Derek Anderson, Replacement Level QB, and that's o.k. right now (analytics post)

[The Frankish Reich]

Question: how do mid-season veteran QBs signed off the street perform when forced into action?

Short Answer (long explanation below, so I'll cut to the chase): "Not as bad as you'd think." In fact, unless you've got an established very good-to-great QB as your regular starter (see Deshaun Watson and Aaron Rodgers examples below), there might not be any drop-off at all. And if you've got a really bad (or very raw rookie) QB as your starter, the signed-off-the-street replacement guy will probably be an upgrade. In short, my money's on Derek Anderson and the Bills this weekend.

 

Long Answer: I'm talking here about classic replacement level QBs here (see the explanation below if you're wondering what I'm talking about*) - the pool of "freely available talent" veteran that can be signed for roughly the minimum salary.  Think Thad Lewis, or Jordan Palmer, or (as one analyst put it) "anyone named Billy Joe." Or Kellen. Or in our case, Derek Anderson.

 

First, the baseline - where do we stand pre-Anderson? 

Football Outsiders ranks QBs by value over replacement level. The results through Game 6 of 2018 shouldn't surprise anyone:

 

https://www.footballoutsiders.com/stats/qb

 

Yes, that's Josh Allen trailing the pack at -504 (as in NEGATIVE 504 yards) "Defense Adjusted Yards Above Replacement." (The terms are defined at the link) Of course, Peterman is worse on a pro rata basis: -254 in his very limited action. Open the link. Please. The chasm between Allen and the rest of the QBs is pretty astounding (although Tyrod Taylor gives him a run for his money on a pro rata basis). By more traditional stats, Allen has a 61.8 QB rating. He really has been that bad.

 

So ... sign a Derek Anderson off the street and what can you expect? Well, clearly better than the horrifically sub-replacement level we've been getting. I can't really find a way to search for this on pro football reference or any other site, so please help me here. I'll add the QB rating for the team's primary QB that season so we can compare how the "replacement player" did compared to the "regular QB." This is what I've got so far:

 

- Thad Lewis (two go rounds for him as an "off the street" emergency signing):

        1. Browns 2012. 1 start, 0-1. 83.3 QB rating. Main QB Brandon Weeden put up a 72.6 in 15 games.

        2. Bills 2013. 2-3 record. 81.0 rating (hey, why didn't we call him in again?) Main QB EJ put up an 81.0 rating.

 

- TJ Yates (he seems to specialize in this role; I won't count his playoff games because he was on the roster the whole season that year, not "signed off the street")

        1. Texans 2017: signed off the street, put up a 65.8 QB rating in 4 games.  0-3 record in his starts. Main QBs Watson and Savage put up 103.0 and 71.4 QB ratings. Awful ... almost Allen-Awful. Kind of the floor for what you can expect from any "Professional NFL QB" walking around in street clothes?

        2.  Texans 2015: signed off the street in October, put up an 80.3 QB rating in 4 games (2 starts). 2-0 record in his starts. Main QB Hoyer put up a 91.4 rating.

 

- Charlie Whitehurst, Browns 2016. Signed off the street before Game 3. Didn't start, but went 14-24, 1 TD/1 INT for a 78.8 QB rating. Main QB Cody Kessler put up a 92.3 rating; RG3 a 72.5.

 

- Matt Flynn, Packers 2013. Signed off the street in November. 2-2 record, 86.1 QB rating. Aaron Rodgers put up a 104.9 QB rating; Scott Tolzien a 66.8

 

- Kellen Clemens, Rams, 2011. Signed off the street by the Texans in November, then released, then claimed by the Rams in December. 0-3, 73.8 rating. Sam Bradford was 0-9 with a 70.5 QB rating.

 

Yes, it really seems to be the case that you can sign any old veteran off the street and get a better performance than you're getting out of Josh Allen (or, for that matter, Josh Rosen) or many other "developmental" QBs.

 

-------------------------------------------------

 

*The concept of the "replacement level player" started with baseball stats (where it is now absolutely essential in any GM's roster building). It's more difficult to transfer to football, and most analysts set the bar way too high. For example, Jay Cutler last year, because he was unsigned at the start of the season, has been called "replacement level." But Jay didn't play until someone offered him $5 million guaranteed and a chance to earn $10 million. That's not what I'm talking about. Likewise Kaepernick. I'm also not talking about the rookie/young QB being groomed as a starter or backup (think Jacoby Brissett) - these are guys that have significant trade value. Nor am I talking about the practice squad rookie like Jeff Tuel. This is about the Derek Andersons and EJ Manuels of 2018, not those guys back when teams fought over them.

