Milehighhawk":2j01z10g said:
5_Golden_Rings":2j01z10g said:
90 is WAY too low. A 90 passer rating is what an 80 passer rating was in the 80s/90s. Your data is biased because of this choice. Raise that number to 95 minimum, and it probably should be closer to 100. Kaepernick had a 90.7 passer in 2016 and was awful.
Due to rule changes, a 90 passer rating is easily achievable by AVERAGE qbs. Using this number GREATLY skews your data away from great defense.
Your own data showed that 61% of qbs hit that mark, so why are you surprised that the pool with 61% hitting the mark wins more championships than the pool with 15.6% hitting the mark.
As hard as you worked, with 90 as your passer rating threshold, this analysis is worthless. All it is saying is that having a below average qb makes it hard to win championships. It says NOTHING about elite, Franchise qbs, since 90 is not even close to Franchise qb play.
I appreciate you taking the time to respond, however I am not sure your response makes a whole lot of sense. Remember the question is, is it easier to be successful as a team with very good QB play or very good defensive play. Using a lower rating threshold actually puts more emphasis on the answer that a team should pursue better QB play over attempting to form a "very good" defense because it should be even easier to reach that lower threshold.
I would ask that you take the time to understand the information a bit more before formulating assumptions. I said nothing of:
- Evaluating QBs league wide (61% is not league wide but only of the data set defined, only just under 40% of QBs were 90 or better in 2017 I believe, a very similar number to total teams that make the playoffs (12 vs. 13))
Defining Elite or Franchise in any way as being relevant to the data set.
Having a "Great" defense.
Try to think about it a bit more as it seems you have jumped to some conclusions not addressed here.
Most importantly, it's a cool thing to think about and a lot of work clearly went into this, so a serious and sincere :2thumbs:
When doing this type of statistical examination, however, you have to be open to criticism of it, as small choices can have big effects on the outcome you come to (which holds a lot of power as it can be written in a sentence that obscures all those small choices.
5_Golden_Rings is (correctly) noting an error in one of those choices. I'll try to explain it clearly, and then explain my thinking on another choice you made that is worth considering.
CHOICE 1: By using a ranking cut-off to measure defense (Top 5) and a threshold cut-off to measure quarterbacks (>90 quarterback rating) you're not measuring apples to apples.
This is a particularly big problem if you're asking a "which matters more" question, which you are.
As you correctly note that QB rating increases over time, the simplest and easiest way to compare applies to applies is to use a ranking cut-off for BOTH QBs and Defenses (e.g. top 10 starting QB Rating and Top 10 defense; top 5 starting QB rating and Top 10 defense).
That's an apples to apples comparison to measure which apple is better
CHOICE 2:
Your outcome variable is making it to the championship game. Why? Once you get to the divisional game that's a dichotomous event, which by definition will come with a whole bunch of noise (i.e. things that aren't overall quality QB play or overall quality defense -- think things like randomly bad fumble luck in the divisional game, or getting matched up against a juggernaut opponent, or your top WR randomly being injured that week).
To decrease noise, a MUCH more straightforward measure is simply winning percentage.
If you are measuring QB play and team defense on the same scale (top x for both for that year) you essentially have a 2x2 table as based on overall winning %:
Winning % for top QB but not top defense
Winning % for top defense but not top QB
Winning % for top QB and top defense
Winning % for not top QB and not top defense