Tua...''He's the greatest prospect ever''

I have NEVER done that. Either you can't comprehend what I'm arguing or you refuse to read what I write. Just take the first paragraph of the first post I made in response to your claim:

Now pray, tell me how from that you could ever get that ascribe wins and losses to a single player. Repeatedly in post after post, including all the discussion with danmarino I showed it was a smaller fraction.

Point is, since the QB influences the outcome, the outcome tells you something about the QB. That is unavoidable logic. And that's precisely why with large sample size wins become a fairly good individual stat.

 

Individual examples don't help your case at all when I'm arguing 10-20% of the outcome is due to QB. That obviously means there are many examples where you can point to other factors that were more important.

What you cannot argue against is that wins become a good correlate to individual ability over large sample size. And that's why wins ARE an individual stat, even if they aren't anywhere near as strong as other team stats like passer rating that also can be treated as individual stats because the individual affects the outcome.

 

I think we're closer to agreement here than not.

If someone wants to claim that QB play affects wins and losses, typically, more so than any other individual player I'll agree. That's why I think the QB position is the most important. However, using wins and losses to judge a QB is going to lead to a false narrative. Like Dilfer was a better QB than Marino and on the same level as Rodgers and Favre. Or Plunkett and Eli were on par with Peyton.

if wins and losses are individual stats they are nearly meaningless. I'm sure we could find out that some team, over 50 years, historically played badly on September 3rd when they play in the late game. Technically that's a stat, but it's pretty meaningless.

 

Trent Dilfer had a career 58-55 record. Plunkett had a career 72-72 record. Eli Manning had a career 117-117 record.

Compare:

Marino had a career 147-93 record. Rodgers had a career 148-75-1 record. Favre had a career 186-112 record. Peyton Manning had a career 186-79 record.

Win% is a VERY good individual stat with large sample size!

 

Yeah I'm not upset or not upset with the situation, I just find it interesting. Generally when a player is being touted as elite, or borderline, the team does what it needs to, to resign a bit earlier cause it's usually cheaper. So I'm just sort of perplexed by it all.

 

I don't necessarily agree that the outcome tells you much of anything about the QB, just because the QB influences the outcome.

 

Without actually comparing the players on those teams I don't find your conclusion to be particularly true.

 

What part of a QB's play is, or should be, attributed to the players around them? If Brady was drafted by the Lions in 1999, with everything else being equal, do you think he wins even one SB? I think the Lions had like 4 winning seasons from 1999 to 2020. The Lions had 3 playoff wins (all WC round) from 1958 to 2023. That had something like 15 winning seasons over that period in time. That's 15 out of 65 years. Are you to tell me that ONE play would have changed that for them? I doubt it.

I don't think that ONE player is responsible for 10-20% of a teams success. How do you even quantify that when there are 10 other players that directly affect the play of the QB? When there are 52 other players that directly affect the outcome of any game? This seems more like the "chicken or the egg" question. What's more important, the QB who threw the pass, the WR who caught the pass, the other WR who set up the pick, the o-lineman that blocked for the QB, or even the coach who called the play?

I mean, yeah, we can look at wins and losses, who the QB was for those wins and losses and try to make a "stat", but it's not telling us anything except that the QB with the most wins played on some really good teams. Teams that ENABLED him to play well and TEAMS that played well themselves...hence the wins > losses.

 

It's math resnor. That's not something that can be argued. The only issue here is how much it tells you. If I'm saying 10-20% then you need LARGE sample size. See the post above on comparison of win% over a career for Plunkett, Eli, Dilfter vs. Marino, Favre, Rodgers and Peyton. You can see how good an individual stat win% becomes with large sample size.

Also note that I didn't select those examples, danmarino did. It actually proves my point not his.

 

Those aren't really cherry-picked examples. You'll find that in general to be true (very few exceptions). A general trend like that, where career win% is highly correlated with perceived individual ability, can't be explained by assuming that in almost every case it was a systematic difference in everything EXCEPT the QB.

No, statistically speaking that hypothesis is dead in the water.

