skywlker32 Posted October 31, 2019 Share Posted October 31, 2019 22 minutes ago, ramssuperbowl99 said: The best, most accurate pain models are based on aggregating people's 1-10 scores for how much pain they're in. There are 0 "facts" in there, just "opinions". But your post surgery care is going to be the best it can be when you're coming out if the nurse asks you how you feel, scale of 1-10. So, again, this is just wrong. Is that a stat or a rating scale though? ESPN likes to use QBR as a statistic and claim to have an elaborate statistical model, but it equates closer to a PFF rating. That completely ignores how they gave a completely mediocre game the best rating ever, then tried explaining it away by saying he didn't play enough snaps, despite him playing the whole game. Are you going to ignore the rest of the post or just accept that the QBR "stat" is horrible at best. Link to comment Share on other sites More sharing options...
ramssuperbowl99 Posted October 31, 2019 Share Posted October 31, 2019 2 minutes ago, skywlker32 said: Is that a stat or a rating scale though? Those are arbitrary categories of data you're making up, and then disregarding based on personal preference. You use what works best. The goal isn't to be "right". You're never right. It's a model, not reality. The goal is to be useful and predictive. 2 minutes ago, skywlker32 said: That completely ignores how they gave a completely mediocre game the best rating ever, then tried explaining it away by saying he didn't play enough snaps, despite him playing the whole game. Are you going to ignore the rest of the post or just accept that the QBR "stat" is horrible at best. I don't have an opinion about QBR, but I see bad posts from people who don't know modeling, and I comment. 1 bad data point doesn't necessarily mean the model is bad. It's not good obviously, but it's not conclusive either. And them seeing a bad data point, then changing their model doesn't bother me at all. See how your model does, tweak it, and see if you can make it predict better is how this stuff works. Progressive trial and error, where your model is still wrong, but more useful. Link to comment Share on other sites More sharing options...
skywlker32 Posted October 31, 2019 Share Posted October 31, 2019 11 minutes ago, ramssuperbowl99 said: Those are arbitrary categories of data you're making up, and then disregarding based on personal preference. You use what works best. The goal isn't to be "right". You're never right. It's a model, not reality. The goal is to be useful and predictive. I don't have an opinion about QBR, but I see bad posts from people who don't know modeling, and I comment. 1 bad data point doesn't necessarily mean the model is bad. It's not good obviously, but it's not conclusive either. And them seeing a bad data point, then changing their model doesn't bother me at all. See how your model does, tweak it, and see if you can make it predict better is how this stuff works. Progressive trial and error, where your model is still wrong, but more useful. Again, they changed it, but the reason they gave for the change was troublesome. And this was never about 1 or 2 data points, there are MANY instances that make no sense. I don't think there is a combination of statistics that could justify any of them. Link to comment Share on other sites More sharing options...
ramssuperbowl99 Posted October 31, 2019 Share Posted October 31, 2019 3 minutes ago, skywlker32 said: Again, they changed it, but the reason they gave for the change was troublesome. And this was never about 1 or 2 data points, there are MANY instances that make no sense. I don't think there is a combination of statistics that could justify any of them. "Because it looked ridiculous and stupid" is a pretty routine reason for changing a model. Link to comment Share on other sites More sharing options...
skywlker32 Posted October 31, 2019 Share Posted October 31, 2019 2 minutes ago, ramssuperbowl99 said: "Because it looked ridiculous and stupid" is a pretty routine reason for changing a model. No, they claimed they fixed it by having a snap count minimum, yet he played the whole game. Link to comment Share on other sites More sharing options...
SkippyX Posted October 31, 2019 Share Posted October 31, 2019 4 hours ago, Danger said: QBR is closer to the Dunk Contest than it is a legit statistic. I don't remember who it was, but multiple posters here before have said that ESPN changed someone's QBR days later after a game because it seemed so poorly graded compared to the opposing QB's performance. Charlie Batch has one of the 5 highest QBR grades of all-time. The game is so embarrassing that ESPN hides it from the rankings. Link to comment Share on other sites More sharing options...
ramssuperbowl99 Posted October 31, 2019 Share Posted October 31, 2019 1 minute ago, skywlker32 said: No, they claimed they fixed it by having a snap count minimum, yet he played the whole game. Sure, but we're adults and both know the real reason was that it looked stupid. If all they did was raise the minimum snap count, that's lazy, but it's all proprietary, so we don't know what changes they made. There may have been some value-added changes, or it may have just been to increase the minimum. Link to comment Share on other sites More sharing options...
MWil23 Posted October 31, 2019 Share Posted October 31, 2019 6 minutes ago, ramssuperbowl99 said: "Because it looked ridiculous and stupid" is a pretty routine reason for changing a model. Counter argument: Link to comment Share on other sites More sharing options...
skywlker32 Posted October 31, 2019 Share Posted October 31, 2019 Just now, ramssuperbowl99 said: Sure, but we're adults and both know the real reason was that it looked stupid. If all they did was raise the minimum snap count, that's lazy, but it's all proprietary, so we don't know what changes they made. There may have been some value-added changes, or it may have just been to increase the minimum. This makes it more ridiculous actually. They didn't adjust his QBR, they just took him off the all time rankings. That game is still a 99.9 QBR (https://www.espn.com/nfl/player/gamelog/_/id/1490/type/nfl/year/2010). There is no defending a stat like that. 3 Link to comment Share on other sites More sharing options...
ramssuperbowl99 Posted October 31, 2019 Share Posted October 31, 2019 17 minutes ago, MWil23 said: Counter argument: I blame you for tagging me in this thread. 15 minutes ago, skywlker32 said: This makes it more ridiculous actually. They didn't adjust his QBR, they just took him off the all time rankings. That game is still a 99.9 QBR (https://www.espn.com/nfl/player/gamelog/_/id/1490/type/nfl/year/2010). There is no defending a stat like that. That's lazy and deceptive, so it's a red flag. But that still doesn't prove conclusively how useful it is either way. Link to comment Share on other sites More sharing options...
Shanedorf Posted October 31, 2019 Share Posted October 31, 2019 Next Gen Stats: SNF cut to Patrick Mahomes 40 different times during the recent broadcast and he wasn't even playing. That helps Mahomes attain a very high Q - score for the Madison Avenue crowd https://en.wikipedia.org/wiki/Q_Score Link to comment Share on other sites More sharing options...
bucsfan333 Posted November 1, 2019 Share Posted November 1, 2019 Not good. Link to comment Share on other sites More sharing options...
RaidersAreOne Posted November 4, 2019 Share Posted November 4, 2019 (edited) Edit: profanity in tweet Edited November 4, 2019 by RaidersAreOne Link to comment Share on other sites More sharing options...
bucsfan333 Posted November 4, 2019 Share Posted November 4, 2019 Link to comment Share on other sites More sharing options...
RandyMossIsBoss Posted November 4, 2019 Share Posted November 4, 2019 Hopkins quietly having a down year by his standards, which is odd considering the Texans and Watson seem to be thriving for most part. On pace for 121 catches, which would be a career high, however, just 1,182 yards and 7 TDs. His paltry 9.8 Y/C is a far cry from his 14.1 Y/C average entering the year. Link to comment Share on other sites More sharing options...
Recommended Posts