A Mathy Long View of UCLA Football
by KnudsenRockne on Oct 7, 2011 7:23 AM PDT
Email
PrintI compared UCLA's Offense and Defense with the "Pac-Ten core teams" (henceforth 'Pac-core') - our traditional rivals - from 1928 to 2010. I calculated season averages of scores for each game played by Pac-core teams (both teams). I was feeling lazy so I didn't explicitly reject duplicate Pac-core teams when they played other Pac-core teams, so they entered into the calculation twice. I also calculated UCLA's average points scored and points allowed for each season since 1928.
Results:
via i52.tinypic.com
Discussion:
The top figure (in black) shows the Pac-core wide average scores (with error bars representing the SDOM - if you don't know what that is you probably don't care). There is a quasi-periodic variation in the scores with distinct peaks around 1927, 1950, 1970, and perhaps 2000. The scores also show a growth trend- the solid black line is the linear fit to the data. Most data points fall along the line (within error bars) except for the peaks. The slope of the fit is about 0.25 points per year. (The Boelter Hall types will want to know the X intercept - it is 1887, the same year Notre Dame started playing football. This is pretty good evidence that there are coincidences that are completely meaningless). In the past few years the scores fall below the trend line but almost within error bars. [Any variation smaller the the width of error bars probably isn't important]
The middle figure (in blue) shows average points scored by UCLA for each season. It, too, shows the overall upward trend and quasi-periodic variation (the linear fit is plotted on all three figures to provide a reference). Notably, the UCLA periodic fluctuations have a larger magnitude than the Pac-core but the peaks don't coincide exactly - and these variations are significant compared with the error bars. A five-year running mean is plotted (in blue) to highlight short-term trends in UCLA's offensive performance. There is a distinct downward trend from ~1997 to present. This is similar to a downward trend from ~1954-1963.
The bottom figure (in red) shows average points allowed by UCLA for each season. It, too, shows the overall upward trend and quasi-periodic variation (the linear fit is plotted on all three figures to provide a reference). Notably, the bulk of the averages lie below the line suggesting that UCLA has typically had a better defense than the Pac-core teams (for the Boelter crowd: there are very few seasons in which UCLA was above the Pac-core average fit by more than an error-bar, there are more seasons when it is below the Pac-core average fit by more than an error bar) . The five year running mean rarely exceeds the black line. Surprisingly, from 1997 to present there has been and overall downward trend in points allowed: during UCLA's decline the defense has been performing better compare with our core rivalries.
Conclusion:
The slump of the past decade and a half seems to be more due to underperforming offense than bad defense: we are scoring about 10 points fewer each game than our traditional rivals and this is the largest dip UCLA has seen. If we want to fix UCLA football this would be a good place to start.
[Added this bit] Over the near century of data, UCLA routinely performs better on offense and on defense than the rest of the 'Core Rivals' (Pac-10 schools) so who says we're "just a basketball school"?
[Added bit of math 10/6] As requested below, I calculated correlation coefficients between wining percentage and points scored as well as points allowed: The coefs are 0.685 and -0.381 respectively. Then I subtracted the trend-line from UCLA points scored and allowed and found those correlation coefs: 0.754 and -0.623 respectively. Those are pretty strong correlations. (I also generated scatter plots available upon request) the difference between UCLA's performances and the conference average are better predictors of winning percentage than raw UCLA points.
This is a FanPost and does not necessarily reflect the views of BruinsNation's (BN) editors. It does reflect the views of this particular fan though, which is as important as the views of BN's editors.
35 comments
|
1 recs |
Do you like this story?
Comments
holy smokes batman
Neat post – well done my friend. I can safely say you don’t find many posts like this on other SN sites. You should get university credit for this, or at least an honorary visit to the football locker room for a presentation. Apparently no one in there knows what the hell’s going on.
This is very cool . . . but these numbers are misleading.
For example, we can see that the most recent scoring peaks in the last 15 years seemed to coincide with our last three 10-win seasons (i.e., 1997, 1998, and the one KD got thanks to MJD). We also scored a tone of points in 1973 or thereabout . . . I was 2 so Im a little weak on that. Anyhoo, in none of those seasons did we fair exceptionally . . . no Rose Bowl wins and 50/50 against SUC. UCLA’s great teams during the Beban years won a Rose Bowl and didnt score as much. And TD’s ultra-conservatism kept scores low though we forget he won 3 Rose Bowls in a 4-year span and something like 7 straight bowl games. The 1988 team led by T—- Aikman (I dont like to write his first name) was ranked #1 in the country. TD coached boring football, but he could do that because he always had a great defense in the 80s. I’d love to see how w/l percentage correlates . . . otherwise, the math is great but common sense dictates that teams with underperforming offense and bad defense will tend to lose.