 

How to quantify the percentage is a tricky question. But like I said in that first paragraph in the first post to TDK it starts with some assumption about what portion of passer rating is due to the QB. Maybe you should tell me how much of passer rating you think is due to the QB and I can take it from there because the variance explained in win% by passer rating across all NFL history is about 40% (square of the correlation).

So let's say you think 50% of passer rating is due to the QB. That forces you to accept (statistically) that 20% of win% is due to the QB, over large sample size of course. If you say 1/3 of passer rating is due to the QB, then that forces you to accept that statistically 15% of win% is due to the QB. If you say 1/4 of passer rating is due to the QB, then it's 10%. I think my estimate of 10-20% is pretty good.

As far as your question about Brady, I wouldn't make predictions about SBs, but there's no question the Lions would have been MUCH better with him. Again, 10-20% of win% is likely due to the QB.

 

Are you using wins and losses to prove that wins and losses make a QB better? That's some circular logic dude. lol

What would Marino's win/loss record be if he played for the Pats from 1999 to 2020?

Again, how much of a QB's success should be attributed to their team? Amy ideas? I mean, there are 1000's of games where a QB had some good numbers, the offense as a whole had good numbers, but they lost because their defense played like crap. Or maybe some of those losses attributed to Rodgers was when the Packers were playing against Peyton's Colts? What if, for arguments sake, Manning was 10-0 against a Rodger's QB'd teeam. Yet when we look at their over all PR's in those games Payton had an 80 and Rodgers had a 105. Who was the better QB? I mean, in this example Rodgers is 0-10 in the wins and losses "individual stat". lol

 

I don't think anyone believes a QB doesn't affect the outcome of a game. Kickers, LB's, DB's, etc all affect the outcome of a game, also. However, that 10-20% seems awfully high since we can't really know how much of the QB's success in everything else besides wins and losses is due to the players around him.

 

No, I'm making reference to your own perception of how good QBs are. In your own post you implicitly made a distinction in ability between those QBs. So career win% is highly correlated to YOUR view of those QBs.

Probably similar to Brady. Both were elite QBs. What would have been interesting is to see how Belichick handled Marino lol.

Yeah, one can adjust passer rating and other passing stats based on other stats, like defensive points allowed. I actually did that years ago if you're interested, but those stats aren't updated for the last 4-5 years. Estimating how much of QB success is due to other offensive weapons is technically possible, but much harder with the data we have. You'd have to look at how passing stats change when individual players are replaced, like a WR or an OL. I can't do that analysis easily because writing a program to keep track of changes in starters is just way too time consuming, but that's how you would do it.

 

See my post above on how to estimate the effect of the QB on win%. I'll copy the relevant quote here. 10-20% seems like a very good estimate unless you think the QB is responsible for less than 25% of passer rating.

 

Well, there's incorrect math, no?

You know I'm a stat lover. I think they tell us A LOT. However, the stat has to be good. Wins and losses, for individual play, just isn't a good stat. And looking at wins and losses to prove that wins and losses show good or bad individual players is not going to cut it for me. I get that there's a correlation, but you know the rest. And this is a perfect example.

I mean, did the Ravens win the SB because Dilfer was the QB or not? it seems you're claiming they did. I think the Ravens won the SB because they were a great team.

 

What I posted isn't incorrect math lol. It's a simple inference from the concept of variance explained. And yes win% is a VERY good individual stat once you have large enough sample size. Why are you continuing to post examples of individual games when I'm saying large sample size is needed? That's almost like ignoring my argument. For individual games it's not a good stat. But that applies to every stat including passer rating.

Back to Tua. Tua's career 32-19 record tells you SOMETHING about Tua that you aren't seeing in someone like 30-32 Herbert. It's not irrelevant that Tua's win% is much higher. That's not ALL the surrounding cast because the QB influences the game a lot. It's also not irrelevant that Tua hasn't done well in do-or-die games. But as I said repeatedly it's absurdly small sample size, which is why for that particular argument it doesn't matter too much (yet). Doesn't mean wins can't be treated as an individual stat.

 

I think it's very difficult to determine individual stats in a game like football.

