The main argument for playing Joey Harrington as long as we did this year is that we needed to give him enough time to see if he could resurrect his career and start playing well in a different environment. In other words, we used part of the 2006 season to conduct the “Joey Harrington Experiment,†which for the sake of brevity I’ll refer to as “JHE†from here forward.
However, as much as many of us wanted to give the guy a chance to see if he could resurrect his career and become a good QB for us, at no point in the season did I personally think the JHE was more important than the success of the Miami Dolphins’ 2006 season, and I think I’m probably not in the minority in saying that.
Ironically, what I’m going to try to show here is that JHE was for all intents and purposes (read: probably unintentionally) treated by the coaching staff as though it was more important than the success of the Miami Dolphins’ 2006 season.
The people here (like me) who thought Harrington should’ve been yanked even though we were winning games with him at QB for a stretch will find their point of view supported here. It’ll take a lot of words and numbers to do that effectively, so please be patient and hear me out.
In a nutshell, what I’m going to try to show with statistics is that, even though we were winning as a team for a stretch, overall, Joey Harrington still wasn’t doing the things from the QB position that help teams win games. In fact, what I hope to show is that he was predominantly doing things from the QB position that typically help teams lose games, which means he should’ve been shelved even in favor of a QB with no regular season experience like Cleo Lemon. What’s more, he was helping the team lose games at a time when we needed to win every game because our record was near .500 relatively late in the season and our playoff hopes hung in the balance.
What Do QBs Do to Help Teams Win and Lose Games?
First the statistics. I have the data for the following offensive and defensive variables for every team in the league from the five seasons between 2001 and 2005 (159 teams total; the Texans weren’t around until 2002):
Number of regular season wins
Points scored and allowed
Pass attempts
Pass completions
Completion percentage
Yards per attempt
Yards per completion
Pass attempts per INT
Number of rushes
Rushing yards
Yards per rush
I ran a general stepwise regression analysis (http://www.kellogg.northwestern.edu/...B/Modeling.htm) to determine which of the above variables are the strongest independent predictors of wins. I excluded points scored and allowed, which are obviously very strong predictors of wins and would therefore block other variables from entering the equation.
According to the regression analysis, the strongest independent predictors of wins between 2001 and 2005 were the following variables, in order of strength of predictability; their zero-order correlations with wins are in parentheses:
1) Number of rushes allowed (-.70)
2) Yards per pass attempt allowed (-.46)
3) Yards per pass attempt (.54)
4) Pass attempts per INT thrown (.45)
5) Pass attempts per INT caught (-.34)
6) Pass attempts allowed (.43)
7) Number of rushes (.57)
These were also the only variables that significantly predicted wins, and taken together they account for 74% of the variance in wins. That’s pretty huge in statistical terms, and it means that if you know the value of these variables you can predict wins with 74% accuracy (gamblers would love that wouldn’t they?).
So what we have here is an analysis of team variables almost all the way across the board (almost every area of offense and defense), with the finding that among those team variables, two that are strongly associated with the QB position, pass attempts per INT thrown and yards per attempt (YPA), are among the seven significant predictors of wins. In other words, if a QB throws more attempts per INT and has a good YPA, he’s going to significantly help his team win games.
Likewise, of the seven predictor variables that aren’t offensively oriented, two of those are the defensive counterpart of pass attempts per INT and YPA, meaning that four of the seven predictor variables center around pass attempts per INT and YPA. More specifically, the offensive and defensive counterparts of those two variables account for 22% of the 74% I noted above, so the QB is affecting quite a large percentage of the game with those two variables alone.
To reiterate, the upshot is that if a QB’s pass attempts per INT and YPA are good, he’s significantly helping his team win games, and if they’re bad, he’s significantly helping his team lose games.
Why The Joey Harrington Experiment?
Back to the main topic. Again, the JHE was conducted to see if he could resurrect his career and become a good QB for us. So let’s compare Harrington’s performance in these areas in Detroit to his performance for us this season. If he improved significantly in these areas (pass attempts per INT and YPA), obviously we can say he turned his career around and started performing well for us, at least in terms of the variables that strongly and significantly predict wins and losses.
In Detroit, Harrington played in 58 games and started 55 between the years of 2002 and 2005 (almost the same years used above by the way). Overall, Harrington’s number of pass attempts per INT was 29.1, and his YPA was 5.6.
