Right, but I'm not ready to discard traditional passer rating as the gold standard for QB ability over the long haul, precisely because QBR isn't fully transparent. However, given that traditional passer rating is so much less valid a measure of performance for single games than for larger samples in my opinion, I don't feel like there's a sacrifice involved in replacing it with QBR for single games. In other words, if both measures can't be distinguished from garbage for single games, then you might as well go with the one that has the better ingredients. Over the long haul, however, there's no reason to discard traditional passer rating because it does a valid job of measuring performance without the transparency concerns. I suspect this is reflected objectively as well, where there is likely a convergence between traditional passer rating and QBR as a function of sample size. And so the measures are probably interchangeable over the long haul, in which case it makes sense to go with the fully transparent one. Over the short haul, however, if both are garbage then the more valid one is likely the one that has more ingredients related to winning. Again I think you have to give QBR some credence when you watch last night's game for example and find that the traditional passer rating differential was about 9 points, whereas the QBR differential was about 20. Mahomes certainly outplayed Garoppolo by the bigger margin of the two measures in my opinion.
You're making so many illogical arguments at once. In this quote alone you say that "traditional passer rating is so much less valid" than QBR for single games, but then say that "both measures can't be distinguished from garbage". How does that work?? If both are garbage then one can't be "so much less valid". You also admit there's a cost to QBR not being transparent. So if "both measures can't be distinguished from garbage for single games" how is it that the cost of being non-transparent doesn't force you to choose the transparent method? Finally, the "validity" (how accurate it measures QB ability) of passer rating and QBR do not depend on transparency. Their validity is whatever is it regardless of transparency. So you can't argue one is more valid for small sample sizes but not for larger ones. No you either argue QBR validity > passer rating validity for all sample sizes or vice versa. Trying to argue some sample size dependent relation puts you in an even worse position: arguing that logic doesn't matter.
I thought so too. He actually set a Super Bowl record for the most rushing yards by a WR! It's one of those meaningless stats but still, it's pretty darn cool as well...
Granted he had a 78 passer rating in the Super Bowl to a 49er defense that averaged giving up 77.7, but his first two post season games this year he posted a 135 and 120. What would he need to do to impress you?
I started a thread recently about Williams, because I feel the same way. It got only a handful of replies, and I was told a Williams is just playing on a team with great players which makes him seem better. To me, he looks like a different player altogether than he did in Miami.
Almost like having a QB at Mahomes' level suddenly getting hot opens things up. You're just being needlessly edgy and contrarian. Dude just became the youngest Super Bowl MVP after blowing the doors off the Texans and Titans and you are trying to emulate Shania Twain? Just rings hollow and smells like sour grapes. Overall, aside from their first TD drive, the San Fran defense was holding and doing their job and exploiting the KC weaknesses (namely, their OL against the SF DL). But when **** was really on the line, dude killed it. That blink and you'll miss it explosion he can do is just something else. The Niners defense had to take a moment to breathe and he took that moment and just crushed them.
Let me see if I can state this more clearly. Here is how I view the situation: Validity for small sample sizes Traditional passer rating: poor QBR: unknown due to incomplete transparency but intuitively better than traditional passer rating in my opinion Validity for large sample sizes Traditional passer rating: good QBR: unknown due to incomplete transparency and likely non-significantly different from traditional passer rating So if I'm dealing with a small sample size (like one game), I'm willing to sacrifice the incomplete transparency for something that to me is intuitively more valid. For larger sample sizes there's no sense in exchanging traditional passer rating for QBR because we know the former is valid, and we don't know the validity of the latter. Does that help at all?
In my opinion, the MVP of the Super Bowl before the fourth quarter was the San Fransisco front line. They were dominating and Mahomes never had a chance to get comfortable.
This is again where we need something better than traditional passer rating as a measure of single-game performance, because the 78.1 traditional passer rating says nothing about Mahomes's clutch play in the fourth quarter. There is nothing in traditional passer rating that contextualizes the gravity of that performance. Likewise when a team is up by three touchdowns and the opposing quarterback amasses a ton of "garbage time" stats, there is nothing in traditional passer rating that contextualizes the meaninglessness of that performance. So you could have Mahomes doing what he did in the fourth quarter last night, and another quarterback who's down by 35 points and does exactly the same thing Mahomes did in terms of traditional statistics, and those two QBs would have the same traditional passer rating during that part of the game.
Yes that helps in that it gets rid of all problems except one: even if QBR has unknown validity it's either always greater than passer rating or always less regardless of sample size. And validity here means "accuracy in estimating QB ability". So pose this question to yourself: why would you ever choose any method that has less validity regardless of its transparency? You wouldn't. Transparency should have nothing to do with the decision IF you already assume one has greater validity in ANY situation. The reason I go with passer rating is because I make no assumptions about validity. We really don't know what the validity of either measure is, including for traditional passer rating. You can remove a lot of random variation with large sample size, but there's no independent measure of QB ability to test their validity against so we don't really know. And so it's clear, it should be obvious that passer rating can be improved by incorporating win probability. I'm not arguing against that. I'm just saying that under no circumstances do you choose a non-transparent method as a replacement. I mean.. you can directly incorporate win probability as a weight on passer rating and it's fully transparent.
Bosa and Buckner had huge games. If Jimmy had connected on that pass to Sanders and they won though, it would have gone to him. But that's just it, the SF DL is one of the best we have seen, and for 3.5 quarters they did what they needed to do. But that last half quarter ... and they weren't even really bad, Mahomes was just that good (and the playcalls were good and supporting cast did their jobs).
