Bet the Process - Cade Massey

Podcast Release Date:
Podcast Link:
hagrinBPR Review Date: 2024-04-30 19:56:21
Rating: 7.25

Notes:
This was a super hard one for me to grade for the opposite reason than usual (which is I don't know anything). I have consumed all of Cade's content 10s of times so there's nothing new in here for me, but Cade is awesome and paved the road for others. If you're on top of Cade's work, there's not much new in here, but if you're someone who isn't, Cade is very smart and it's a good listen.

2:45 - "If TOR wasn't tanking would he have done this?" re: the prevalence of tanking.

9:00 - "If all you could bet is NFL Draft, what are the 3 things you do in the year preceding to get ready to bet it?" Historical data. Look for tendencies in teams. Look at historical mocks at different timeframes out. Try and determine biases in the process. Get historical market pricing. Obviously, using connections.

12:23 - Cade interview starts.Full time professor in Information and Decision Dept at Wharton. Talks about the findings from 25 years ago that from a value perspective the value of 1st round draft picks was actually flipped. 15:00 - Jeff asks how this has changed and Rufus interjects that the rookie pay scale has changed. Cade expands that the pay scale changed, but in terms of getting better at scouting / draft order, based on whatever performance metric you want, we don't seem to be getting much better at this.

17:00 - Rufus asks if we are any better predicting one position over another? Cade answers I don't know, but does say "different positions are differently predictable". Gives example of IOL having a tighter range of variance. 18:40 - Mentions technology has gotten much better in the recent years and maybe the increase in predictions and decisions are a few years away as we better understand how to use this technology better - ex: in-game analytics vs just using the Combine like in the past.

19:00 - No predictive value from surplus value. Talks about a consulting project he did for a team who asked "Is someone good at drafting?" so that they could copy them. Cade finds that basically the surplus value derived from a pick's expected value is basically entirely chance driven, always very low, less than 10% of the variance.

22:00 - NFL teams very rarely do anything consistently, i.e. selection issues. Hard to see patterns, policies pan out. Even within organizations, there are differing opinions / competing forces. (Since there is so much noise at the organization level, I wonder if some type of scouting prediction market would work which would be something I would actively participate in.) Selection bias - the guys who perform worse (-surplus value), they get selected out / fired.

26:00 - "What's the optimal strategy if you have job security for 10+ years?" "We tend to think about it a pick at a time. It's a bundle of picks for a bundle of picks." "It's a huge portfolio problem that no one has really modeled, you can't make one decision without affecting the constraints of other decisions." Says he would need 10 years and a clean run to build. I think if you're reading this and you want to be in a front office one day, this is the research project to start doing in the public space. Later describes that most of his team work these days is organizational and "blending" all of these opinions.

I'd also like to bring up that he mentions "we have to be humble" several times and I feel like we should probably send this clip to all of Football Analytics Twitter which is the thinnest skinned set of analysts in pro sports and by quite a wide margin. "Acknowledge the limit of what you don't know." "Irreducible uncertainty." (Sidenote - searching this term on twitter leads to some very interesting threads including one which, in response to this podcast, called "Cade overconfident about other's overconfidence" which, 1) funny but 2) IMO, you should challenge yourself to formulate a response to a statement like that).

33:15 - "Mechanical aggregation". Possibly using mechanical aggregation to "set the board" to "start the conversations". 38:50 - "What about wins, does this stuff matter in wins?" They found it did - 1 std dev movement, 1.5 wins a year (Pats and Eagles, 8-10 window). (Sidenote - part of me wonders how you measure the effect of pick portfolio management when confounded by Brady and Wild Bill, but generally I agree & makes sense).

42:00 - Cade corrects Jeff who said "never trade up" and Cade offers the small window where it's possible trading up might be correct strategy and it's a good refresher.

47:40 - Rufus closes with an EXCELLENT question re: the finding that even if you chose the right player that the chart would say you still shouldn't pick and you're still overvaluing the right to choose/draft. "The actual trade chart was steeper than the 'Fortune Teller's' chart." Even with perfect knowledge which player will be best!