His 5.6 YPA was significantly lower than the league average of 6.8 between the years of 2001 and 2005 at a significance level of <.0001, which means there’s greater than a 99.999% likelihood that the difference between his YPA and the league average wasn’t due to chance.
His number of pass attempts per INT (29.1) was significantly lower than the league average of 33.8, and that difference was significant at the same level. In other words, it’s clear Harrington was underperforming in these areas in comparison to the average NFL QB, and therefore he obviously wasn’t helping the Lions win games with his performance in these areas.
Again, however, equipped with only the knowledge we had about him at the beginning of this season, we could speculate that his performance could’ve been more about the environment in Detroit than anything else. And things were looking up when we heard things during camp and the preseason like, “Joey is happy in Miami; Joey’s having fun playing football again; Joey likes the system here better because it tells him exactly what to do, whereas in Detroit it was a mess.â€Â
How Did the Joey Harrington Experiment Work Out?
Now back to this regular season. The question at the beginning of this season was whether Harrington’s performance in Detroit could be improved on by a change of scenery for him, hence the JHE. However, if the JHE wasn’t more important than the overall success of the 2006 Miami Dolphins, then he should’ve been yanked when it was clear that his performance hadn’t improved to the point that he was winning games for us at a time when we had to win games to stay alive in the playoff race and salvage the season.
That point of the season in my opinion was just after the Jacksonville game, before the New England game at home. Before the Jacksonville game we were 5-6 on the season, coming off a four-game win streak, and still very much alive in the playoff race. We lost the game and fell to 5-7, which made our playoff hopes a huge longshot. We needed to win out and get help to make the playoffs at that point.
In my opinion, since we absolutely had to win every remaining game at that point, the only way Harrington should’ve been left in was if he had truly resurrected his career and his performance through the Jacksonville game had been consistent with helping the team win games. So let’s check that out and find out.
Through the Jacksonville game, Harrington played in eight games (half a season) and had thrown 25.1 passes per INT, with a YPA of 6.09. Obviously his number of pass attempts per INT was still significantly under the league average of 33.8, and in fact it was significantly worse than the 29.1 figure he had in Detroit. And although his YPA increased from his 5.6 figure in Detroit to 6.09, it was still significantly below the league average of 6.8.
This means Harrington hadn’t been helping us win games, yet he was left in at a time when we had to win every game to make the playoffs. In other words, at that point, for all intents and purposes, continuing the JHE became more important than the season.
But what about the four-game win streak you ask? Maybe he was turning a corner then and should’ve been left in? Again, I’m suggesting Harrington should’ve been removed after the Jacksonville game, and looking at his statistics during the five games including win streak and the Jacksonville game, he threw an INT every 30 pass attempts and had a YPA of 5.96.
Both of these numbers were still significantly lower than the league averages, so Harrington wasn’t turning any corners. Really when it comes right down to it, he had one stellar game in Detroit that probably kept him hanging on as the starter much longer than he should have.
The results of the JHE since the Jacksonville game? We know them all too well. Sure we beat New England convincingly, but then of course we went up to Buffalo and laid a total egg, ending any hopes for the playoffs.
Since after the Jacksonville game we were basically continuing the JHE at the expense of the season, let’s take a look at his stats after that game and see if we made good at least on the experiment and gave Harrington the extra playing time he may have needed to turn his career around.
Against New England and Buffalo combined, his number of pass attempts per INT was 23.5 and his YPA was 4.47, both still well below the league averages and very significantly so. His performance in Buffalo effectively ended the season, as he went 5 for 17 passing for 20 yards, no TDs, and 2 INTs. His number of pass attempts per INT for that game was 8.5, and his YPA was 1.18, both ridiculously low numbers. At that point the experiment crashed and burned, Harrington played only another half (the first half against the Jets), but by then the damage had been done and the season had ended in Buffalo.
The better approach in my opinion would’ve been to let Cleo Lemon start and get his feet wet against New England, a home game our defense was sure to be hyped for, and then start him the rest of the season. Perhaps the Buffalo and Jets games would’ve gone dfferently and we’d still be in the playoff race. As it was the “Joey Harrington Experiment†ruined them both. Against the Jets he threw no INTs in 15 attempts, but his YPA was a meager 2.8.
