Comparing Tannehill to Luck,

"Finster wrote:

What stats show why the QB17 Dolphins are better vs the Pats(3-5) than we are vs the Bills(2-6), and also why we are just as good vs the Jets(3-5) as we are vs the Pats, when the Bills(29-35) and Jets(28-36) are clearly inferior teams, compared to the Pats(48-16), and why the Dolphins(29-35) aren't better vs the Bills and Jets.

Miami's offense matches up better against New England's defense than it does vs the ones run by the Jets and Bills. Thats especially been true at home. Its difficult to quantify, but the Bills have just had our number, even though they've sucked vs the rest of the league for years.

"Unlucky 13 wrote:

Miami's offense matches up better against New England's defense than it does vs the ones run by the Jets and Bills. Thats especially been true at home. Its difficult to quantify, but the Bills have just had our number, even though they've sucked vs the rest of the league for years.

While that's a valid theory, at least a partial theory, it is a theory, and therefor an opinion, not a statistical analysis, which was what I am specifically asking for.

"Unlucky 13 wrote:

Miami's offense matches up better against New England's defense than it does vs the ones run by the Jets and Bills. Thats especially been true at home. Its difficult to quantify, but the Bills have just had our number, even though they've sucked vs the rest of the league for years.

-47.9 should cover it.

http://football-players.pointafter.com/stories/10227/least-valuable-worst-players-nfl-2015#28-Miami-Dolphins-Dallas-Thomas

Dallas Thomas
PFF grade (position rank): -47.9 (82nd)
2015 salary cap hit: $761,594

Thomas was the worst guard in the NFL this season, according to Pro Football Focus. He ranked in the bottom two for both pass blocking and run blocking, and Ryan Tannehill's development certainly wasn't aided by Thomas' poor play.

Thomas gave up 10 sacks, the most of any guard, and 36 QB hurries, which placed him in the top three. It'd be a shock if new Dolphins coach Adam Gase decided to give Thomas another shot at a starting job in 2016.

"Finster wrote:

What stats show why the QB17 Dolphins are better vs the Pats(3-5) than we are vs the Bills(2-6), and also why we are just as good vs the Jets(3-5) as we are vs the Pats, when the Bills(29-35) and Jets(28-36) are clearly inferior teams, compared to the Pats(48-16), and why the Dolphins(29-35) aren't better vs the Bills and Jets.

You say you want an answer through statistics. Not sure one can provide that given the paucity of stats we have, but the first step is to see how likely/unlikely those events are if you assume the Dolphins are an average opponent of those teams (average opponent takes into account whatever effect that team has on its opponents so it's not a bad start).

Once again, you do a coin flip (with a weighted coin) and ask what the probability is of obtaining X wins out of N games for that coin. For NE, the weight on the coin is 0.75 = 48/(16+48), and the probability of an average opponent winning at most 3 out of 8 games is 88.6%. So the Dolphins having a 3-5 record against NE is highly likely if you assume the Dolphins are an average NE opponent and there's nothing really to ask there.

The comparable stats for the Bills is 9.14% and for the Jets is 23.7%. Normally, statisticians won't reject a hypothesis unless those numbers are less than 5%, but forget that for a moment and you see at least what needs to be explained here: The one real anomaly is how the Dolphins do against the Bills relative to how other teams do against them.

And maybe there's a simple explanation for that: our crappy OL vs. their great DL + a QB with not-too-great pocket presence.

Just read the Colts allowed the most QB hits in the last 4 year span (Miami second). Some advanced stats such as time to pressure, time to hit etc. would reveal more. But man, Luck's sack numbers are great compared to the hits they allow.

Question: Fantasyland. Indy offers Luck straight up for Tannehill. Who doesn't make the trade.

Following up on the "you build your team to win the division theory" w.r.t Indy having an abnormally large 17-2 record against the AFC South, the win % of a division winner from 2012-2015 against other division opponents is 75.5% while the win % of a division winner against non-division opponents is 68.1%.

So that is at least some evidence that division winners are built to win in their division more than they are against the rest of the NFL.

How does that change the likelihood of a 17-2 record? As I pointed out before the probability of an average opponent (68.8% win probability) of the AFC South winning 17 or more games out of 19 was 3.68%. Well.. the 68.8% win probability of an average AFC South opponent is eerily similar to the 68.1% win probability of a division winner against a non-division opponent. All I'm saying is that simplifies any analysis that attempts to "adjust" for a team being built to win a division (otherwise it's a more convoluted process).

