"Dissecting QB Value In Fantasy Football -- The Zero QB Theorem"

[darin3]

Article is now up on the main page: http://www.thehuddle...sy-football.php

 

 

 

Some evening reading for sure.

[delusions of grandeur]

The easiest way that I've always explained why it's good to wait on QB, is the opportunity cost of drafting one early. That means that by drafting a QB early, you're sacrificing the opportunity to gain strength at other positions, and will have to rely on hitting on fliers later and on the WW to compensate (unless of course, as BC always mentions, that the scoring is just so skewed towards QBs that the top ones just score exponentially more, and in that case the other positions barely matter, but most leagues are not skewed that way, as you can tell by looking at scoring differential between the best and worst starter per position).

 

I'm glad to see actual documentation that the difference you gain by having a top QB is usually more than offset by the strength you lose elsewhere. For example, you can almost always find a plenty good QB in the mid to later rounds, even double up on them if you wait too long, but unless you're just the king of sleepers, it's not nearly as easy to predict at the other positions, and your team will tend to suffer as a result.

 

Good stuff Keg

[electricrelish]

Great write-up, keg! I loved your analysis. I'm glad most readers are scoffing so I can continue to draft QB's late.

 

I am curious as to how it will influence you in drafting. I know waiting has lots to do with who's on the board, but do you see yourself waiting until Elil Manning is available or do you go later than that and settle for a Flacco?

[delusions of grandeur]

Great write-up, keg! I loved your analysis. I'm glad most readers are scoffing so I can continue to draft QB's late.

 

I am curious as to how it will influence you in drafting. I know waiting has lots to do with who's on the board, but do you see yourself waiting until Elil Manning is available or do you go later than that and settle for a Flacco?

 

 

Not answering for Keg (quite the opposite, would love to see what he thinks) but I always go QBBC if I wait that late, with either 2 safe picks, or a combo of a safe/high-upside, maybe even 2 high upside if you're feeling really ballsy.

[keggerz]

Great write-up, keg! I loved your analysis. I'm glad most readers are scoffing so I can continue to draft QB's late.

 

I am curious as to how it will influence you in drafting. I know waiting has lots to do with who's on the board, but do you see yourself waiting until Elil Manning is available or do you go later than that and settle for a Flacco?

 

It's all relative but in the mocks I've been doing Eli is the highest ranked QB I've taken...Going forward in the rest of my drafts and mocks I feel safe saying I won't be taking a QB until 16-18 at a minimum are off the board.

[keggerz]

Before it happens I just want to make sure that no one turns this thread into a pile of worthless crap. If you agree that's fine...if you don't agree that's fine...if someone doesn't agree with you that's fine...but do not turn this into something that can't continue to be a positive experience for forum users.

[stethant]

Many thanks for the article - exactly the type of fresh-look analysis we appreciate here.

 

I would agree with your concern about putting a "metric in context." Did you consider performing the same zero analysis with the RB1 or WR1 position? If zeroing out those positions resulted in a change in the win percentages that were statistically significant, that would strongly buttress your contention that the QB position is unique in this regard.

[delusions of grandeur]

Many thanks for the article - exactly the type of fresh-look analysis we appreciate here.

 

I would agree with your concern about putting a "metric in context." Did you consider performing the same zero analysis with the RB1 or WR1 position? If zeroing out those positions resulted in a change in the win percentages that were statistically significant, that would strongly buttress your contention that the QB position is unique in this regard.

 

 

(again, not trying to speak for keggers), those could definitely be some good followups to consider, but I think one of his primary purposes here was to dispell this notion in recent years that you need an elite QB to win.

 

Perhaps you might find something similar if you did RB or WR, but then I don't think it's as easy to find a fairly consistent player with high-scoring potential at those positions later, like it is to find a QB that isn't ever going to actually get you near zero, like even your top WRs can get shut out on an off week. He's merely quantified what some of us already saw, that QBs scoring more points doesn't necessarily give them more value in most leagues.

 

Though I'll of course defer to him to elaborate further.