Edited October 19, 2018 by The Frankish Reich

[CaptnCoke11]

?

[KingRex]

Thanks for the details, but I thnk the key analysis for this game is less likely to be the analytics of what are the combined stats of the middlin to  poor QBs in the #2- #3 roles for the offense but instead an analysis of the D of the opponents.

 

The Indy comes into this game with zero film on Anderson's strengths and weaknesses and zero record on Daboli's play call tendencies with Anderson at the helm.

 

My GUESS is the Bills come in with an O playbook limited to the plays the existing O has run well (next to nothing actually as Allen racks up about 100 yds a game).  They add onto his whitebread base a few misdirection plays they have practiced with Anderson and anything Anderson/Benjamin (who have played together some as bak-ups for NC feel that they have a chemistry for.

 

If I am Indy  I:

 

1. Assign a spy to Shady as he is the only Bills O player who scares me

2. Emphasis aggressive and changing blitzes on the newly installed and thus limited knowledge of second and third reads and facing pressure from Anderson.

3.  I do not fear a shootout if it comes down to whether Luck or Anderson performs.

 

What does analytics say re Indy?

 

 

[fansince88]

We shall see on Sunday. If he moves the ball and throws as many TD's as Nate does INT's we should be fine.

[BigBuff423]

To the OP, I appreciate your time and the effort it took to put this thoughtful post together. However, I fear your diligent efforts will be lost on this board as many will simply dismiss your analytic information for Anderson apologetics. 

 

That said, while the analytics provide some empirical evidence for a scintilla of hope for Sunday, to me it would be somewhat obvious. Mainly, because as you pointed out, Allen has just been so dismal except for the Minny game where he pretty brilliant overall, but still lacked pocket presence and ran a bit too much. I think Anderson will play relatively well - meaning, not great but at about a slightly below average NFL level - and IF the Bills D shows up the way it has since the second half of the Chargers game, I think the Bills have a good chance at winning. Bills D stays on the plane, and I don't think there's really much of any viable chance at getting the W. 

[Brianmoorman4jesus]

People really don’t appreciate how much better these QBs get once they are older and have seen it all. Most of the best QBs in the league are very experienced veterans. I am excited for Anderson. The best passing games we have had post Kelly were Orton and Fitz. Both of those guys had been around. I am tired of watching severely limited offenses and half game plan football. Let’s go out and play the game that the rest of the league always plays.

[PlayoffsPlease]

3 minutes ago, Brianmoorman4jesus said:

People really don’t appreciate how much better these QBs get once they are older and have seen it all. Most of the best QBs in the league are very experienced veterans. I am excited for Anderson. The best passing games we have had post Kelly were Orton and Fitz. Both of those guys had been around. I am tired of watching severely limited offenses and half game plan football. Let’s go out and play the game that the rest of the league always plays.

This is the epitome of being a Bill's fan, wistfully hoping for a return to the glory days of Kyle Orton and Ryan Fitzpatrick.   The only explanation is that us Bill's fans have been place on a purgatory island the island in Lost, and none of this is really happening. 

[The Frankish Reich]

1 hour ago, BigBuff423 said:

To the OP, I appreciate your time and the effort it took to put this thoughtful post together. However, I fear your diligent efforts will be lost on this board as many will simply dismiss your analytic information for Anderson apologetics. 

 

That said, while the analytics provide some empirical evidence for a scintilla of hope for Sunday, to me it would be somewhat obvious. Mainly, because as you pointed out, Allen has just been so dismal except for the Minny game where he pretty brilliant overall, but still lacked pocket presence and ran a bit too much. I think Anderson will play relatively well - meaning, not great but at about a slightly below average NFL level - and IF the Bills D shows up the way it has since the second half of the Chargers game, I think the Bills have a good chance at winning. Bills D stays on the plane, and I don't think there's really much of any viable chance at getting the W. 