Point is, we can calculate the probability now of an average division winner having a certain record against the non-division winning AFC South teams. It's the same calculation as before but replacing 68.8% with 75.5%. The result is that the probability of an average division winner having at least 17 wins against the AFC South teams is 12.1%.

I know 12.1% looks small, but given the really small sample size, no statistician would reject the hypothesis that Indy's seemingly abnormally good record against AFC South teams was due to random chance IF we adjust for whatever it is that helps a division winner beat other division teams.

Summary of several posts: none of the stats posted so far (including passer rating) are discerning enough to explain the variances in W/L in this case. Which just really goes back to saying it's maybe better to rely on the collective intelligence of experts in this case than try and figure out what's going on through stats.

Lmao

Comparing two qb's stats is meaningless because of the complexity of the sport. This isn't baseball where you can put a player on a different team and he'll produce the same as he's always been. We also cant easily quantify everything that happens on the football field. You can say tannehill has played as well as luck in their careers up to this point but when you actually watch them play you'd see luck is the better guy

"cbrad wrote:

You say you want an answer through statistics. Not sure one can provide that given the paucity of stats we have, but the first step is to see how likely/unlikely those events are if you assume the Dolphins are an average opponent of those teams (average opponent takes into account whatever effect that team has on its opponents so it's not a bad start).

Once again, you do a coin flip (with a weighted coin) and ask what the probability is of obtaining X wins out of N games for that coin. For NE, the weight on the coin is 0.75 = 48/(16+48), and the probability of an average opponent winning at most 3 out of 8 games is 88.6%. So the Dolphins having a 3-5 record against NE is highly likely if you assume the Dolphins are an average NE opponent and there's nothing really to ask there.

The comparable stats for the Bills is 9.14% and for the Jets is 23.7%. Normally, statisticians won't reject a hypothesis unless those numbers are less than 5%, but forget that for a moment and you see at least what needs to be explained here: The one real anomaly is how the Dolphins do against the Bills relative to how other teams do against them.

And maybe there's a simple explanation for that: our crappy OL vs. their great DL + a QB with not-too-great pocket presence.

However, reality can't be overlooked when dealing with stats Brad, the weighted coin theory falls flat when compared to reality, the Pats were 12-4 each of the last 4 seasons, so obviously 3-5 cannot be expected, because that would leave the Pats at around 10-6 every year, taking into account that they play good and bad teams as well as avg, so it all comes out in the wash, and 4 years is a solid sample size.

So since this is not theoretical, but based on actual data, the weighted coin theory has to be thrown out, since it is proven to be inaccurate in this example.

Since we play in the same division as the other teams, we are playing the same schedule, and our record is basically the same as the Jets and Bills, yet they have a better record vs us, when one would come to the conclusion that we should be going near 500 vs those teams, based on common opponents and nearly identical records.

So, we are still left wondering where are the stats to explain my original question(s).

"The Finest wrote:

Comparing two qb's stats is meaningless because of the complexity of the sport. This isn't baseball where you can put a player on a different team and he'll produce the same as he's always been. We also cant easily quantify everything that happens on the football field. You can say tannehill has played as well as luck in their careers up to this point but when you actually watch them play you'd see luck is the better guy

That is complete bull.

yes the sport is complex, however you can compare two quarterback's statistics and find meaning.

"Finster wrote:

However, reality can't be overlooked when dealing with stats Brad, the weighted coin theory falls flat when compared to reality, the Pats were 12-4 each of the last 4 seasons, so obviously 3-5 cannot be expected, because that would leave the Pats at around 10-6 every year, taking into account that they play good and bad teams as well as avg, so it all comes out in the wash, and 4 years is a solid sample size.

So since this is not theoretical, but based on actual data, the weighted coin theory has to be thrown out, since it is proven to be inaccurate in this example.

Since we play in the same division as the other teams, we are playing the same schedule, and our record is basically the same as the Jets and Bills, yet they have a better record vs us, when one would come to the conclusion that we should be going near 500 vs those teams, based on common opponents and nearly identical records.

So, we are still left wondering where are the stats to explain my original question(s).

No you're not understanding what I did. Probably my explanation wasn't good enough.

OK.. Pats have been 12-4 past 4 seasons, so they have a steady 75% win probability. The question I answered was NOT what the probability of exactly 3 wins out of 8 games is. I answered what the probability of at MOST 3 wins out of 8 is (so you win either 0,1,2 or 3, but not more).

Why did I do that? Because the probability of any specific record is relatively small. For example, the probability of winning exactly 3 out of 8 against the Pats if you randomly choose an opponent is 20.77%. The probability of winning exactly 2 out of 8 against them is 31.14%. Winning exactly 2 out of 8 is the MOST likely thing to occur, but it's not 100% as you see (and as should be obvious).

