I rarely read books these days, life is too busy and the lure of the internet and TV is strong at night after a long day. But this a really interesting book and a light, entertaining read. It delves into how we form opinions on things, and how confirmation bias consistently leads us down the wrong path. It starts off with a chapter on Rockets GM Daryl Morey — and how he has tried (with mixed success) to build a statistical model that could predict player potential for success better than either gut instinct or traditional performance stats, or even raw physical attributes. Lots of great antecdotes, like how over a several year period a number of slender, light-skinned black guys were overdrafted b/c they reminded scouts/GMs of Steph Curry.
Give it shot — I guarantee you’ll be hooked in the first few pages and the concepts in the book are a good primer for the upcoming draft debates!
(Also, it’d be great if anyone else has any other good book ideas — now that I’ve hit 50, I’m trying to make a point of getting off Rotoworld and reading a bit more, cheers)
AMZN - ( New Window )
Will check it out, thanks.
Give the first chapter a shot — it might not confirm your bias that Lewis is whatever you think he is!
How was Homo Deus? I loved Sapiens.
But every model make assumptions, has boundary conditions, and has to operationally define things. These are just a few places they can be biased or go wrong.
Take the lending crisis of 2008. One issue is that the algorithms used to predict default were based and validated on people carefully screened and meeting certain income criteria, credit scores, etc. But those standards were drastically relaxed as the mortgage boom was happening (because mortgage backed securities were in high demand as graded as really solid). Those assumptions about who got these loans were violated so the assumptions underlying those default models were wrong. Thus they failed at levels the models didn’t predict.
Or a researcher on something like say violence can define it operationally for a study as threats, documented violence in police reports, arrest, conviction, a certain level of violence, etc. Each drastically changes meaning when “x is linked to violence.” So too each element of algorithms have to be defined. Use something “neutral” like zip code to define something and it can end up being proxies for race, income, urban, etc. Likewise what variables to include, exclude, and control for matter.
Understanding humans build these things and thus they cannot be free from bias be it intentional or not will help people be appropriately skeptical.
Quote:
just finished Sapiens and Homo Deus this summer - great reads although longer. I am finishing up The Art of Logic in an Illogical World right now (should be a must-read for BBI LOL), and was wondering what might be next. This looks like a good recommendation.
How was Homo Deus? I loved Sapiens.
Good - not as good as Sapiens. I went into it having been told that it looks forward, and while it does that, it seems to rehash a bit of Sapiens. Overall though if you liked Sapiens and want a thought experiment on the future of the species, I definitely recommend Homo Deus.
I couldn't stop talking about these books - everyone in my circles probably sick of hearing me share anecdotes and insights, etc.
Understanding humans build these things and thus they cannot be free from bias be it intentional or not will help people be appropriately skeptical.
Love this - amazing how some of the stuff I read which is presented and accepted as fact is so full of assumptions. This is the key point for students in my HS data science class - if you can't articulate this concept clearly you cannot pass the class.
Quote:
Understanding humans build these things and thus they cannot be free from bias be it intentional or not will help people be appropriately skeptical.
Love this - amazing how some of the stuff I read which is presented and accepted as fact is so full of assumptions. This is the key point for students in my HS data science class - if you can't articulate this concept clearly you cannot pass the class.
Great posts by Daniel and Dan above ... all Dans are great posters, got my model.
The Lewis book is worth it alone for his recount of Morey’s attempt to build models ... and how kept being led down dead ends, and drafted numerous busts.
One of the pleasures of maturing, I think, is to realize just how little you know about things, how your best guesses based on the knowledge you have readily accessible can be completely off. The more I delve into stuff, the more I realize how much less I “know” compared to when I was a 30 year old.
So when I bust on Sy 55 for ranking LaQuon Treadwell and Corey Coleman and 15 other busts ahead of an historically great Michael Thomas — a Greek god from a high profile program—-. it’s less a slam on him and more an inquiry on how are we so bad at predicting performance.
His personal life accounts and how he has dealt with fear and overcoming obstacles in life.