Thanx for the feed back.
I gotta say, I do not see any discrepancy between these plots and the statement “common sense dictates that teams with underperforming offense and bad defense will tend to lose.”
If anyone can point out any discrepancies between these plots and any common sense expectations, I’d love to hear it. I may have messed up in my analysis, or maybe there is actually something new and exciting to learn. (Supporting the same ol’, same ol’ just isn’t fun to a scientist).
BTW: The assertion that Donahue had great defense is supported by the chart but the assertion that he “kept scores low” is not (at least as compared with the rest of the conference).
Play with so much passion nothing else matters
by KnudsenRockne on Oct 5, 2011 9:47 PM PDT up reply actions
Your wish is my command
I added the correlation coefficients to the above discussion – it turns out points scored and points allowed are good predictors of wining percent (especially after subtracting the conference average).
Thanx for the suggestion – keep ’em coming!
Play with so much passion nothing else matters
by KnudsenRockne on Oct 6, 2011 1:51 PM PDT up reply actions
No! No, touche - we found something out. I THANK you for raising the issue!
I may have been glib in how I presented it but, Alcides, I really appreciated the challenge. I don’t know if you are a South campus type or not, but we thrive on challenging ideas. It is the refiner’s fire. (Ad hominen attacks aren’t so welcome — but you didn’t do any of that)
I apologize if I’m abrasive in my responses on any challenges but I really like them. (In fact, Ill tone them down henceforth). I had some scatter plots about this – and the interesting this – I’d call it a “finding” is that UCLA’s raw points score (or points allowed) are not as good of predictors of winning seasons as points-conference average. This tells me that it is appropriate to discuss UCLA’s performance in context of our conference and no ‘in a vacuum’
Play with so much passion nothing else matters
by KnudsenRockne on Oct 7, 2011 7:56 PM PDT up reply actions
I think I'm starting to get this math business
Are you saying that if we get a certain number of points in a game, and the team we’re playing gets less than that number, there is somehow some way to predict the outcome? That’s a little far fetched, in my opinion. Nowhere do I see any discussion of the all-important “time of possession” statistic.
Dammit Fox. You know that when you were punching those computer cards in school that the key stat was time outs remaining
I would wager that the team with the most timeouts at the end wins at least 51% of the time . . . except for CBH of course, but thats a different game
Uh, Knudsen...do you mean to suggest that
the team that scores more, wins? Amazing! just teasing…really nice work on this, no doubt…i’m a political scientist and I really appreciate analysis like this.
Hey *F*lorida *U*niversity to all Y'all1 ;)
I have two responses to all this drivel – one is condescending and the other is sarcastic (no, I tell a lie – the other is oppressively precise – but it’ll take a while to write and to read).
I should offer a poll of which response to write but I won’t. This weekend depending on my mood I’ll explain in a manner which suits me. LOL (that should have an evil tone to it, BTW).
Play with so much passion nothing else matters
by KnudsenRockne on Oct 7, 2011 8:00 PM PDT up reply actions
Yeah I didn't see...
the discrepancy either. Your data and that assertion don’t seem to be mutually exclusive….
One comment on the defense graph
Since the season averages include games against non-Pac-10 teams, typically worse teams, I would expect most Pac-10 teams to have higher points scored averages than points against averages. In theory, our defensive line (and the line for every Pac-10 team) should be somewhat below the average points scored line — the games against bad teams deflate the points against average while also pushing the points for average up a little. A comparison between our points against average and the seasonal points against average might be a better way to look at how our defense has matched up to the rest of the conference.
All in all, though, this is pretty cool.
Good point. OOC games will influence the values
In some cases, those OOC teams are like Notre Dame and Nebraska and in others, SJSU or Montana St – so some of the effect is offset. Even if it weren’t, I don’t think it is a killer – the error bars are wide enough that I doubt this effect would lead to drawing bad conclusions. Still, we should keep this in mind when trying to interpret data.
BTW: I am glad you mentioned this because on the top chart the way I calculated average scores included each Pac team twice when it played a Pac team (once as the home team and once as visitor). I could have written a function to search opponent names and deal with that but I was feeling lazy ;) But maybe that is better because it emphasizes Pac teams. (I did not count Pac games twice for UCLA’s scores so that is something to keep in mind.)
I’m not sure I understood your recommendation, if you could spell it out again I’d like to hear it. Maybe I’ve already done it.
Thanx again.
Play with so much passion nothing else matters
by KnudsenRockne on Oct 6, 2011 8:57 AM PDT up reply actions
My thought was to essentially do a version of the first graph, but looking at average points against.