So what statisticians usually do is calculate the probability of a range of possibilities. In this case, winning at most 3 out of 8. And because the Patriots have been so steady at 12 wins per year past 4 years, the coin flip model is about as realistic as you can get, with no other information provided.

So, knowing that read my previous post again. There's nothing to explain w.r.t our record against NE, but the Buffalo record and to a lesser degree the Jets record might need explanation. I think the one I provided is the most obvious one based on an "eye test".

"jdang307 wrote:

I was being facetious ;)

I completely agree with your last statement. His first few years were historic. After that, he was one of the best but not the greatest.

I think he was the greatest. he took average teams and at least made the playoffs. those teams were horrible on defense at times. 34 pts a game and you still lose , u g l y

"cbrad wrote:

No you're not understanding what I did. Probably my explanation wasn't good enough.

OK.. Pats have been 12-4 past 4 seasons, so they have a steady 75% win probability. The question I answered was NOT what the probability of exactly 3 wins out of 8 games is. I answered what the probability of at MOST 3 wins out of 8 is (so you win either 0,1,2 or 3, but not more).

Why did I do that? Because the probability of any specific record is relatively small. For example, the probability of winning exactly 3 out of 8 against the Pats if you randomly choose an opponent is 20.77%. The probability of winning exactly 2 out of 8 against them is 31.14%. Winning exactly 2 out of 8 is the MOST likely thing to occur, but it's not 100% as you see (and as should be obvious).

So what statisticians usually do is calculate the probability of a range of possibilities. In this case, winning at most 3 out of 8. And because the Patriots have been so steady at 12 wins per year past 4 years, the coin flip model is about as realistic as you can get, with no other information provided.

So, knowing that read my previous post again. There's nothing to explain w.r.t our record against NE, but the Buffalo record and to a lesser degree the Jets record might need explanation. I think the one I provided is the most obvious one based on an "eye test".

Ok, so the coin flip theory is basically saying that we are doing about as well as can be realistically expected, which is pretty evident to most of us fans, which is part of my point, because why then are we 5-11 vs the Jets/Bills over the last 4 years, who have also been avg teams.

Keep in mind Brad, that I'm looking for a statistical analysis to explain it, as I have my own eye test theory, and that it's really an exercise, but I do appreciate you input.

"dolphin25 wrote:

I think he was the greatest. he took average teams and at least made the playoffs. those teams were horrible on defense at times. 34 pts a game and you still lose , u g l y

I agree completely, I think a very strong case can be made that no QB has ever done more with less.

"jdang307 wrote:

Just read the Colts allowed the most QB hits in the last 4 year span (Miami second). Some advanced stats such as time to pressure, time to hit etc. would reveal more. But man, Luck's sack numbers are great compared to the hits they allow.

Funny, I just saw that as well.

Listen dudes, Tanne will not make it and be considered great until we see him on TV, on a national commercial ad, being a shill for some big s--tbag corporation, pedaling some s--tbag product.

"dolphin25 wrote:

Tannehill is not football smart though.

Tannehill's segment with Gruden contradicts this statement.

"Finster wrote:

Ok, so the coin flip theory is basically saying that we are doing about as well as can be realistically expected, which is pretty evident to most of us fans, which is part of my point, because why then are we 5-11 vs the Jets/Bills over the last 4 years, who have also been avg teams.

Keep in mind Brad, that I'm looking for a statistical analysis to explain it, as I have my own eye test theory, and that it's really an exercise, but I do appreciate you input.

OK.. well like I said I doubt there is any statistical analysis that will pin down a solution for you because you'll probably need stats they either don't publish or even record. In any case, the only way I see doing this is to take whatever stats you think are relevant and then calculate how many fewer wins one would expect given each stat.

For example, take sacks given up vs. Buffalo compared to what we normally give up. We gave up 184 sacks from 2012-2015, or 2.875 sacks allowed per game, or 23 sacks over 8 games. The total number of sacks given up vs. Buffalo in those years was 27, so that's 4 extra sacks. The question now is how do sacks relate to wins? They are negatively correlated at -0.28 to wins meaning each extra sack decreases the number of expected wins, but by how much I have no idea.

Figure that out and now you can redo the coin flip thing I did to see how much that changes the result. Do this for more and more stats until you get a result you think explains the win-loss record. One thing you'll have to be careful of is "double counting" the effects of two different stats that are really measuring the same thing.

Anyway, if you or anyone else wants to actually go get the relevant data for some hypothesis you have on this Buffalo anomaly and you need help analyzing it, I'm glad to help. But like I said, I doubt you'll find your answer through stats.