Saying that, I believe PR is a good stat to judge individual performance. Yes, someone had to catch and block, but it seems to be the best we currently have. Saying that, correlating PR with wins seems foolhardy. I mean, a QB with a great PR is typically going to be a better QB, but there are literally 1000's of games where the QB with the better PR lost. I understand that there are also 1000's of games where the QB with the better PR won, but why was the other team's QB's passer rating less? Defense has a HUGE affect on that. So if the team that had the better QB wins, was that because the defense he played was just worse than the other?

I think before I believe that wins and losses are an individual stat I need to know these things.


Last season, Tua's PR was about 10 points higher than Mahomes'. Why? Both KC and Miami were 11-6 in the regular season. Why? KC won the SB. Why? Well, I believe because KC had one of the best defenses in the league while the Dolphins had one of the worst. KC was the overall better TEAM.

 

OK.. so you believe that a TEAM stat — passer rating — is a good individual stat. You do realize the exact same issues you are raising with win% as an individual stat apply to passer rating, just not as much because passer rating is more influenced by the QB than win%. But in principle there is NO difference in what the underlying issue is. I think you need to reconcile that aspect first (in your own mind) before proceeding. You can't tell me to explain the contribution of X, Y and Z players other than the QB to win% before you can accept that as an individual stat when you refuse to do the same for passer rating. It's the same issue.

As far as explaining individual game outcomes (or even within-game outcomes), that's difficult. The best stats to follow there are win probability and EPA per play because that tells you how much each play contributed. Then it's up to you to decide subjectively how much each player contributed, until of course stats become advanced enough to help estimate that. That NFL Next Gen thing is at least attempting to go there.

Either way, I think it's clear now that win% over large sample size correlates with who YOU think are better QBs. That alone should give you pause in arguing it's not a good individual stat with large sample size.

And finally, correlating any stat with win% is extremely important because that's precisely how you determine which stats are important for winning! The goal of the game is to win.

 

btw.. these are the QBs in NFL history with at least 60% win% and 100+ wins in their career. Pray tell me that win% with large sample is not a good individual stat after seeing this list.

Brady: 251 wins, 75.4% win%
Favre: 186 wins, 62.4% win%
Peyton: 186 wins, 70.2% win%
Brees: 172 wins, 60.1% win%
Roethlisberger: 162 wins, 67% win%
Elway: 148 wins, 64.3% win%
Rodgers: 148 wins, 66.3% win%
Marino: 147 wins, 61.3% win%
Unitas: 118 wins, 65.2% win%
Montana: 115 wins, 71.3% win%
Wilson: 115 wins, 61.4% win%
Bradshaw: 107 wins, 67.7% win%
Kelly: 101 wins, 63.1% win%

If you're going to say that a stat that picks out those QBs as the best in history is not a good individual QB stat with large sample size, then you have a VERY different view of who the best QBs are in history than the great majority of people, and of course whoever votes for induction into the HoF.

Wins with large sample size is a VERY good individual stat!

 

Maybe incorrect math wasn't the correct phrase.

How about, applying correct math in an incorrect manner?

2+2 does equal 4...but I can't then just say, "See, 2+2=4 so that means all butterflies are yellow." lol

Again, looking at a correlation isn't necessarily proving anything. For example, speeding kills. Referring to car crashes here. You've heard this phrase, right? And it's true that if you're in a car crash going 100mph you're more likely to die than if you were going 10mph. Well, what if I then said, "Corvettes are one of the fastest cars on the road so if you drive a Corvette you're more likely to die"? That's not true at all.

These win/loss records of seemingly all time great QB's means they also played on good teams. Look at Steve Young in Tampa compared to Steve Young in San Fran. Look at these same QB's wins/losses when they played on teams with good to great defenses vs bad defenses. Their PR's may be very similar, but their wins/losses aren't. In 1984 and 1985 the Dolphins had the 4th and 11th ranked defenses, respectively. They went 14-2 and 12-4. Then, over the next 4 seasons, they were 8-8, 8-7, 6-10, and 8-8. Their offense (including Marino) was good to great ranking 1st, 7th, 17th, and 15th in the NFL. What happened? Their defenses ranked 26th, 16th, 24th, and 22nd. Overall, Marino played the same, yet the Dolphins failed to win more than 8 games and didn't make the playoffs.

