football metrics
by Jay , San Diego, Tuesday, September 24, 2013, 10:58 (3858 days ago)
edited by Jay, Tuesday, September 24, 2013, 11:14
I whipped this up thinking about the different ways to analyze football games, and the different levels you can look at. Games are made up of drives, and drives are made up of 4-down series, and series are made up of plays. And some measurements have been developed to look at each level. Supe's "y-score", for instance, goes down to the play level. Fremeau's FEI looks at drives and possessions.
I think most people look at outcomes and then dive straight down to per-play measurements. I like FEI and drive-based measurements (like Available Yards) because plays don't happen in isolation: oftentimes it's the combination of plays that make for a good offense (or bad offense).
"Drive Efficiency" is another good one that I don't see anybody doing for college football (FO does it for the NFL). It's basically the percentage of first downs + touchdowns / # of 4-down series.
At a high level...
by Jim (OFD) , Naptown, Tuesday, September 24, 2013, 16:33 (3858 days ago) @ Jay
I think a high level metric like total yards per point scored could go a long way in describing overall offensive efficiency. An efficient offense would net out around 10 yards per point.
Comic Sans, eh?
by oaknd1 , Notre Dame, Tuesday, September 24, 2013, 15:32 (3858 days ago) @ Jay
Courageous.
passive aggressive smiley face :)
--
It's really pathetic when someone includes a link to their Twitter profile in their signature.
Chalkboard!
by Jay , San Diego, Tuesday, September 24, 2013, 16:02 (3858 days ago) @ oaknd1
- No text -
(grumble grumble)
by oaknd1 , Notre Dame, Tuesday, September 24, 2013, 16:51 (3858 days ago) @ Jay
- No text -
--
It's really pathetic when someone includes a link to their Twitter profile in their signature.
Oh my.
by irishoutsider , Tuesday, September 24, 2013, 13:54 (3858 days ago) @ Jay
- No text -
THIS IS AWESOME
by Supe , VA, Tuesday, September 24, 2013, 13:40 (3858 days ago) @ Jay
At work now - I will respond more later
I've played with some series data.
by LaFortune Teller , South Bend, Tuesday, September 24, 2013, 12:19 (3858 days ago) @ Jay
I think I want to do more, but I haven't had an opportunity to dig into as much data there. See the top right charts:
I've been thinking about a way to visualize a drive chart
by Jay , San Diego, Tuesday, September 24, 2013, 15:13 (3858 days ago) @ LaFortune Teller
Something similar to a baseball inning line, you know, like:
0 0 0 0 2 0 0 1 0 | 3 1 0 0 0 0 0 3 0 - | 4
I think you could it with symbols or colors and easily show opportunities missed & made on a per drive basis.
On yards-per-play
by Tim, Chicago, IL, Tuesday, September 24, 2013, 11:40 (3858 days ago) @ Jay
This stat just feels incomplete by only looking at averages without also looking at standard deviations, which would bring some sense of how consistently or "chunkily" a team gets its yards.
Consider a team that has a 50% chance of an 8-yard gain and a 50% of no gain on every play. That team is going to have a drive stall when only one-of-three after a first down come up 8 yards, which will be 50% of the time. It's YPP will average a 4.
Consider another team that has a 100% chance of a 4-yard run on every play. That team is going to have a time consuming TD on every drive. It will never lose. It's YPP will average 4.
I guess the mean in and of itself isn't useless, and coming up with standard deviations is probably unbelievably time consuming work, but I just can't bring myself to care too much about the average yards per play without more of the picture being exposed. Does anybody in the #SPORTSMATH universe produce these numbers?
i think you can guess my bias
by Supe , VA, Tuesday, September 24, 2013, 17:24 (3858 days ago) @ Tim
I like the ypp because I think it's non-judgmental. While the big plays can skew it - over time, the larger numbers help to average it out.
So - at this point in the season, we can determine's ND's success by 4 data points (3-1 record). Or by 560 data points (assume 70 plays game for two teams for four games).
Granted, with the mix of big vs small plays can skew that, but the volume of data creates a very powerful simple metric.
I think what ypp is missing is a measure of ultimate drive/series success. So, I find measures that are based on a drive basis compelling. Although, I would argue that, at its core, the ypp is a core metric of ability. And a mix of drive/series + a measure of turnovers/special teams would be a very compelling measure.
exactly why I like the drive-based stats
by Jay , San Diego, Tuesday, September 24, 2013, 15:10 (3858 days ago) @ Tim
You're exactly right that the long plays can skew the yards per play averages quite a bit. So which is the better offense: the one that hit an 80 yard bomb but went three and out the other three times, or the one that moved the ball effectively on all four drives, but without the long bomb?
