Is It Tebow Time Yet?

I am a statistics guy. I believe in numbers and data. I believe in the logical and objective application of what can be measured. Look, I don’t believe statistics are perfect or that they capture the whole picture, but I have a lot more faith in numbers than in “gut feelings” or “intangibles” or the idea that there might be some kind of divine intervention in football games. This is why I thought I would never utter, let alone write (in a public forum), these words:

The Jets should consider starting Tim Tebow.

Ok, ok. Let’s all take a deep breath. There’s a lot of hedging to be done, and it’s not exactly clear cut, but I’ve run the numbers and, well, they were surprising. In the interest of full disclosure, part of my inspiration to write this article was to show that Tebow was not some inexplicable winner, but rather that his wins could be explained by the high support numbers he received in 2011. And, to some extent, the numbers bore out my initial hypothesis. But it was not quite as dramatic as I would have first guessed. I was surprised to learn that, adjusting for QB Support, Tebow was about “average” at winning games. That’s decidedly better than where his QBR would rank him (31^st out of 34 ranked QBs in 2011).  

If that was not enough, there was something else for which I was not quite prepared: Mark Sanchez is actually really really bad at helping his team win. Ok, maybe I should have been prepared for that. Still, it was worse than I expected.

In examining this question, I looked at all of Tim Tebow’s starts from 2011, including the playoffs, and I looked at all of Sanchez’ starts from the 2011 and 2012 season. Here’s what I found:

SANCHEZ AND TEBOW: BOTH BENEFITING FROM HIGH SUPPORT LEVELS

So far in the 2012 season this is how all NFL QBs performed across different support levels:

QB SUPPORT	Lower than 1.5	1.5 to 2.99	3.0 to 3.99	4 to 4.99	5+
WINS	4	29	26	21	40
LOSSES	48	39	17	11	1

There are two things I want point out here. First, there is the obvious point that QB Support, especially at the extremes, makes a big difference. Second, I want point out a breakdown of how support levels are distributed. This is only about half of one season, but it suggests that, on average, a QB gets a QB Support level of 4 or above around 30 percent of the time, and above 5 around 17 percent of the time. Meanwhile, QBs thus far in the 2012 have had QB Support levels below 3 about half the time.

Looking at all the games Mark Sanchez played from the 2011 and 2012 seasons we get a distribution for him that looks like this:

MARK SANCHEZ

QB SUPPORT	Lower than 1.5	1.5 to 2.99	3.0 to 3.99	4 to 4.99	5+
WINS	0	0	4	2	5
LOSSES	4	3	4	1	1

There are 24 games here, and we can see that Sanchez enjoyed high QB support slightly more often than the league average. His QB Support levels were above 4 in 37 percent of games (league average = 31 percent). Furthermore, 63 percent of Sanchez’ wins came from those games. Sanchez had to play with QB Support levels less than 3 in only about 30 percent of his games in 2011 and 2012 (league average = 50 percent).

How about Tim Tebow? Well Tebow enjoyed high QB Support levels even more often than Mark Sanchez. Here’s his breakdown.

TIM TEBOW

QB SUPPORT	Lower than 1.5	1.5 to 2.99	3.0 to 3.99	4 to 4.99	5+
WINS	0	1	0	5	2
LOSSES	2	2	0	0	1

Although games with QB Support levels above four make up only 31 percent of all games played in 2012, they made up 61 percent of Tebow’s 2011 starts. Not only that, they made up 87 percent (7 of 8) of the games he won.

WHO DOES MORE WITH MORE?

Both Sanchez and Tebow receive high support on a somewhat consistent basis. And neither have stellar records when their support is below 4. Still, I wanted to get a sense of how effectively the two QBs played across their support levels, even if they were significantly higher than league averages.

First, I looked at the average winning percentage for all QBs across support levels during the 2012 season. I used that winning percentage and the number of games Sanchez and Tebow played at different support levels to calculate how many wins we should have expected from each of them. Here are the results:

SANCHEZ

QB SUPPORT	Games Played	Expected Wins	Actual Wins
LOWER THAN 1. 5	4	0.308	0
1.5 to 2.99	3	1.279	0
3 to 3.99	8	4.837	4
4 to 4.99	3	1.969	2
5+	6	5.854	5
TOTAL	24	14.247	11

TEBOW

QB SUPPORT	Games Played	Expected Wins	Actual Wins
LOWER THAN 1. 5	2	0.154	0
1.5 to 2.99	3	1.279	1
3 to 3.99	0	0.000	0
4 to 4.99	5	3.281	5
5+	3	2.927	2
TOTAL	13	7.641	8

Sanchez won 11 games, a number that is 3.247 less than what we should expect from “an average NFL QB” given the support levels the Jets provided. Obviously, a QB can’t rack up fractional wins, but it serves as our reference point. Sanchez played 7 games in 2011 and 2012 where QB Support was below 3, and he lost each and every one of them. He simply is not going to carry the Jets to a win on any type of consistent basis. Sanchez even lost one a game with a support level above 5 (a 2011 game against the Ravens), something that’s happened only ONE time in 2012. Then again, so did Tebow (the 2011 season finale against the Chiefs). Regardless, Sanchez’ QB Support levels suggest that, with an average QB, the Jets should be 14-10 in their last 24 games. Instead, they’re 11-13.