The black line establishes the trend in offenses, but it might be useful to also see the trend for defenses. Something like doing a search on all games involving Pac-10 teams, then picking the score from the other team and averaging that.
My bad.
The black lie contains points scored and allowed. I shoulda explained that more clearly-ier. and all that.
Play with so much passion nothing else matters
by KnudsenRockne on Oct 7, 2011 8:02 PM PDT up reply actions
Ah
I still think it would be interesting to isolate it with different lines for points scored and allowed. And since they’re combined right now, it kind of makes sense that our offense tends to be above the line, while our defense tends to be below it. I’ll try to explain what I mean:
In conference games, the points scored are zero-sum. Points scored by Stanford against Cal are also points given up by Cal against Stanford – totaled over the whole league, there’s a zero point differential in conference games. On average, the offenses are going to look just as good as the defenses, and we can divide the total number of points by the number of teams involved.
In non-conference games, points scored for and against the conference aren’t automatically equal. Since the average non-conference opponent is likely somewhat worse than the Pac-10 team involved (more SJSUs than Notre Dames), the league is going to win more games and gain points. In non-conference games, the conference likely usually has a positive point differential. The average points scored should be somewhat higher than the average points given up. When we combine that with the zero-sum conference games, there will still be those same extra points, more points scored than allowed. The average allowed and average scored would be somewhat different.
I think with the black line, those two lines (‘points scored’ and ‘points against’) have basically been averaged out. In effect, it’s like saying all of our non-conference games are ties, giving the other team some of the points we scored.
And that is why you cannot separate the black line for conference play
consider the Cal-Stanford game as well as the Stanford-Cal game both scored and allowed points will enter into the average.
But if you want to help, go grab a bunch of Historical records (I’ll provide the website if anyone wants to help out on this endeavor) and scrape the screens. If you have lynx it is pretty easy to convert a HTML page into a text file and then just edit out the header and footer and send them too me.
Then I could do this analysis on all of FBS
Play with so much passion nothing else matters
by KnudsenRockne on Oct 9, 2011 11:43 AM PDT up reply actions
One line for conference games, probably two for non-conference games.
For conference play, the two lines (points allowed and points scored) would be the same, something like the black line. But for non-conference games, they would be two different lines. Combining the two sets (conference and non-conference) into a whole season points per game calculation, you’d still wind up with two lines. With those two lines, it might explain why our offense tends to be above the line you have and our defense tends to be below it.
(Unless, that is, if the line you have is only for conference games, and the UCLA points for and points against data you have is also only for conference games…)
But if you link the website, I’ll see what I can find.
HEY!!!!! ALRIGHT!!! Start here-> http://www.jhowell.net/cf/scores/byName.htm
Knock yourselves out with this URL – it’s always better to have many eyes look at data. (It helps determine if neutrinos did go faster than light, for example). In fact, anyone who wants to collaborate -or share – or ask for tips – directly with/from/of me must figure out this ‘charades clue’, my e-mail address sounds like:
Saint Thomas of Canterbury’s lastname + UCLAlumni.net
I really, really, really, hope someone rises to this challenge and we can have a vigorous discussion on this.
Play with so much passion nothing else matters
by KnudsenRockne on Oct 9, 2011 8:07 PM PDT up reply actions
Wouldn't it be easier
to go through each team’s schedule and only count how many points the team scored in conference games. That would be more efficient, plus throwing out the non-conference games means you actually get a well-defined quantity for average points per game.
Not if it is done by computer and
Not with the way this data set is stored. I’ve been screen scraping for years to get my data.
My brother is a (semi-retired) database engineer and if I can ever bribe him to help me import all of these disparate files (with pseudo-unique formats) into one comprehensive DB then we’ll be running my rod on AvGas.
But he didn’t go to UCLA and is afraid that I’ll give the Bruins an unfair advantage. (Damn that book & movie Moneyball !!!)
Play with so much passion nothing else matters
by KnudsenRockne on Oct 9, 2011 8:14 PM PDT up reply actions
I'm a north campus type
so all I heard was numbers numbers numbers, hire an offensive minded coach.
Is that about right? ;)
The best thing you can do for your children is to love their mother. John Wooden
Don't know about "North Campus type"
but it does remind me in High School when I had a crush of Suzy Weiner. She ended up marrying Mark Spitz. In other words…she was WAY out of my league.
Good Job KR!!!
Did she hyphenate her last name?
by Westwood is the best wood on Oct 7, 2011 10:20 AM PDT up reply actions 1 recs
-1
you are a bad man. I assume you’re a man. I think I’m right.
Then again, that probably made Jay Leno’s Headlines segment…
But hey, what do I know. I’m just the 800 lbs bruin in the room.
