The CardBoard
All PAC-12 teams - Printable Version

+- The CardBoard (https://thecardboard.org/board)
+-- Forum: C-House! (https://thecardboard.org/board/forum-4.html)
+--- Forum: The CardBoard (https://thecardboard.org/board/forum-5.html)
+--- Thread: All PAC-12 teams (/thread-8403.html)

Pages: 1 2

Re: All PAC-12 teams - CTcard - 12-03-2014

(12-03-2014, 03:37 PM)stupac2 link Wrote:Basically I use F+ because it's the only game in town,

It's the best known, but it's not the only statistical work up that tries to adjust for strength of schedule.

One of my favourites is Steve Durrett's Relative Performance Ratios (RPRs). Steve comes to this board to publish links to his photos, but for whatever reason doesn't come to link his RPRs and game projections. If you are comfortable thinking in ratios, you can develop intuitive understanding of what they mean rather quickly. They are only singly-connected in adjusting for strength of schedule (all ratios are for opponents' averages), but that is at least the most important step.

He's still one week behind as is the Football Outsiders' web site. He doesn't try to find a single number to define how good a team, or a unit, is - but he does find Stanford defense at #1 in the country in RPR yds allowed per play, #2 in yds/allowed per game, and #3 in pts per game.

http://durrettphoto.com/html/football_statistics.htm

p.s. While Steve goes to great lengths to note that his individual game analyses are "projections, not predictions" they do seem to provide good insight later into seasons. Last week, for example, he projected Stanford over UCLA (though by a much smaller score), BYU over c.a.l. (though by a bigger score), and had all of the correct winners.

Re: All PAC-12 teams - BostonCard - 12-03-2014

Well, then let me put in a plug for our very own Steve Durrett's RPR (Relative Percentage Ratios).  As you wanted, it adjusts statistics for opponent quality.  As CTcard likes, it can easily be reproduced, and, in fact, Steve makes his spreadsheets available to anyone for download.  Also, it has a pretty straightforward explanation, the RPR is a ratio of "what a given team has done to what  its opponents have collectively allowed in other games."  Thus, an RPR of 1 means that a team does as well against its opponents as other teams have; ratios above 1 for offense and below 1 for defense mean that the team does better against its opponents than other teams have; ratios below 1 for offense and above 1 for defense suggest a worse performance than what would have been expected.

EDIT: Link here: http://www.durrettphoto.com/html/football_statistics.htm

By that metric, the Stanford had an outstanding defense.  Here are our defensive RPR's (remember, numbers less than 1 are good, except for turnovers) and ranks by category:

Rushing
Attempts 0.92 (40)
Yards per game 0.65 (15)
Yards per attempt 0.71 (13)
TD 0.51 (19)

Passing
Attempts 0.95 (40)
Completions 0.86 (19)
Completion percentage 0.91 (19)
Yards per game 0.64 (2)
Yards per attempt 0.67 (1)
TD 0.43 (5)
Interceptions 1.38 (28)
Interception percentage 1.45 (21)

Total defense
Plays 0.93 (26)
Total yards 0.64 (2)
Yards per play 0.69 (1)
Pts per game 0.48 (3)


In other words, our opponents scored less than half as many points against us as they did against their other opponents and got 31% fewer yards per play against us than they did against their other opponents.

BC

Re: All PAC-12 teams - CTcard - 12-03-2014

(12-03-2014, 04:06 PM)Boston Card link Wrote:Well, then let me put in a plug for our very own Steve Durrett's RPR (Relative Percentage Ratios). 

I think this is where I say the internet bulletin board equivalent of "jinx".

Re: All PAC-12 teams - BostonCard - 12-03-2014

Haha... Yeah.  I started typing before you posted, so I hadn't seen your post.  Great minds think alike... But so do feeble ones.

BC

Re: All PAC-12 teams - terry - 12-03-2014

(12-03-2014, 03:26 PM)CTcard link Wrote:The next step is to think about F+ and what does it really measure versus everything else one could use to measure a defense.

F+ does seem to have some strengths, but there are several things that concern me. One is that while they describe the metric in some depth in words, I don't believe they have published enough detail that one could reproduce their numbers on one's own.

More fundamentally, I am suspicious of two things that are done here: one any statistic that is too complicated to simply understand [inherently what does a defensive F+ of 14.3% mean?]; and I also think that a complex, qualitative question like "how good is a defense" cannot be answered by one number regardless of the algorithm. 

I absolutely agree. I don't like metrics that just don't make any intuitive sense.

And I agree that Steve Durrett's RPRs do make intuitive sense.