Tebow may have blown a game despite receiving 5+ support, but he also won every game where he received a QB Support level between 4 and 4.99. Tebow’s QB Support levels suggest that, during his 13 starts, the Denver Broncos should have been 7-6 or 8-5 with "an average NFL QB." Tebow was able to help them get to the higher of those two. It is not exactly phenomenal, but it suggests he performed within the range of average. Now, a lot can be made of that single game that Tebow won with a QB support level between 1.5 and 2.99. For those playing at home, if you guessed that it was Tebow’s playoff win against the Steelers (His best game as rated by QBR as well), you are right. That was a game that clearly could have gone either way, and it does make up Tebow's only win with QB Support below 4. However, even if you pretend that the Broncos lost that game, Tebow’s “Expected Wins” are within a point of his actual wins (unlike Sanchez’ -3.25 figure).  

QBR, MARGIN OF VICTORY, AND WHY WE NEED TO BE CAUTIOUS 

QBR suggests that, although Sanchez is pretty poor compared to the league average (29^th out of 33 ranked QBs in 2011), Tebow is even worse. So what gives? There are two things that I suspect are going on, but they point in opposite directions.

First, QBR measures how well a QB performs over the course of the game on average and over all the games combined. When a QB is getting high QB Support levels (meaning, in many cases, a low scoring game), it is likely more valuable to have the QB play great for one quarter and terrible for the other three opposed to mediocre for the whole game. Two good drives, as Tebow demonstrated time and again, is sometimes all it takes. This is why Tebow’s fourth quarter QBR, which is above average, is important to consider. Tebow actually IS better in the fourth quarter; he actually has played better in the clutch. If your defense holds your opponents to 13 or 10 points, one solid quarter of play may get you all the points you need. In many games, it was Tebow's inconsistency that helped him. I do not, however, think that is the full explanation.

I also think Tebow DID get lucky. Consider this: four of Tebow’s starts went to overtime, and he won every single one of those. Sure, you have to give him some credit for coming through when it counts, but given that QBs sometimes lose in overtime without ever touching the ball, you can’t expect Tebow to remain undefeated in overtime forever. The fact that half of his wins came in such a manner suggests that his "average" winning ability could be a statistical fluke. That, and the small sample with which we are working,  means we have to be careful about reading too much into his record, even after QB Support is taken into account. If we assign those overtime games a margin of victory of 0 (because they were tied at the end of regulation), Tebow won eight games by an average of 3.5 points. He lost 5 games by an average of 23.8 points. Compare that to Sanchez who (using the same rules for overtime games) won 11 games by an average of 14.9 points and lost 14 by an average of 15.5 points.

So, is it time for the Jets to bring in Tim Tebow? Maybe. If they do, they shouldn't expect much more than “average” from him in terms of his ability to win, and that’s IF he continues to win all the close ones. It’s very possible that he would be fairly comparable to Sanchez moving forward. But that's the best argument for Tebow: the status quo for Jets QB play is already so close to the bottom. Why not try something new? 

Regardless of whether the Jets start Tebow, keep him on the bench, or cut him, it’s definitely past time for the Jets to ditch Sanchez. His 0-7 record for the last season and a half when support is below 3 is not encouraging. It suggests that he has been simply incapable of putting the team on his back and carrying them to victory. He is no franchise QB. What’s worse though is that he is decidedly below average in games where support is high. That suggests that he will be a primary reason for a number of losses for an otherwise decent team.

Given how significant Sanchez' struggles have been across all QB Support levels, I offer these (famous) last words on Tim Tebow: How bad can he be?

UPDATE 01/2013
Since writing this article, I have been able to more accurately rank a QB's performance across support levels using TAVA. What was interesting, however, is that when I calculated Sanchez's TAVA for the 2012 season and Tebow's TAVA for the 2011 season, it gave me similar results: Sanchez is still terrible, and Tebow may indeed give the Jets a better chance at winning, but Tebow is likely not the long-term answer to the Jets QB problems. 

Sanchez ranked 35th out of 38 ranked QBs with a TAVA of -2.208. Tebow would have ranked 30th (between Blaine Gabbert and Nick Foles) with a TAVA of -1.055. Although Tebow would have ranked just about average in Wins Added (15th compared to Sanchez's 32nd), his weak ranking in Points Added (36th compared to Sanchez's 35th) brings him closer to the bottom.