Thanks. I appreciate the kind words. I think some people here will get my point. It's not that Josh Allen shouldn't start; he probably needs to play (when healthy) to develop. Rather, it's that we don't realize how high the veteran QB "floor" is relative to a young/raw QB like Allen. In the 7 "veteran signed off the street" games I looked at in my original post, it's fair to say that the replacement vets were better than the guys they replaced in 3 games; worse in 1 game; roughly the same in 3. (I'm saying Flynn replaced Tolzien, not Rodgers; Yates replaced Savage, not Watson.)  None of the replacement level QB games I looked at involved a QB who'd been with the team for an extended time "learning the system" before getting thrown in to real games. (You could argue that Flynn and maybe Yates knew the systems they were put in from previous stints with their teams, but that was a while ago.)  So that's what you can expect: in the Bills current situation, better QB performance from the signed off the street free agent veteran than what we saw from Allen.

 

48 minutes ago, Brianmoorman4jesus said:

People really don’t appreciate how much better these QBs get once they are older and have seen it all. Most of the best QBs in the league are very experienced veterans. I am excited for Anderson. The best passing games we have had post Kelly were Orton and Fitz. Both of those guys had been around. I am tired of watching severely limited offenses and half game plan football. Let’s go out and play the game that the rest of the league always plays.

Exactly. For other reasons, I kind of hope Derek Anderson doesn't get more than 2 or 3 starts, but I expect him to be better than Allen (and Peterman!) nonetheless.

[BuffaloBillies]

I actually liked Thad. Thought he did a nice job in very difficult situations.

Edited October 19, 2018 by BuffaloBillies

[IgotBILLStopay]

The replacement QB relative win loss record is a function of the opposition as well as over the year improvements  in D and ST even if offense is the same. As OP points out we have not had stellar (or even mediocre) QB play and yet we are 2-4. Our D has improve tremendously the past 4.5 games. 

 

I am more optimistic than most even about this season since the Bills schedule does get easier. Additionally our D has started playing lights out. FWIW, I even thought Allen was turning a corner with his two strikes to KB (one called back). The offense is what it is - the system can’t overcome lack of talent - but hoping that the non Peterman offense can play mistake free ball.  More than the Offense, my concern is with ST. In that light, players specifically working on that is a good sign. 

 

To sum up, I am hopeful we can get a few wins thanks to our defense so long as we clean up ST and play mistake free conservative offense. It won’t look pretty but can be effective considering the remaining opposition.

[Coach55]

This is a great post.  With average QB play this season we are at minimum 3-3 and probably 4-2.  There are plenty of 30-something journeymen that could've been slotted into this role while Allen is developed.  How Nate Peterman is still on the team is beyond me.  I am really looking forward to seeing what Anderson could do this weekend.  With our defense, all you need is a QB that isn't going to screw it up, they don't have to be great.  

[Rebel101]

I believe we will be better. As long as Anderson has a handle of the offense and get gain chemistry with the receivers. Our defense is playing at an extremely high level and they’ve put us in a chance to win every game since the chargers. I believe a veteran QB takes advantage of of defense and we actually pull out more wins then previously thought was possible. But the real question is if Anderson does start winning us games do you pull him when Allen gets healthy?

[Dopey]

1 hour ago, The Frankish Reich said:

Thanks. I appreciate the kind words. I think some people here will get my point. It's not that Josh Allen shouldn't start; he probably needs to play (when healthy) to develop. Rather, it's that we don't realize how high the veteran QB "floor" is relative to a young/raw QB like Allen. In the 7 "veteran signed off the street" games I looked at in my original post, it's fair to say that the replacement vets were better than the guys they replaced in 3 games; worse in 1 game; roughly the same in 3. (I'm saying Flynn replaced Tolzien, not Rodgers; Yates replaced Savage, not Watson.)  None of the replacement level QB games I looked at involved a QB who'd been with the team for an extended time "learning the system" before getting thrown in to real games. (You could argue that Flynn and maybe Yates knew the systems they were put in from previous stints with their teams, but that was a while ago.)  So that's what you can expect: in the Bills current situation, better QB performance from the signed off the street free agent veteran than what we saw from Allen.

 

Exactly. For other reasons, I kind of hope Derek Anderson doesn't get more than 2 or 3 starts, but I expect him to be better than Allen (and Peterman!) nonetheless.

It's a long shot, but if we win 2 out of those 3 starts and are in the playoff hunt, do you put Allen back in?