[delusions of grandeur]

Other than the "funny" little post about golden wisdom, my contribution to this thread has been to discuss the content of the article itself--and provide some counterpoints to the interpretation of statistical data reported therein. I would suggest that everyone else do the same, since having 20 retorts about what an idiot I am after every post I make really does more to derail a thread than anything I have said.

 

 

First off, if you really cared to discuss the article objectively, you wouldn't declare it flawed and meaningless, and take the opportunity to stroke your own ego.

 

But if you'd rather I dispute your flawed anazlysis, fine:

 

Your fourth table shows “the percentage of games that a team still won once their points at QB were zeroed out.” Since these percentages hover around 60%, and the percentage of all games won by all teams with any or all of their players zeroed out must, by definition, equal 50% (each game will have exactly one winner and one loser), I must conclude that the key word here is “still”--the percentage of games that a team STILL won after zeroing out their QB.

 

 

"The percentage of all games won by all teams with any or all of their players zeroed out, must by definition, equal 50%" is entirely irrelevant to the success of lack thereof of individual teams measured with the QB removed. You cannot compare ALL with individual results. It's meaningless. What is relevant is that ~60% of the time, a lack of QB did not make a difference in their winning the game.

 

So you say the 61% win rate equates to 7.9 games over a 13 game season, but it doesn’t. If the team in question would have gone 10-3, your data shows that they would have won 61% of their 10 wins even with no QB--that’s only 6 games, for a record of 6-7. The only way that the team in question would win 8 games over the course of the season is if they WOULD have gone undefeated (13-0) with their QB in the lineup. I would say that zeroing your QB makes quite a bit of difference if it means that you go from 10-3 to 6-7.

 

A win rate is a win rate. If a team went 10-3, then their win percentage is 76.9, but on average, their win rate was 61% without a QB. You can't just take a random example and then take a win percentage out of that. It doesn't make any sense whatsoever to take a win % out of a win %. No idea where you're going with that.

 

This, of course, affects your next table (and your conclusion regarding it)--the average wins by top 3 ADP QBs. You show that the top-3 ADP QBs average 6.6 wins on a season--but then you compare that to the 7.9 games you erroneously claim that a team can win with a zero QB. The comparison should be to the 61% of those wins that those teams would have recorded without their top-3 ADP QB, which is 61% of 6.6 wins--or 4.03 wins. So you’d be costing yourself two and a half wins, on average, if you went with no QB instead of a top-3 ADP QB.

 

What? It doesn't affect anything, because this is a different measure. 61% related to whether when ANY QB was zeroed out for a team, what was their win percentage. This measure is about the winning percentage of ONLY the top 3 ADP QBs, and how many games they personally put over the top. You're comparing apples and oranges in a feeble attempt to make a point. The two are seperate analyses, and thus not affected by one another.

 

 

The fact that the playoff records of top-3 ADP QBs is about 50% isn’t surprising, given that you’re talking about the best teams in the league now--all of which either drafted or acquired a solid QB (and a solid supporting cast around them) during the season. I suspect that you could say the same for the top players at RB, WR and TE as well. In fact, I suspect that you could find similar results for zeroing out ANY position in your starting lineup--and the results would probably be even more robust, given that the QBs contribution is, on average, the largest contribution to the overall point total of any single starting position.

 

 

Perhaps you could say the same about other positions, but that was not the purpose of the article, as far as I can tell, which was to dispell the notion that having an elite QB is a recipe for success, which this shows that more often than not, it is not.

 

Further, it doesn't matter if QBs tend to have the largest contribution to score. It only proves his point, when almost all QBs tend to score consistently well, epseically over a season. Point differential is what matters, and there doesn't appear to be enough for an elite QB to push you over the top here. Quite the opposite, building a strong team elsewhere appears to be more consequential.

Edited July 18, 2013 by delusions of grandeur

[delusions of grandeur]

Continued (since it limits the amount of quotes:

 

So of all the games a team wins, they would probly still win 70-80% of them without an RB1 or a WR1 as well. In the end, this is all just a lot of meaningless statistical descriptions with little to no predictive value--the best way to maximize your win total is to maximize the points your team produces on a week to week basis, period.

 

This is something else Keg might want to follow up and look at, but if you think this is the case, then take the time and show that not having good WRs or RBs is more consequential than not having a top QB.

