockets General Manager Daryl Morey. Morey, dubbed “Basketball’s Nerd King”, is the poster child for the popularization of Basketball Analytics. He started with a basic statistical model built to predict the future performance of basketball players, and continually tweaked and adjusted the model to try to improve organizational decision making.
- He compiled box scores from every college game over the past 20 years as a starting point for the data set
- Traditional counting stats can be “wildly misleading” but can be useful when looked at per minute or pace adjusted
- The Rockets began gathering their own stats such as REB% and on/off
- They looked at a variety of player data such as parental status, college defensive scheme, and bench press which were all apparently nonpredictive
- Rebounds per minute were useful for predicting bigs and steals per minute for guards
- Length is more important than height
- In his 2007 model the odds of getting a “good NBA player” were 5-8% at the end of the lottery and 1% at getting a starter
- He later had to adjust the model by raising the weight on player age and the weight to games played against strong opponents versus weak opponents
- They expanded the model by adding physical traits – speed off the ground and quickness of first two steps
- They began questioning the value of workouts in creating confirmation bias based solely off a small sample size
- One day they noticed that they had systemically overvalued their own players and passed on positive value trades – a bias known as the Endowment Effect
NBA front offices are notoriously secretive, so it is rare to see this level of detail publically available. The specific details are interesting and could provide guidance for others looking to create their own draft models. However, there is an inherent difficulty in trying to predict how a young man will develop both physically and mentally upon becoming a professional athlete. Teams can only slightly affect the odds of selecting the right player, but only have a couple of tries (on average 2 draft picks and a couple roster spots) to play those odds. It ironically is analogous to the player who has a sudden uptick in 3-point shooting percentage in a contract year. How much of that shooting improvement is real? How much is the model’s evaluation of that player real? Morey experienced some early success by drafting Aaron Brooks and Carl Landry (both solid rotation players for many years) at the end of the 1st round. He followed that up by drafting Joey Dorsey with the 25th pick the subsequent year. Dorsey would proceed to be out of the league after his 4th season (with a 1 season comeback with Houston 2014-15).
The rest of the book about Daniel Kahneman and Amos Tversky was interesting in its own way by revealing some of the systemic biases we are all subject to. One particularly interesting bias is that people will respond to the same situation differently depending on framing. For example most people would rather take a sure gain of $500 versus a 50% chance to win $1000. Conversely, most people also prefer a 50% chance to lose $1000 versus a sure loss of $500. Could the framing of a trade proposal between GMs actually have an effect on whether a trade is ultimately agreed upon?
Many people also tend to misperceive probabilities when decision making. For example, most people treat a 99% chance of Event X as not quite 99% certain and a 1% chance of Event Y as more common than 1% of the time. Hopefully awareness can mitigate the distorting effect of these systemic biases.
Another takeaway I found particularly relevant to myself the power of Confirmation bias. In my current work as a consultant, there is an inherent pressure to deliver positive news to clients. I’m sure you can imagine that everything goes far smoother when clients hear what they want to hear. Of course nobody intentionally tries to mislead clients, but there is no way to gauge how much subconscious Confirmation bias is present in my work. As consultants, we are trained to craft a story around the data to present our findings, and for good reason. A story is more digestible and memorable than simply dumping hundreds of charts of data onto clients. However, there is always the worry that you are unknowingly gravitating towards the good and positive while analyzing and interpreting data.
I also noted some unique advice from Amos Tversky which I hope implement in the future.
On accepting invitations:
“Wait a day and you’ll be amazed how many of those invitations you would have accepted yesterday you’ll refuse after you have had a day to think it over.”
On leaving an event you don’t want to attend:
“Just start walking and you’ll be surprised how creative you will become and how fast you’ll find the words for your excuse.”