[Mickey]

Someone needs to explain these types of analytics to me. I don't get it. Past outcomes do not influence future outcomes unless there is a causal link between a particular past outcome and a future outcome. Flip a coin 10 times in a row and it comes up heads, the odds of the eleventh flip being heads is still 50%. However, if you are playing four card burn and are hoping for aces, your chance of getting one diminishes with every ace dealt to another player. Independent trials vs. dependent trials.

 

Aren't these numbers just a whole pile of coin flips banded together to look like a deck of cards? The fact that some other QB's on some other teams in some other games running some other offense whose plays are being called by some other coaches added up results in an average performance of x,y or z seems to me to be information of very little value. We aren't starting an average QB playing an average defense behind an average offensive line in average weather conditions with average coaching and an average amount of injured players effecting the game. We are starting Derek Anderson and much information about him and his abilities are known. That information is, to me, more valuable than how a bunch of other guys in a bunch of other games, on average, performed. Our offense, its strengths and its weaknesses are well known to us and that seems way more informative than the average of what other offenses have done at other times in other circumstances.

 

I guess my questions is whether or not Derek Anderson's performance on Sunday is a dependent trial or an independent trial? Or perhaps the DVOA of my math skills indicate that this is all just too complicated for a senior citizen who played without a helmet for way too long?

 

[dave mcbride]

6 minutes ago, Mickey said:

Someone needs to explain these types of analytics to me. I don't get it. Past outcomes do not influence future outcomes unless there is a causal link between a particular past outcome and a future outcome. Flip a coin 10 times in a row and it comes up heads, the odds of the eleventh flip being heads is still 50%. However, if you are playing four card burn and are hoping for aces, your chance of getting one diminishes with every ace dealt to another player. Independent trials vs. dependent trials.

 

Aren't these numbers just a whole pile of coin flips banded together to look like a deck of cards? The fact that some other QB's on some other teams in some other games running some other offense whose plays are being called by some other coaches added up results in an average performance of x,y or z seems to me to be information of very little value. We aren't starting an average QB playing an average defense behind an average offensive line in average weather conditions with average coaching and an average amount of injured players effecting the game. We are starting Derek Anderson and much information about him and his abilities are known. That information is, to me, more valuable than how a bunch of other guys in a bunch of other games, on average, performed. Our offense, its strengths and its weaknesses are well known to us and that seems way more informative than the average of what other offenses have done at other times in other circumstances.

 

I guess my questions is whether or not Derek Anderson's performance on Sunday is a dependent trial or an independent trial? Or perhaps the DVOA of my math skills indicate that this is all just too complicated for a senior citizen who played without a helmet for way too long?

 

Disagree with this. Tom Brady's past outcomes are a friggin' great predictor of his future outcomes because ... he's really good. The point is that past performance is often quite indicative of future performance because it's a reflection of talent.  Granted, we're talking about this issue in the context of QB dumpster diving, so it's a little hard to see!

[mannc]

Josh McCown might be the best example of this.  He actually stepped in and played very well for the Jets last year.  He would have been a great guy to have backing up JA this year.

[The Frankish Reich]

12 minutes ago, Mickey said:

Someone needs to explain these types of analytics to me. I don't get it. Past outcomes do not influence future outcomes unless there is a causal link between a particular past outcome and a future outcome. Flip a coin 10 times in a row and it comes up heads, the odds of the eleventh flip being heads is still 50%. However, if you are playing four card burn and are hoping for aces, your chance of getting one diminishes with every ace dealt to another player. Independent trials vs. dependent trials.

 

Aren't these numbers just a whole pile of coin flips banded together to look like a deck of cards? The fact that some other QB's on some other teams in some other games running some other offense whose plays are being called by some other coaches added up results in an average performance of x,y or z seems to me to be information of very little value. We aren't starting an average QB playing an average defense behind an average offensive line in average weather conditions with average coaching and an average amount of injured players effecting the game. We are starting Derek Anderson and much information about him and his abilities are known. That information is, to me, more valuable than how a bunch of other guys in a bunch of other games, on average, performed. Our offense, its strengths and its weaknesses are well known to us and that seems way more informative than the average of what other offenses have done at other times in other circumstances.

 

I guess my questions is whether or not Derek Anderson's performance on Sunday is a dependent trial or an independent trial? Or perhaps the DVOA of my math skills indicate that this is all just too complicated for a senior citizen who played without a helmet for way too long?