 

Either way, his point stands that having an elite QB is not the end all be all like many seem to think they are nowadays. I believe that was his point to this analysis.

 

Keg, feel free to correct me if I'm wrong on any of this, or just get rid of this cancer who has nothing positive to add unless it's a sad attempt to stroke his internet ego.

Edited July 18, 2013 by delusions of grandeur

[keggerz]

Many thanks for the article - exactly the type of fresh-look analysis we appreciate here.

 

I would agree with your concern about putting a "metric in context." Did you consider performing the same zero analysis with the RB1 or WR1 position? If zeroing out those positions resulted in a change in the win percentages that were statistically significant, that would strongly buttress your contention that the QB position is unique in this regard.

 

One of the first things that Jake (Swiss Cheezhead) said to me...and while it seems to make sense to do that comparison there were many reasons why I didn't think it was as easy as that. If I didn't have a deep feeling that the QB spot was the position to use for the theorem then that might have been the way to go but what is a RB1? WR1? The first RB a team selects, same for a WR? Instead I wanted to find a way to measure the value of each position and then let that math dictate what position to use for the theorem. Had the math ruled in favor of the RB position then that is what the theorem would have been, "the Zero RB Theorem" but that isn't what the math "said"

 

Also, another part of it was how many games that RBs and WRs miss in a fantasy season...be it due to injury or just being benched due to matchups...QBs are much more likely to be plug and play...but again, for me it was all about the math.

 

(again, not trying to speak for keggers), those could definitely be some good followups to consider, but I think one of his primary purposes here was to dispell this notion in recent years that you need an elite QB to win.

 

Perhaps you might find something similar if you did RB or WR, but then I don't think it's as easy to find a fairly consistent player with high-scoring potential at those positions later, like it is to find a QB that isn't ever going to actually get you near zero, like even your top WRs can get shut out on an off week. He's merely quantified what some of us already saw, that QBs scoring more points doesn't necessarily give them more value in most leagues.

 

Though I'll of course defer to him to elaborate further.

 

You hit on a big part of it, which I touched on above...and yes, just because a QB scores more points doesn't mean they are worth more from a value standpoint....think about it like this...if Kickers got 10 points per FG and 5 pts for an XP would that change when you would draft them just based on them scoring a ton more points?

[keggerz]

Continued (since it limits the amount of quotes:

 

 

 

This is something else Keg might want to follow up and look at, but if you think this is the case, then take the time and show that not having good WRs or RBs is more consequential than not having a top QB, or else stop poo-pooing on other's articles with a bunch of blatantly flawed faux analysis.

 

Either way, his point stands that having an elite QB is not the end all be all like many seem to think they are nowadays. I believe that was his point to this analysis.

 

Keg, feel free to correct me if I'm wrong on any of this, or just get rid of this cancer who has nothing positive to add unless it's a sad attempt to stroke his internet ego.

 

Another way to put it is that a good team's wins aren't dependent on QB production. The metrics go on to prove why the QBs value isn't what it is cracked up to be my many. Another point is that the average loss for a winning team if you zeroed out their QB points is 10.93(Sofa Classic), 11.15 (Sofa IDP) & 11.36 (BDFLP!)...of the 704 losses from the 1805 games researched 436 were by less than 13 points...I'm not advocating playing without a QB, rather to illustrate their lack of importance...if you take the zeroed out points and just give the QB 13 points scored (not a tall order to achieve) then the % of games that those teams would win jumps up to 85.2%.

 

If those numbers don't speak volumes...especially when looked at in conjunction with the value study for the positions that is included then you more than likely won't ever "hear" what this article is saying.

[Big Country]

So you're telling me I shouldn't draft a QB early?

 

 

Alright, but you;re going to have to twist my arm to get me to go along with it.....