I think plays can be "won" or "lost" independent of yardage
by Jeff (BGS) , A starter home in suburban Tempe, Tuesday, September 24, 2013, 18:20 (3858 days ago) @ Jay
Not completely independent, but gaining 1 yard on 3rd and 1 is a "win" for the offense, while gaining 9 yards on 3rd and 10 is a "loss". A 6 yard gain on 1st and 10 might also be considered a "win" while a 6 yard gain on 3rd and 7+ is again a "loss". I suppose that is what you are going at with the series measure, but the per-play metrics always seemed a bit suspect to me.
This is an excellent point I hadn't considered.
by Tim, Chicago, IL, Wednesday, September 25, 2013, 07:05 (3857 days ago) @ Jeff (BGS)
- No text -
definitely. We kicked this around before
by Jay , San Diego, Tuesday, September 24, 2013, 20:16 (3858 days ago) @ Jeff (BGS)
Maybe dumb question: series is a subset of drives?
by MHB (Rakes of Mallow), Chicago, IL, United States, Earth-199999, Tuesday, September 24, 2013, 11:39 (3858 days ago) @ Jay
- No text -
just shorthand for a 4-down series
by Jay , San Diego, Tuesday, September 24, 2013, 11:41 (3858 days ago) @ MHB (Rakes of Mallow)
A drive may have one or more 4-down series.
Thx. I guess I'd never verbalized it that way. Makes sense.
by MHB (Rakes of Mallow), Chicago, IL, United States, Earth-199999, Tuesday, September 24, 2013, 12:16 (3858 days ago) @ Jay
- No text -
I assume series begins on 1st down
by Tim, Chicago, IL, Tuesday, September 24, 2013, 11:41 (3858 days ago) @ MHB (Rakes of Mallow)
- No text -
BTW, Pat Forde was complaining about the NCAA site.
by domer.mq , Tuesday, September 24, 2013, 11:11 (3858 days ago) @ Jay
I was too busy yesterday, but I noticed he was complaining about the new site layout on Twitter yesterday. I wanted to dig up the e-mail address this gang had found and send it to him. He had, essentially, very similar complaints as this board.
--
Sometimes I rhyme slow sometimes I rhyme quick.
Mark posted this the other day
by Domer99, John Wesley Powell's Expedition Island, Tuesday, September 24, 2013, 11:52 (3858 days ago) @ domer.mq
http://bluegraysky.com/forum/index.php?id=172083
Apparently the name is Jeffrey S. Williams.
jswilliams@ncaa.org
that's awesome
by HumanRobot , Cybertron, Tuesday, September 24, 2013, 11:08 (3858 days ago) @ Jay
I've been thinking of something similar as well.
The other important concept might be how data is cleaned (what is the definition of garbage time or other dropped data) and normalized.
Getting the data is more difficult, but you also have the
by Jason93 , Raining debris all over Europe, Tuesday, September 24, 2013, 11:21 (3858 days ago) @ HumanRobot
individual player performances/decisions on every play on both sides of the ball. The play is essentially the aggregation of the individual performances.
EDIT
This would also vary from team to team and would require knowledge of the play call, etc, so probably only available as internal metric.
--
I think the children like it when I "get down" verbally.
Individual Player Grades
by PBurns , Ah Denver, the Sunshine State. Beautiful, Tuesday, September 24, 2013, 11:51 (3858 days ago) @ Jason93
If they could be somewhat normalized, could realistically be combined with some of these statistics. To what end, I have no idea.
--
Just run 4 verts
garbage time, SOS, FCS competition, kneeldowns
by Jay , San Diego, Tuesday, September 24, 2013, 11:13 (3858 days ago) @ HumanRobot
short yardage situations. All things that can prejudice the data. (Is that a real phrase? "Prejudice the data"?)
And there's there situational cuts: Home/Away, Night/Day, for example. In the Red Zone. Or 3rd down. Running the ball. Passing the ball.
Probably "skew" or "confound" rather than "prejudice".
by Savage, Around Ye Olde Colonial College, Tuesday, September 24, 2013, 15:09 (3858 days ago) @ Jay
- No text -