I know it would have gone to him. I would have disagreed since that line was why they were winning that game
The traditional passer rating reflects that the SF defense held Mahomes in check for 3/4s of the game. It includes the effect of 2 bad passes being intercepted. The fact that Mahomes was able to overcome the deficit is purely due to the Chief’s defense being able to make enough plays to keep the game in reach at the start if the 4th quarter.. However as it is also passer rating it does not include some very valuable runs Mahomes made. On the passer rating differential Jimmy G’s rating for the games was 69.2, for a 8.9 pr differential in the Chief’s favor. The traditional passer rating reflects that - Mahomes did not play well all game - Mahomes played better than Garoppolo Traditional passer rating did it’s job well in this game. Edit to add My view os ESPN’s QBR is that cbrad is too kind and generous to it. However ESPN says the QBR for Mahomes was 61.6 (i.e. a little above average) and Garoppolo’s was 40.6.
Hold on now, the 49ers were up 20-10 and 95% likely to win the game with 7:26 left in the fourth quarter. The fact that Mahomes overcame that situation surely shouldn't be attributed entirely to the Chiefs' defense. The Chiefs went from 5% to 100% likely to win that game in the span of a half a quarter. I suspect quarterbacks are associated with those kinds of swings in win probability very rarely.
I agree with others- Mahommes struggled all day long and then had an amazing 4th quarter to make up for it. Jimmy G played okay but never did enough to seal the victory...I hate it that he missed on that bomb. It was a good game though and the defenses were the stars of the day; I don't think anyone expected a low-scoring game.
That's a stark difference from their traditional passer ratings in the game, however. Traditional Passer Rating Z-Scores Mahomes: -1.22 Garoppolo: -2.09 QBR Z-Scores Mahomes: 0.45 Garoppolo: -1.30 So there's a 0.79 difference in Garoppolo's z-scores, yet a 1.67 difference in Mahomes's z-scores.
You won't find many QBs with YPA below 7, two picks, and four sacks, and at the same time nearly a half a standard deviation above average in QBR on the game.
I'm beginning to wonder when players like him and Tannehill move on and do better; is it because of the new team, or did the transfer cause them to refocus and that made the difference. I would love to know from those players if there was something they learned under Gase that helped them in their new situation.
Umm.. you're calculating z-scores wrong for ESPN's QBR. The 0-100 scale represents "percentiles". https://en.wikipedia.org/wiki/Total_quarterback_rating So you need to convert percentiles to z-scores directly. You can use this tool (make sure to use "One-sided"): https://measuringu.com/zcalcp/ So 61.6 corresponds to a z-score of 0.295 while 40.6 corresponds to a z-score of -0.2378 for a difference of 0.5328. So it's actually the other way around. ESPN's QBR says there was a smaller difference in z-scores. Regardless.. it doesn't matter what ESPN's QBR says. No one knows what it's actually calculating.
I generated z-scores by using the QBR mean and standard deviation in the league in 2019, same as for traditional passer rating.
Why would that have any bearing on how much the QBs last night deviated from the league norm in 2019 in terms of the mean and the dispersion involved in that distribution?
Because they already transformed the scale for you. So if you want to compare on the same z-score scale you have to use the percentile to z-score conversion like I said. And when you do that you get the opposite result from what you were saying.
But they haven’t transformed the ratings in terms of the 2019 season distribution. What if hypothetically every quarterback in the league in 2019 had a QBR above 80? A QBR of 50 couldn’t possibly then represent the 50th percentile in terms of the league distribution in that season. Isn’t this the same reason why you adjust traditional passer ratings by year, because the league mean passer rating changes by the year? QBR could likewise have a different mean every year.
According to that link it's percentile based on all games since 2006. Obviously, they don't go back and change prior ratings, but that's how it's calculated. So this is actually another flaw in that rating system. Percentiles aren't "measures" lol (I didn't even know that until I read that link carefully). They wanted a 0-100 scale so some idiot thought let's use percentiles. Except that going from 50th percentile to 51st percentile is FAR easier in general than going from 90th to 91st. z-scores are measures in that one unit z-score difference means the same thing everywhere on the scale. So they tried to create a rating system that is already adjusted for era, but they should have used z-scores not percentiles lol. More reason to ignore ESPN. They didn't even create a measure in a technical sense.
Bruh. He had great games against the Texans and Titans...and had a great clutch comeback in the SB. Who in history, especially at his age, would you put ahead of him? He's gonna be a legend when it's all said and done.
Williams was always a strong RB but he had injury concerns and had a few years there where he wasn't reliable. We always had a slightly better RB ahead of him in the rotation when he was healthy so he never really got his shot in Miami except when others were out hurt themselves. The time he played for us though was productive and he was excellent catching out of the backfield. In fact, Gase recently said that he messed up with Damien and should have featured him more...even back then he had all the tools to be a star.
I was always a big Williams fan. He hasn't changed much - great 3rd down back and a key reason why KC is a Super Bowl Champion
Dunno, he didnt impress me in the SB like the greats did when they faced an elite defense. I just look at Brady and that Seattle defense few years back when he had an amazing game facing an elite defense.
What are the alternatives for quantitatively representing a quarterback's performance in a single game, and how do they compare to QBR?
True...but just compare 2002 Brady to 2019 Mahomes and tell me they're anywhere in the same universe for skill / talent... Took Brady several years to reach "elite" status...not Mahomes.
And I wouldn't call it an amazing game...it was good. Took a bonehead playcall on the Seachickens part for them to even win, they had lost it until that last second miracle.