As important as the QB position is in the NFL, unfortunately the eulogy for the 2006 Miami Dolphins will be that we experimented with a QB at the expense of the season, and both the season and the QB failed.
However, as much as many of us wanted to give the guy a chance to see if he could resurrect his career and become a good QB for us, at no point in the season did I personally think the JHE was more important than the success of the Miami Dolphins’ 2006 season, and I think I’m probably not in the minority in saying that.
Ironically, what I’m going to try to show here is that JHE was for all intents and purposes (read: probably unintentionally) treated by the coaching staff as though it was more important than the success of the Miami Dolphins’ 2006 season.
The people here (like me) who thought Harrington should’ve been yanked even though we were winning games with him at QB for a stretch will find their point of view supported here. It’ll take a lot of words and numbers to do that effectively, so please be patient and hear me out.
In a nutshell, what I’m going to try to show with statistics is that, even though we were winning as a team for a stretch, overall, Joey Harrington still wasn’t doing the things from the QB position that help teams win games. In fact, what I hope to show is that he was predominantly doing things from the QB position that typically help teams lose games, which means he should’ve been shelved even in favor of a QB with no regular season experience like Cleo Lemon. What’s more, he was helping the team lose games at a time when we needed to win every game because our record was near .500 relatively late in the season and our playoff hopes hung in the balance.
What Do QBs Do to Help Teams Win and Lose Games?
First the statistics. I have the data for the following offensive and defensive variables for every team in the league from the five seasons between 2001 and 2005 (159 teams total; the Texans weren’t around until 2002):
Number of regular season wins
Points scored and allowed
Pass attempts
Pass completions
Completion percentage
Yards per attempt
Yards per completion
Pass attempts per INT
Number of rushes
Rushing yards
Yards per rush
I ran a general stepwise regression analysis (http://www.kellogg.northwestern.edu/...B/Modeling.htm) to determine which of the above variables are the strongest independent predictors of wins. I excluded points scored and allowed, which are obviously very strong predictors of wins and would therefore block other variables from entering the equation.
According to the regression analysis, the strongest independent predictors of wins between 2001 and 2005 were the following variables, in order of strength of predictability; their zero-order correlations with wins are in parentheses:
1) Number of rushes allowed (-.70)
2) Yards per pass attempt allowed (-.46)
3) Yards per pass attempt (.54)
4) Pass attempts per INT thrown (.45)
5) Pass attempts per INT caught (-.34)
6) Pass attempts allowed (.43)
7) Number of rushes (.57)
These were also the only variables that significantly predicted wins, and taken together they account for 74% of the variance in wins. That’s pretty huge in statistical terms, and it means that if you know the value of these variables you can predict wins with 74% accuracy (gamblers would love that wouldn’t they?).
So what we have here is an analysis of team variables almost all the way across the board (almost every area of offense and defense), with the finding that among those team variables, two that are strongly associated with the QB position, pass attempts per INT thrown and yards per attempt (YPA), are among the seven significant predictors of wins. In other words, if a QB throws more attempts per INT and has a good YPA, he’s going to significantly help his team win games.
Likewise, of the seven predictor variables that aren’t offensively oriented, two of those are the defensive counterpart of pass attempts per INT and YPA, meaning that four of the seven predictor variables center around pass attempts per INT and YPA. More specifically, the offensive and defensive counterparts of those two variables account for 22% of the 74% I noted above, so the QB is affecting quite a large percentage of the game with those two variables alone.
To reiterate, the upshot is that if a QB’s pass attempts per INT and YPA are good, he’s significantly helping his team win games, and if they’re bad, he’s significantly helping his team lose games.
Why The Joey Harrington Experiment?
Back to the main topic. Again, the JHE was conducted to see if he could resurrect his career and become a good QB for us. So let’s compare Harrington’s performance in these areas in Detroit to his performance for us this season. If he improved significantly in these areas (pass attempts per INT and YPA), obviously we can say he turned his career around and started performing well for us, at least in terms of the variables that strongly and significantly predict wins and losses.
In Detroit, Harrington played in 58 games and started 55 between the years of 2002 and 2005 (almost the same years used above by the way). Overall, Harrington’s number of pass attempts per INT was 29.1, and his YPA was 5.6.
