Taking a closer look at ESPN's win probability model and this spike in Cleveland's WP was caused by a 9 yard Baltimore run on 1st & 10, which caused CLE WP to go from 55% to 75%
????????????????
In ESPN's win probability model, a 4 yard Trace McSorley run on 1st & 10 with more than 13 minutes left increased Cleveland's win probability from 18% to 34%
Packers with a 4% chance of winning down by 3 with 2nd & 10 from the opponent 47 with 45 seconds left
Anyway hopefully ESPN fixes this sometime but the point of this thread:
--> Because 4th down and 2-point decision models are built on underlying win probability models, view ESPN's numbers for these with a very, very, very, very healthy dose of skepticism
Or in other words, if ESPN's model makes one recommendation on a 4th down and a different model (PFF for example) makes a different recommendation, it's a cop out to say that different models give different recommendations because we know for a fact that ESPN's isn't good
Important clarification from Brian here- the numbers on the WP charts on ESPN's page are indeed wrong in some cases but the actual numbers used in decision tools are the corrected version, which is good
My model actually thinks the Browns going for 2, down 8 last night was a bad decision- it didn't think the difference between being down 6 vs 7 was worth the additional risk
So I think it's time to cancel this chart (sorry whoever made the chart) because it assumes "convert 2 --> kick PAT" results in a win and that is obviously not the case
"They’ll only have him for another 16 games" is what would happen if he DID sign the tag, no long-term deal were reached, and he weren't tagged against next year.
So again, this tweet doesn't make sense
Note that this appears to be a change in strategy by Frank Clark's agent, who said this on the record in October 2018 espn.com/nfl/story/_/id…