Then the very next season they went 12-4...why? Defense ranked 4th that season. The next year they were 8-8 (24th ranked defense). The next year 11-5 (defense ranked 11th).

Marino played nearly identical each season and yet "his" wins and losses were up and down.

 

They saw his late season collapses + the concussions previous to last season, probably some clauses and money thing they are arguing about, but I think they’ll get it done before the season. I don’t think Tua wants to play with only 1 more year left on his contract.

 

Except that's nothing like what I did. I calculated variance explained (look it up) between passer rating and win%. That's the ONE part of the inference you cannot argue with unless you want to be accused by mathematicians of saying it's equivalent to arguing 2+2 is not 4. There's no arguing the correlation part.

See my post above. Look at the QB win% with large sample size list. Those are the QBs this stat you're disagreeing with says are the best QBs in history. Do you have serious issues with that list?

And let's turn this around. Which QB has a losing record with large sample size that YOU think is one of the best is history?

The evidence is staring you in the face: with large sample size wins are a very good individual stat.

 

I realize that. I also realize that there are far fewer people involved in PR than wins and losses. And to me that makes it a BETTER individual stat...not a PERFECT individual stat.

I mean, I think wins and losses can be claimed as an individual stat in the same way that I believe one paver on my back patio can be called my house. Technically it is my house, but if I showed you one paver you'd have no idea what my house looks like. If a paver is wins/losses then the framing, sheetrock, cabinets, insulation, subfloors and roof is PR. If I showed you a picture of my house in that unfinished stage you'd have a much better accurate picture of the finished product.

 

Look at the evidence of who the best QBs are based on win% with large sample size.
https://thephins.com/threads/tua-hes-the-greatest-prospect-ever.98555/page-47#post-3641473

It's showing you directly how good an individual stat win% is with large sample size. The only way out now is if you disagree with that list or show lots of great QBs with losing records.

So? Which "great" QBs had losing records over a long career? And how many can you find?

Win% is nothing like a paver lol.. it's almost like the entire house itself if you get enough data!

 

All of those QB's are white, also. Correlation doesn't equal...

lol

How about this, what are these QB's passer ratings in each of their wins and losses? I showed in Marino's case, and even though he played relatively the same each year from 1984-1992, that the Dolphins wins and losses fluctuated greatly. Did Brady win so many games because he was consistently playing on better teams? (Remember, in his only 2 seasons with below average defenses, both ranked 17th, he won 9 and 10 games) That's now Marino and Brady that had extreme differences in their wins and losses because their teams were below average even though they played relatively the same from season to season.

 

Don't dismiss my argument by misdirection lol. I think I've proven quite conclusively with that list that win% with large sample size is a VERY good individual stat.

Can you acknowledge that now? The evidence is overwhelming.

Also you don't have causation with passer rating either. So this is about correlation not causation.

As far as QB passer ratings for wins and losses the relationship is linear! That means that if you accept passer rating as a good individual QB stat you HAVE to accept win% as a good QB stat. See the relationship for 2002-2019.

 

There's no problem with the list. I think there's a problem assigning wins and losses to them. Why? Well, read above. I haven't checked everyone on this list, but Marino and Brady kind of prove that the TEAM needs to be good to great in order to win a lot of games even if their PR's are good to great.

So 10-20% is the entire house?

If you give Tiger Woods my 7 yr old sons plastic golf club set he's not going to beat me (most of the time...lol) on the golf course. If you put Brady on the 2007 Dolphins they aren't winning an appreciable more amount of games.

 

The list was generated purely through wins and losses. So if there's no problem with that list you are acknowledging that using wins with large sample generates a very good list of how good QBs are, i.e., wins are a very good individual stat with large sample size. Obviously, there are still other factors at work.

Any non-zero percentage becomes "the entire house" if you have large enough sample size. It's how statistical significance works (how to determine if something was due to random variation alone or whether there's a real difference). You can take the smallest effect size (actual difference between two distributions) and if you have large enough sample size a hypothesis test will show there is a "real" (i.e., statistically significant) difference between the two.