[flemingd]

By zeroing out only the wins, the data is badly skewed. You're not saying zero QB teams will win 61% of their games, you are saying they still win 61% of the 50% of the games they originally won. If a team plays 13 games and wins 8, and you zero out their QB and get a 61% success rate, that's just 61% of 8, or ~4.8. That's not a 7.9 win season. So you are no longer measuring the value of an "elite" QB vs. the 16th QB, you are measuring the overall strength of the winning team at any moment and demonstrating that, if the rest of your team is good enough you can get by with any donkey. If you want to truly understand the Top 3 impact, do a data set where you zero out only their wins (not all games) and see what % of their games were still won once zeroed.

 

Curious about methodology - when you say Top 3 QB are you going with literally the first three drafted, or are you going with the top 3 as they finished the season statistically?

 

If would be more on-point to zero out the RB1, the WR1, and the TE against each data set and see what the Remaining Win % is. Do teams only 61% with a zero QB theorum but only 49% of a Zero RB1 theorum?

 

=====

 

I like the CV analysis. This is close to how I do mine, and really illustrates the supply/demand imbalance between QB and RB/WR. I always thought it was crazy to start 1 QB and 2 RB when really each team in the NFL only had one "primary" running back. This is really the heart of the issue and why QB's get devalued. If the raw data is easily accessible, run it against a 2 QB starting model if you really wanna blow your own mind.

 

=====

 

Miss rate is interesting. Any consideration for severity of miss? StDev of the miss, or comparing miss value to PPG differential?

 

=====

 

If you have raw data you'd like to share I'd love a look at it.

 

Better yet would be a true Replacement Value analysis - how often was the (Top 3 QB score - replacement QB score) the key to a win/loss? In my homer league the Top 3 averaged 26.9 and the 13-15 QB's averaged 21.4. How many games were won by 5 or less points?

[chiroacademy]

I feel comfortable with a later round (rd 8-9) QB pick. In mocks I usually end up with Luck, Stafford or Romo and I'm more than fine with that. But to slide to a Cutler or a Palmer makes me nervous because at that point I'm not seeing backs or WR that I would start. Not much WR or RB value left so why not pick up a Luck or Stafford at that point.

[delusions of grandeur]

Fleming, I'm confused where you're getting all this.

 

First you say, "By zeroing out only the wins the data is badly skewed. You're not saying zero QB teams will win 61% of their games, you are saying they still win 61% of the 50% of the games they originally won", I don't think that is what he's saying (and I also don't get where 50% comes into play when talking about winning teams who won more than 50% of their games).

 

He zeroes out the QB for all the games, with no regard to win or loss with the QB, and then determines win percentage from that. This win % without is completely separate from the win % with a QB, as far as I understand. He also determines the average points they lost by in zeroing out the QB, which as he mentions above is equally noteworthy that they only lost by an average of about 10 or 11 points (so even a mediocre QB could easily raise their win percentage).

 

Then you go on to advocate zeroing out only the wins for the top 3, the exact thing that you mistakenly thought skewed the original data?

 

As for methodology, he said top 3 ADP (drafted) QBs.

 

And like I said earlier, he could do a similar analysis for other positions, but the point was more to dispell the notion that a top QB is essenital to winning nowadays, when winning teams with a QB zeroed fared better than those with a top 3 QB.

 

There's always room for mroe research, but I think some are missing the point of the analysis.

Edited July 18, 2013 by delusions of grandeur

[darin3]

Solid data mining for sure.... took me my entire dinner break to read

 

Was a long road to get to the "you can wait on QB" endpoint, but a fun ride nonetheless.

 

Nice job!

[loaf]

Solid data mining for sure.... took me my entire dinner break to read

 

Was a long road to get to the "you can wait on QB" endpoint, but a fun ride nonetheless.

 

Nice job!

 

this minus the dinner break

[stevegrab]

Fleming, I'm confused where you're getting all this.

 

First you say, "By zeroing out only the wins the data is badly skewed. You're not saying zero QB teams will win 61% of their games, you are saying they still win 61% of the 50% of the games they originally won", I don't think that is what he's saying (and I also don't get where 50% comes into play when talking about winning teams who won more than 50% of their games).

 

He zeroes out the QB for all the games, with no regard to win or loss with the QB, and then determines win percentage from that. This win % without is completely separate from the win % with a QB, as far as I understand. He also determines the average points they lost by in zeroing out the QB, which as he mentions above is equally noteworthy that they only lost by an average of about 10 or 11 points (so even a mediocre QB could easily raise their win percentage).