His 5.6 YPA was significantly lower than the league average of 6.8 between the years of 2001 and 2005 at a significance level of <.0001, which means there’s greater than a 99.999% likelihood that the difference between his YPA and the league average wasn’t due to chance.
His number of pass attempts per INT (29.1) was significantly lower than the league average of 33.8, and that difference was significant at the same level. In other words, it’s clear Harrington was underperforming in these areas in comparison to the average NFL QB, and therefore he obviously wasn’t helping the Lions win games with his performance in these areas.
Again, however, equipped with only the knowledge we had about him at the beginning of this season, we could speculate that his performance could’ve been more about the environment in Detroit than anything else. And things were looking up when we heard things during camp and the preseason like, “Joey is happy in Miami; Joey’s having fun playing football again; Joey likes the system here better because it tells him exactly what to do, whereas in Detroit it was a mess.â€Â
How Did the Joey Harrington Experiment Work Out?
Now back to this regular season. The question at the beginning of this season was whether Harrington’s performance in Detroit could be improved on by a change of scenery for him, hence the JHE. However, if the JHE wasn’t more important than the overall success of the 2006 Miami Dolphins, then he should’ve been yanked when it was clear that his performance hadn’t improved to the point that he was winning games for us at a time when we had to win games to stay alive in the playoff race and salvage the season.
That point of the season in my opinion was just after the Jacksonville game, before the New England game at home. Before the Jacksonville game we were 5-6 on the season, coming off a four-game win streak, and still very much alive in the playoff race. We lost the game and fell to 5-7, which made our playoff hopes a huge longshot. We needed to win out and get help to make the playoffs at that point.
In my opinion, since we absolutely had to win every remaining game at that point, the only way Harrington should’ve been left in was if he had truly resurrected his career and his performance through the Jacksonville game had been consistent with helping the team win games. So let’s check that out and find out.
Through the Jacksonville game, Harrington played in eight games (half a season) and had thrown 25.1 passes per INT, with a YPA of 6.09. Obviously his number of pass attempts per INT was still significantly under the league average of 33.8, and in fact it was significantly worse than the 29.1 figure he had in Detroit. And although his YPA increased from his 5.6 figure in Detroit to 6.09, it was still significantly below the league average of 6.8.
This means Harrington hadn’t been helping us win games, yet he was left in at a time when we had to win every game to make the playoffs. In other words, at that point, for all intents and purposes, continuing the JHE became more important than the season.
But what about the four-game win streak you ask? Maybe he was turning a corner then and should’ve been left in? Again, I’m suggesting Harrington should’ve been removed after the Jacksonville game, and looking at his statistics during the five games including win streak and the Jacksonville game, he threw an INT every 30 pass attempts and had a YPA of 5.96.
Both of these numbers were still significantly lower than the league averages, so Harrington wasn’t turning any corners. Really when it comes right down to it, he had one stellar game in Detroit that probably kept him hanging on as the starter much longer than he should have.
The results of the JHE since the Jacksonville game? We know them all too well. Sure we beat New England convincingly, but then of course we went up to Buffalo and laid a total egg, ending any hopes for the playoffs.
Since after the Jacksonville game we were basically continuing the JHE at the expense of the season, let’s take a look at his stats after that game and see if we made good at least on the experiment and gave Harrington the extra playing time he may have needed to turn his career around.
Against New England and Buffalo combined, his number of pass attempts per INT was 23.5 and his YPA was 4.47, both still well below the league averages and very significantly so. His performance in Buffalo effectively ended the season, as he went 5 for 17 passing for 20 yards, no TDs, and 2 INTs. His number of pass attempts per INT for that game was 8.5, and his YPA was 1.18, both ridiculously low numbers. At that point the experiment crashed and burned, Harrington played only another half (the first half against the Jets), but by then the damage had been done and the season had ended in Buffalo.
The better approach in my opinion would’ve been to let Cleo Lemon start and get his feet wet against New England, a home game our defense was sure to be hyped for, and then start him the rest of the season. Perhaps the Buffalo and Jets games would’ve gone dfferently and we’d still be in the playoff race. As it was the “Joey Harrington Experiment†ruined them both. Against the Jets he threw no INTs in 15 attempts, but his YPA was a meager 2.8.
As important as the QB position is in the NFL, unfortunately the eulogy for the 2006 Miami Dolphins will be that we experimented with a QB at the expense of the season, and both the season and the QB failed.