"Dol-Fan Dupree wrote:

That is complete bull.

yes the sport is complex, however you can compare two quarterback's statistics and find meaning.

Meaningless was the wrong word but the fact of the matter is is that there are still so many variables that go into evaluating the skillset of two different players. You would be mislead into believing that tannehill is a slightly better qb than luck by looking at statistics

"Sceeto wrote:

Listen dudes, Tanne will not make it and be considered great until we see him on TV, on a national commercial ad, being a shill for some big s--tbag corporation, pedaling some s--tbag product.

He doesn't have a cool name like luck that can be used as a double entendre

"The Finest wrote:

Meaningless was the wrong word but the fact of the matter is is that there are still so many variables that go into evaluating the skillset of two different players. You would be mislead into believing that tannehill is a slightly better qb than luck by looking at statistics

There are so many variables, however the person who used these statistics did a good job utilizing the variables to show that at the very least they are not far off.

Too much of football analysis is based on "eyes" and "feeling". I see a lot of that when it comes to Luck. Heck I see a lot of that when it comes to quarterbacks period. With "X factor" Or "put the team on their back" or whatever feeling/eyes words they use.

"Pauly wrote:

Passer rating has also continued to rise across the league.
2015: 90.31
2014: 88.71
2013: 86.01
2012: 85.63
Average 87.67

Over Luck's career the league average has been 87.67. His in division defenses have allowed 90.62.

Over Tannehill's career he has had the same league average of 87.67, but his divisional defenses have allowed a passer rating of 83.41

Returning to this point for a minute.

Hypothetically, what would an average NFL QB have done for Indy or Miami over the last 4 years.

First step is to adjust the NFL average to remove your in division opponents.
Miami. Divisional opponents: 83.41. Adjusted league average: 88.11
Indy. Divisional opponents: 90.62. Adjusted league average: 87.36

(i.e. taking out the AFC East raises the rest of the league average and taking out the AFC South lowers the rest of the league)

Second Step. Take 6 games against divisional opponents and 10 games against adjusted league average and then divide by 16 to get an average season rating.

Expected career passer rating from an average NFL QB 2012-2015.
In Miami: 86.3
In Indy: 88.6.

Conclusions.
1. If the QBs were of equal skill the Indy QB should have a career passer rating 2 to 2.5 points higher than the Miami QB simply due to differences in the divisions.
2. Tannehill has underperformed against an average NFL QB by 1.1 rating points
3. Luck has underperformed against an average NFL QB by 3.6 rating points.

"jdang307 wrote:

J

Question: Fantasyland. Indy offers Luck straight up for Tannehill. Who doesn't make the trade.

I don't simply because of Luck's performance against the AFC East. Which is even worse when you take the playoffs into consideration.

Based on what he's done so far it would be locking us into 1-5 or 0-6 against the AFC East. Bellichek in particular has treated Luck like he was a middle schooler ina biology class and Luck was a frog.

"The Finest wrote:

Meaningless was the wrong word but the fact of the matter is is that there are still so many variables that go into evaluating the skillset of two different players. You would be mislead into believing that tannehill is a slightly better qb than luck by looking at statistics

I think that a strong case can be made that Tannehill is a better passer than Luck.

As for the other things that go into the assessing a QB, the stats become a lot murkier.

"cbrad wrote:

OK.. well like I said I doubt there is any statistical analysis that will pin down a solution for you because you'll probably need stats they either don't publish or even record. In any case, the only way I see doing this is to take whatever stats you think are relevant and then calculate how many fewer wins one would expect given each stat.

For example, take sacks given up vs. Buffalo compared to what we normally give up. We gave up 184 sacks from 2012-2015, or 2.875 sacks allowed per game, or 23 sacks over 8 games. The total number of sacks given up vs. Buffalo in those years was 27, so that's 4 extra sacks. The question now is how do sacks relate to wins? They are negatively correlated at -0.28 to wins meaning each extra sack decreases the number of expected wins, but by how much I have no idea.

Figure that out and now you can redo the coin flip thing I did to see how much that changes the result. Do this for more and more stats until you get a result you think explains the win-loss record. One thing you'll have to be careful of is "double counting" the effects of two different stats that are really measuring the same thing.

Anyway, if you or anyone else wants to actually go get the relevant data for some hypothesis you have on this Buffalo anomaly and you need help analyzing it, I'm glad to help. But like I said, I doubt you'll find your answer through stats.

Cbrad,

I really want to thank you for doing all this work with the binomal probability. It has been hugely helpful, and I've learned a lot too.