You simply need fewer samples the greater the influence of the QB on the stat. At 10-20% it seems like 100+ wins is sufficient.

 

I don't think PR is a good way to judge wins and losses either. I think PR tells us more, waaaaaaaaay more, about an individual QB than wins and losses.

Nothing in the NFL is in a vacuum so it's hard to break this all down. And I love reading what you're posting here, because I love stats. However, I just think there are too many moving parts in football to attribute any real meaning to wins and losses when it comes to individual players.

And I am just giving you a little bit of a hard time with the correlation ≠ causation stuff and I realize that any correlation means there is at least some sort of statistical indicator between the two. However, I just don't think showing a list of all time great QB's and their win/loss record is enough of a correlation. I explained why with my Brady and Marino win/loss records post. Does this not show that they played on good teams when they won? If they are playing relatively the same from year to year, and they win 14 games with the 4th ranked defense and win only 9 with the 24th ranked defense, that has to mean something, right?

 

Focus on that graph. IF you think PR is a good measure of QB ability, then because the relationship to win% is linear, that means win% is also a good measure of QB ability. There's no escaping that. You could escape that conclusion if the relationship was not linear, but it is.

And yes, obviously there are other factors at work. You'd get a strong correlation between defense and other aspects of the team with win% too. But that doesn't change the relationship with QB stats like passer rating.

 

Not always. Its possible to have a good QB who can't overcome a bad organization. Archie Manning was 38-110-3 for his career and Dan Fouts was 88-92-1. Both were good QBs, but from Win% you wouldn't know it. David Carr is another good example, he was certainly better than his 33-61 record would indicate.

I get what you are saying, a QB is a cog in the team machine so the better the QB the better the overall team performance should be but sometimes the rest of the machines is just too broke to overcome.

 

Correlation isn't 1 so obviously there are some exceptions. That's true for EVERY stat including passer rating. And none of the ones you mentioned are typically mentioned within the top 20 QBs of all time. So yes win% is a VERY good QB stat with large sample size. It's telling no one can provide more than a few counter-examples, which puts win% in the same league as passer rating (again the relationship is linear — if one works the other does too!).

The main difference is sample size. Passer rating works on passing attempts. It's somewhat reliable after only 150 attempts, and after 3-5 seasons you have a pretty good idea of where most QBs will be over their careers. Not true with win%. Probably need closer to 10 seasons. But it ends up being a VERY good QB stat in the end.

 

I think you are missing the point. QBs have a large impact on how effective the offense is so it stands to reason that if you have an elite QB you will have a better offense and likely win more games. So yes, elite QBs will typically have higher win percentages. But the correlation between win percentage and a QBs ability is too weak to be considered a meaningful statistic to use in judging the vast majority of QBs. There are just too many other factors that need to be considered.

Tua has a win% of 64% vs Burrow's 56%. Does that tell you anything about their respective abilities?

 

Are you even reading my posts? I pointed out the main difference is sample size and you probably need ~10 seasons with win%. And you respond by asking about Tua and Burrow, neither of whom have ~10 seasons of play. Read the post first.

Again the relationship is linear. There's no way around this. If you accept passer rating as a decent metric of QB ability then you are forced to do the same with win%. The only difference is sample size required.

 

Ask yourself why it is only correlated when you get to a very large sample size but isn't for smaller samples.

 

Bridgewater often got hurt.

 

I already explained that. If only 10-20% of win% is due to the QB but maybe 25-50% of passer rating is due to the QB, then you obviously need more samples for win%. Been explaining that for 2 pages. It's math.

Point is, win% is a VERY good QB statistic with large sample size. The only way to argue against this is either to say evidence is irrelevant OR to take the position passer rating is not a good measure of QB ability.

That latter position of denying passer rating as a good metric of QB ability would open up a different argument strategy on my part. Let the person arguing that randomly select 100 QBs and rate how good they are, say on a scale from 1-10. We calculate the correlation between those ratings and both passer rating and win% and you'll have confirmation that the correlations to that person's view of which QBs are good or not is very good for both metrics.

 

That is a single game. Do they get there without Mahomes?