 

I get your point. Let me try to explain by using a baseball example (the sport that led the analytics movement). For years, everyone believed that a big reason good pitchers were good was because they had the ability to induce weak contact, and weak contact meant balls in play that were easily converted into outs. Dribblers vs. line drives; pop ups vs.  deep fly balls. But one ingenious study asked this question: we know the worst pitchers in MLB are position players sent out to throw an inning or two in blowout games. Hell, they’re not even professional pitchers, much less major league pitchers. And what did he find? He found that even these worst possible pitchers didn’t do much worse on the Batting Average on Balls in Play (BABIP) they surrendered to opposing hitters as compared to “real” pitchers. (They did do much worse on what are called the “three true outcomes” - strikouts, walks, home runs given up - that is, those things that have nothing to do with balls out in play.). So we learned something. In my example, I have a tiny, self-selected sample (based on my own recent memories) of how veteran QBs signed off the street did when forced into game action. I’m using this as a proxy to determine what I think is an interesting hypothetical question: Let’s say a completely otherwise average team with a completely average strength of schedule doesn’t bother at all to get involved in the off-season QB derby (draft/free agency). Instead, it rolls into Week 4 of preseason and signs the best unsigned 5 year veteran QB it can find prowling the streets and installs him as the starter. My hypothesis is that that team/QB will probably finish at something like 6-10. Maybe 5-11. That can then be defined as how an otherwise average team (the Bills in 2017/18?) would fare with a replacement level QB. And my admittedly sketchy sample suggests (doesn’t prove ... we’d need to dig a whole lot deeper for that) that maybe I’m right?  That maybe that the floor for any otherwise ordinary NFL team signing a Thad Lewis or Derek Anderson just prior to the season for the minimum salary will probably still win 5 or 6 games, perhaps more than they’d win with a Josh Allen, and definitely more than they’d win with a Nate Peterman?  That’s why I’m suggesting the replacement level guy will probably be an upgrade for us in the short term given the clearly sub-replacement level performance we’ve gotten so far ....

... hope that explains it a bit better. Feel free to find examples that refute my hypothesis!

[hondo in seattle]

Good post by The Frankish Reich

 

But let me take a simpler approach that looks at our QBs specifically.   Here are their career  passer ratings:

 

NP:  29.9

JA:  61.8

DA:  71.1

 

Derek Anderson's knowledge of the playbook will be limited.  But we've heard that Daboll was using a simplified version of his EP offense with Josh and Nate.  Whatever DA runs won't be dumbed down much from what NP and JA were running.  

 

And DA has run the EP in the past.  He should be able to get up to speed quickly.  

 

Anderson's 71.1 passer rating isn't great - but it's better than the crappy QB play we've been getting.  

[ctk232]

6 hours ago, Coach55 said:

This is a great post.  With average QB play this season we are at minimum 3-3 and probably 4-2.  There are plenty of 30-something journeymen that could've been slotted into this role while Allen is developed.  How Nate Peterman is still on the team is beyond me.  I am really looking forward to seeing what Anderson could do this weekend.  With our defense, all you need is a QB that isn't going to screw it up, they don't have to be great.  

You want an Alex Smith then, not a Derek Anderson. Again, hopeful with the rest of us, but I can't see this going well outside of the fact that Indy has literally no tape on Anderson. 

 

41 minutes ago, hondo in seattle said:

Good post by The Frankish Reich

 

But let me take a simpler approach that looks at our QBs specifically.   Here are their career  passer ratings:

 

NP:  29.9

JA:  61.8

DA:  71.1

 

Derek Anderson's knowledge of the playbook will be limited.  But we've heard that Daboll was using a simplified version of his EP offense with Josh and Nate.  Whatever DA runs won't be dumbed down much from what NP and JA were running.  

 

And DA has run the EP in the past.  He should be able to get up to speed quickly.  

 

Anderson's 71.1 passer rating isn't great - but it's better than the crappy QB play we've been getting.  

It's skewed by his one solid year - look at his stats year to year for better trends and historical growth/decline. Not to mention, while widely used, Passer Rating has it's own statistical shortcomings.

Edited October 19, 2018 by ctk232

[Big Turk]

I'm scared Anderson is going to hurt our draft positions by winning games we would have lost.