Suzy walked down the aisle to the tune of "When the swallows fly back to Capistrano"
Its okay for me to say this because Im female. :)
Bad man?
Disappointed to be -1ed.
by Westwood is the best wood on Oct 7, 2011 11:14 AM PDT up reply actions
ROFL...
Unlike that Congressman from NY Anthony…her names was pronounced winer. Very funny though. Man you guys are funny.
A geezerly, kinder gentler look.
Following Knudsen’s lead, I decided to try to analyze the Bruin football team but from a different perspective. I selected several poets at random, and then checked the meter and rhyme sccheme used by each, and compared it with (a) UCLA’s first down offense and defense. In fact, I did this for every down (excluding the last two minutes of each half, which I assumed would be anomalies.
I wish you could see the charts and graphs I prepared, because you would be shocked and amazed by the results. In fact, when I used color coding based on “Horatius at the Bridge,” it actually made a picture that looked remarkably like Moore Hall viewed from the back. Anyway, the charts were pretty amazing.
My analysis may still be incomplete, but I have us going undefeated in 2009, but only playing 3 games in 2010. I am not sure of the significance of all of this.
LOL. Too true.
Play with so much passion nothing else matters
by KnudsenRockne on Oct 9, 2011 11:44 AM PDT up reply actions
A Non-Condescending Comment
The point of posting these plots was to show (ie. with graphs) a couple of trends in Pac-10 teams as well as UCLA:
1) An overal linear growth of points scored with a growth rate of nearly 0.25 points per season
(after 20 years scores, on average, are 5 points higher).
2) A series of peaks in the game scores – the two most prominent at around 1950 and 1970.
These peaks are visible in the (more noisy) UCLA points scored – and as troughs in the
points allowed. I wonder were there are rules changes which occurred during these eras?
Interestingly there is another Score-Peak/Allow-Trough for UCLA ~1990 – but it isn’t very
visible in the conf avg. Hmm 20-years between peaks … kinda like the Babcock solar cycle!
3) UCLA’s points scored falls mainly above the conference average trend-line and allowed
below it – but rarely by more than 1 sigma (StDev). Note the line itself is there to compare
the data points in each of the three plots … the main thing to worry about is the trend
over the past 10-15 years.
4) The correlation coefficients were far from surprising – I kindly took the time to do that
calculation since a BN’er gave a response and asked to see it… just call it a “reality check”
So I have two questions for this list: What could have caused those peas around 1950 and 1970? (In the conference avg). Why would the average scores be increasing in this fashion? (Maybe a linear fit is not appropriate – OTOH: It fits the data well).
Play with so much passion nothing else matters
Peaks in scoring in '50s and '70s
I wasn’t around enough for the 50s to know. As a guess, maybe it was a little more prolific passing game. Remember that the entire Big Ten was viewed as a bunch of fullback teams who went up the middle on every play, and they did so successfullly. So was there an increase in passing starting in the 50s? I don’t know – just a guess.
For the 70s, it seemed to me that teams got faster, and tried to exploit that speed a little more. UCLA went back to the single wing of the 50s when Pepper Rogers had the “Blair Pair” (although it was called the wishbone.) But about this time there were all the variations on things, and new offensive theories. Again no data just a hunch – maybe offenses had a little wrinkle or two added that the defense hadn’t caught onto.
I am aware of nothing elise that’s measurable, such as the increase in the height of the mound which caused (or at least contributed to) the huge decrease in ERA in the 60s.
Knudsen – I looked at recruiting rankings and compared them to final AP rankings a year and two years and three years later, but it was not a particularly scientific study. This looks like something that would really be your area of expertise. My conclusion was that there is no real correlation between recruiting class rank and final standings rank, at least for the schools in the bottom half of the top 25. I would reall be interested to see what you could do with this.
Hey Fox
You of all people can unravel my e-mail address… I would really like to do a useful analysis on how well the ranking of athletes matches the actual performance of the teams. What I really crave is a forum where we can kick these data sets and ideas around until we can see something new. Crunching numbers is the easy part – in fact, number crunching tools are so much crazier now than 20 years ago… the trick is to ask the right question.
To anyone else in BN-land I will crunch any and all data (unless it is in effing EBCDIC ;) contact me.
Play with so much passion nothing else matters
by KnudsenRockne on Oct 9, 2011 8:26 PM PDT up reply actions
You give me too much credit, knudsen
I can’t even figure out my own email address half the time. Ask the mods to send your email address to me, as I hereby asks the mods to send my email address to you.
Comments For This Post Are Closed
Secondary Sidebar
User Tools
Related FanPosts on Bruins Nation
FanPosts
Community blog posts and discussion.
