Bet the Process - Sarah Rudd

Podcast Release Date:
Podcast Link:
hagrinBPR Review Date: 2024-03-14 21:30:21
Rating: 4.00

Notes:
Jeff makes 2 football/soccer jokes in the first 30 seconds & boy was this foreshadowing.

Origin story - Sarah ran Arsenal's analytics group for 10 years. Meets Adrian Hanauer (Sounders owner), gets Sarah to start diving into analytics where Sarah writes a paper (believe it's the Markov Chain paper from 2011) that's similar to xT. Starts working for StatsDNA which gets acquired by Arsenal.

13:30 - interesting that one of the pieces of data that Sarah mentions that they didn't (don't?) have is "hip orientation" which would be an analysis gap between the model and video review with non-analytics people when determining how dangerous a situation was. (Sidenote - I actually don't know if this is a solved data issue at the club level, if someone knows ping me). She then brings up the language barriers & I could talk about this for hours when it comes to soccer betting. Incredibly she brings this up so quickly & it being an issue even within soccer analytic departments.

20:30 - At Arsenal they collected data down to the u15 level, but admits that analytics probably not the largest driver of player development. Much bigger edge in recruitment of players as lower tier clubs probably still not utilizing analytics fully & its better understood than youth development.

25:10 - Outside the box idea? With additional subs, we still don't see any real unique sub patterns implemented since sub increase. Great point, not sure I know the answer here, but Sarah's right that there's likely some undiscovered edge here. 27:25 - Players that start their careers earlier (16/17) that the sheer mileage generally means a sooner end to their career (calendar age vs football age) (sidenote - would be curious to see if there's any calendar age curve vs football age curve edge to be found for bettors in futures or late in seasons as legs tire or require lineup rotation). 30:00 - Very interesting. Looking at certain metrics that are captured in lower leagues as proxies for metrics that aren't captured at these lower leagues. Using event data that is usually available to extrapolate possible measurements that aren't captured. She talks about manual data collection, would have liked a question about possibly doing vision learning on lower league video to extract uncaptured metrics.

This pod could have been great - Sarah has been doing this a long time at a very high level and she's very talented. I'm sure there's more recent Sarah content, but I know her from her 2011 NESSIS talk (now 13 years ago). But you have two hosts that know nothing about the sport & apparently put no effort in before recording so they extract almost nothing of value - Sarah basically has to carry the convo herself to anything soccer specific. For example, Sarah brings up how they model where a possession should progress vs the actual. Would have loved a follow up where the ask about progression forwards vs backwards and if a player is playing too often backwards what the fix is from an analytics perspective (change in strategy, talk with player/coach, play a different player, etc.), short range passing vs cross field aerial crosses vs through balls, etc, taking defenders on dribbling, literally anything, but we just simply move on to the next prepared question.