TimTebow @socks0803

Love is patient, love is kind. It does not envy, it does not boast, it is not proud...it is not self-seeking. - 1 Corinthians 13:4-5 (please retweet)

@BraydenYee @HashTaguars Agreed, but if theres not enough data for the model to be reliable then theres no reason to think Shanny has enough data for his intuition to be reliable. Or if he does have some secret data that hes hiding then he can share that with the analytics staff to improve their model.

@BraydenYee @HashTaguars My intuition is that he(and other old school coaches) don't do this because: 1. It would expose him as not really having this secret insight he pretends to have 2. He likes being able to appeal to his "football experience" to make himself an authority who can't be questioned

@BraydenYee @HashTaguars Fair enough, we can agree to disagree on interpretation. But I take issue with your last sentence....if he has insight that the model is overlooking then why not explain that insight to his analytics staff so that they can incorporate it into the model?

@akashanav Whats wrong is that he is working AGAINST his analytics staff instead of WITH them. If his coaching experience allows him to have insight that the model is overlooking....then he should explain that to the analytics staff so that they can incorporate the insight into their model

@BraydenYee @HashTaguars When he says "I'm not playing blackjack 𝗳𝗼𝗿 𝗮𝗻 𝗵𝗼𝘂𝗿....𝗜 𝗷𝘂𝘀𝘁 𝗵𝗮𝘃𝗲 𝘁𝗵𝗶𝘀 𝗼𝗻𝗲 𝘁𝗿𝘆" he seems to be implying that a small sample size can negate a mathematial edge(regardless of whether that edge is estimated or certain). And I disagree strongly with that

@BraydenYee @HashTaguars We never know, its an estimate. This is equally true both for a model's output and a coach's intuition/feelings. I probably confused matters by talking about mathematical certainty in point #2. My fault there. I only brought that up because...

@RyanReynoldsNFL @JakeSirkus My apologies, I'll leave you guys be. I certainly could've made a bad read on you here. Hopefully no hard feelings.

@BraydenYee @HashTaguars I mean, maybe you're right that theres a contingent of dumb people online who are using analytics-based arguments incorrectly. I certainly can't defend all those people. But people who actually spend their lives studying this stuff....you should give them a little more credit.

@BraydenYee @HashTaguars You're wrong. I'm not sure who "the analytics community" is referring to, but... Anyone who has gone to school for computer science(esp someone who goes on to land a 6 figure job in the NFL) 100% understands the difference between a model's estimate and a mathematical certainty.

@JeffreyHuber69 @SharksNSports88 @katsuxbt A large number of variables is an argument in favor of machine learning, not the human brain(which has a very limited amount of "RAM")

@HashTaguars Most everyone(including analytics guys) agrees that the margin of error is over 1%(which means the model is fairly useless in such close cases) #2 is where Shanny is getting pushback from the analytics community, as it sounds as if hes saying "yes I would sometimes take the 49%"

@HashTaguars There are two separate issues here: 1. What is the margin of error on these 51/49 estimates? 2. If you know for a fact that something is 51/49, should only having one try at it(instead of a large sample size of tries) ever cause you to willingly pick the 49%?

@davidsven If you're that involved in basketball then you should be well aware of how tough it is for a 3 man ref crew to make accurate calls during a fast paced(24 sec shot clock, super athletes) NBA game. Of course they will miss calls, but overall NBA refs are amazing at their jobs.

@RyanReynoldsNFL @JakeSirkus Lots of eyetest/feel people don't like discourse surrounding mathematically solved topics(such as 4th downs) because it doesn't allow them to make arguments from authority where they tell us about how we should trust them because "they've watched a tremendous amount of football."

@RyanReynoldsNFL @JakeSirkus I get your concerns, but you're underrating how many "obvious holes" can be plugged by a team of skilled data scientists working on them(for years!). You say "football is a mess", but -110 side/total and -115 player prop markets are ~unbeatable. Stuff is pretty well solved.

@JakeSirkus @RyanReynoldsNFL ☝️ I don't understand why the average person acts as if data scientists are idiots who can't possibly incorporate something super simple(such as quality of QB) into their model. This is trivially easy stuff for even beginner data scientists yet alone those employed by NFL teams.

@SharksNSports88 @katsuxbt Your assumption teams aren't evaluating spot specific factors before arriving at a 51/49 output is incorrect imo. Most teams have very talented analytics guys who have been working on these things for years...they know how to incorporate all factors into 4th down decision models

@RobKuhn_ I'm reminded of this excerpt from Phil Jackson's 1995 book Sacred Hoops in which he talks about how big men shooting 3s "demeans your opponent"

this level's pretty mid, retweet if you agree