- Joined
- Apr 3, 2006
- Messages
- 26,126
- Reaction score
- 52,125
If you're going to bring up Butler, we have to talk about SB 52.
Not sure what you mean…what happened?
Registered Members experience this forum ad and noise-free.
CLICK HERE to Register for a free account and login for a smoother ad-free experience. It's easy, and only takes a few moments.If you're going to bring up Butler, we have to talk about SB 52.
My understanding is that the predicted win probabilities are provided by some form of machine learning algorithm that takes into account information about the current game, but has been trained on a large set of historical play-by-play data, hopefully also including external factors like weather, injuries etc (the total of the information in the training sample is what I referred to as "average"). For a choice between two alternatives the algorithm may predict a 1% difference. Based on my own work with machine learning algorithms in much better controlled and data rich settings, this is unlikely to be a meaningful distinction (supported e.g., by the wild swings one can observed in real-time calculations of these probabilities). What is now a 1% distinction in one direction could easily flip sign, if more data or slightly updated models become available, and even in the best of circumstances is small compared to the swing one good (or bad) tackle, throw, run, kick etc can make.I Truly dont see where they said they gain 1% in average or studied only changes of 1% in average. And I also dont see any averages in this picture...
It says gains of at least of 1%.
And saying with confidence what is the variance of any given equation is... bold. Unless you are saying that any given event has large variance, to which I respond "Obvs"... no studies is determined by any single event
Ps - if you did the math (like, adding all samples, derivating the average based on graphs) and reversed engineering these probabilites, I just applaud you.
So junk in, junk out?My understanding is that the predicted win probabilities are provided by some form of machine learning algorithm that takes into account information about the current game, but has been trained on a large set of historical play-by-play data, hopefully also including external factors like weather, injuries etc (the total of the information in the training sample is what I referred to as "average"). For a choice between two alternatives the algorithm may predict a 1% difference. Based on my own work with machine learning algorithms in much better controlled and data rich settings, this is unlikely to be a meaningful distinction (supported e.g., by the wild swings one can observed in real-time calculations of these probabilities). What is now a 1% distinction in one direction could easily flip sign, if more data or slightly updated models become available, and even in the best of circumstances is small compared to the swing one good (or bad) tackle, throw, run, kick etc can make.
Note added in proof: Going back to the last "big controversy", going for it or kicking vs Tampa Bay, different analytical models gave wildly different estimates of the respective win probabilities, with swings of more than 10%. Prior information not available to the models (e.g. that Mac got his head scrambled by a hit a few minutes prior or that Folk was making 55+ yard kicks with ease in warmup) easily trumps the predictive power of the model for situations close to 50/50.
I, for one, am sick of BB being micro-analyzed on this board by "fans" and pundits who actually believe they "know things" sufficient to disparage the greatest coach this sport ever has seen -- with or without Tom Brady. It's what "fans" do, I suppose, when they're used to being spoiled and the road gets bumpy. BB is nearing the end of an incredible career run and too many here are eager to rush him out the door. Be careful what you wish for, folks. Whoever comes next likely will give cause for some REAL second-guessing.
Many here are tired of being told that, effectively, Belichick is above criticism. And, as far as "Be careful what you wish for", you should have given that lecture to Belichick, before he pushed Brady out the door.
Got it, i never seen this definition of average - being the total of anything -, but Im dont have have that much experience with ML, Im more of a Econometrics kinda of guy.My understanding is that the predicted win probabilities are provided by some form of machine learning algorithm that takes into account information about the current game, but has been trained on a large set of historical play-by-play data, hopefully also including external factors like weather, injuries etc (the total of the information in the training sample is what I referred to as "average"). For a choice between two alternatives the algorithm may predict a 1% difference. Based on my own work with machine learning algorithms in much better controlled and data rich settings, this is unlikely to be a meaningful distinction (supported e.g., by the wild swings one can observed in real-time calculations of these probabilities). What is now a 1% distinction in one direction could easily flip sign, if more data or slightly updated models become available, and even in the best of circumstances is small compared to the swing one good (or bad) tackle, throw, run, kick etc can make.
Note added in proof: Going back to the last "big controversy", going for it or kicking vs Tampa Bay, different analytical models gave wildly different estimates of the respective win probabilities, with swings of more than 10%. Prior information not available to the models (e.g. that Mac got his head scrambled by a hit a few minutes prior or that Folk was making 55+ yard kicks with ease in warmup) easily trumps the predictive power of the model for situations close to 50/50.
Seriously. Hes in the wrong forum.
Bill has shown us all repeatedly that he believes in letting players go too early rather than too late. What's wrong with applying that to coaches?I, for one, am sick of BB being micro-analyzed on this board by "fans" and pundits who actually believe they "know things" sufficient to disparage the greatest coach this sport ever has seen -- with or without Tom Brady. It's what "fans" do, I suppose, when they're used to being spoiled and the road gets bumpy. BB is nearing the end of an incredible career run and too many here are eager to rush him out the door. Be careful what you wish for, folks. Whoever comes next likely will give cause for some REAL second-guessing.
Thats not true. He lets players go that no longer produce up to their cap/salary, there is a younger/better option, or don't want to be here anymore.Bill has shown us all repeatedly that he believes in letting players go too early rather than too late. What's wrong with applying that to coaches?
Bill is 10 games under .500 as a HC without Brady. That's not an opinion, it's a fact.
It's not really about BB. The coaching on this team is questionable right now and it's fair to wonder what's going on.I, for one, am sick of BB being micro-analyzed on this board by "fans" and pundits who actually believe they "know things" sufficient to disparage the greatest coach this sport ever has seen -- with or without Tom Brady. It's what "fans" do, I suppose, when they're used to being spoiled and the road gets bumpy. BB is nearing the end of an incredible career run and too many here are eager to rush him out the door. Be careful what you wish for, folks. Whoever comes next likely will give cause for some REAL second-guessing.
Belichick isn't getting fired by this team now, end of the year, or ever. He will step down when the time comes.Thats not true. He lets players go that no longer produce up to their cap/salary, there is a younger/better option, or don't want to be here anymore.
...and firing BB now or at the end of the year is nonsensical.
They had one timeout but I agree with your point.It's not really about BB. The coaching on this team is questionable right now and it's fair to wonder what's going on.
When you have a rookie QB that is 3 for 3 for leading his team to a score in a 2 minute drill, and you concede a half with 90 seconds and 2 timeouts instead of letting your QB go for it again, and you lose because of that decision, and you give half baked answers why that happened, then it doesn't matter who our coach is there will be questions.
There was no challenge that would have reversed a play on the field during the Cowboys gameThat and he hasn't challenged any calls either. A red flag could have helped their chances last Sunday.
Bill has shown us all repeatedly that he believes in letting players go too early rather than too late. What's wrong with applying that to coaches?
Bill is 10 games under .500 as a HC without Brady. That's not an opinion, it's a fact.
For some reason I remembered 2 but you're right.They had one timeout but I agree with your point.
The Meyers catch forward progress that was over the first down marker that was marked short would have been overturned on a challenge.There was no challenge that would have reversed a play on the field during the Cowboys game
Forward progress only applies when you get hit. You can't catch a ball and fall backwards untouched and get forward progress. Learn the gameThe Meyers catch forward progress that was over the first down marker that was marked short would have been overturned on a challenge.