 

Then you go on to advocate zeroing out only the wins for the top 3, the exact thing that you mistakenly thought skewed the original data?

 

As for methodology, he said top 3 ADP (drafted) QBs.

 

And like I said earlier, he could do a similar analysis for other positions, but the point was more to dispell the notion that a top QB is essenital to winning nowadays, when winning teams with a QB zeroed fared better than those with a top 3 QB.

 

There's always room for mroe research, but I think some are missing the point of the analysis.

 

 

NO, I think fleming is right, only the QB score of the winning team is being zeroed out. Just look at the example he shows at the begiining, the QB of the winning team has their scored zeroed, and not the losing team. That seemed strange to me as well because the QB of the winning team may not be the STUD or top 3 QB that was taken.

 

I understand his basic point, that you don't need a stud QB to win. But I thought that was just common sense and not something that needed a complex theorem to prove. Not knocking kegs work or saying it is wrong, just wondering why such a complex thing was used to explain something that simple.

Edited July 18, 2013 by stevegrab

[flemingd]

He zeroes out the QB for all the games, with no regard to win or loss with the QB, and then determines win percentage from that. This win % without is completely separate from the win % with a QB, as far as I understand. He also determines the average points they lost by in zeroing out the QB, which as he mentions above is equally noteworthy that they only lost by an average of about 10 or 11 points (so even a mediocre QB could easily raise their win percentage).

 

 

 

"The findings in the following tables are based on the Zero QB Theorem. The Win % column is the percentage of games that a team still won once their points at quarterback were zeroed out."

 

Still. You can't "still" win something unless you won it in the first place.

 

50% comes into play because for every win, someone else loses. So if you and I play 13 times and I win 10, you lose 10. So between us we are 10-3 and 3-10. Then you take the 10 I win and the 3 you win, and analyze those winning games. There will always be a 50% win rate when analyzing league-level data.

 

"Instead of zeroing out the quarterback points for the winning teams if I used an average of just 12 points/game, the win % would have climbed to 83.4%, good for an average of 10.8 wins."

 

For the winning teams. Also, think about this - a team getting 12 wins 83.4% of the time? No way that's overall numbers.

[keggerz]

 

 

For the winning teams. Also, think about this - a team getting 12 wins 83.4% of the time? No way that's overall numbers.

 

that is for teams that initially won...sorry it was worded poorly...I'm in meetings the rest of the day so I'll address everything else later tonight

[keggerz]

NO, I think fleming is right, only the QB score of the winning team is being zeroed out.

 

Correct

[darin3]

NO, I think fleming is right, only the QB score of the winning team is being zeroed out. Just look at the example he shows at the begiining, the QB of the winning team has their scored zeroed, and not the losing team. That seemed strange to me as well because the QB of the winning team may not be the STUD or top 3 QB that was taken.

 

I understand his basic point, that you don't need a stud QB to win. But I thought that was just common sense and not something that needed a complex theorem to prove. Not knocking kegs work or saying it is wrong, just wondering why such a complex thing was used to explain something that simple.

 

 

While I certainly appreciate and am envious of the time/research put into keg's article, I kind of have to agree with the bolded (to a certain degree).

 

Last year QB1 scored 26ppg ("standard" scoring, using one of my leagues as an example), QB16 (assuming you're in a 16-team league) scored 19ppg for a 7ppg difference.

 

Now look at RBs. RB1 scored 22ppg and RB32 (assuming that same league starts 2 RBs) scored 9ppg for a 13ppg difference.

 

Obviously a far more rudimentary analysis but as keg showed, the multiple and in-depth angles you can look at this that only proves it further.

 

Awesome article that kind of made my head spin.

[WashingtonD]

Damn it I'm behind, I only got through half on my way to work this morning, you guys better not ruin the ending in this thread before I get there, no spoilers!

[darin3]

Damn it I'm behind, I only got through half on my way to work this morning, you guys better not ruin the ending in this thread before I get there, no spoilers!

 

 

Sorry, but...

 

 

Bruce Willis is a ghost.