clock menu more-arrow no yes

Filed under:

How FEI and S+P see the games: who watches the watchmen

New, 10 comments

Accountability and reliability is hugely important. How good are these predictions from advanced stats in seeing how games go? How good is Kalon in interpreting them?

Quis custodiet ipsos custodes?
Quis custodiet ipsos custodes?
Steve Dykes

Something that has been asked for more than almost anything is the simple question: how good are these predictions? How useful are they in determining anything? I think this is an absolutely fair question to ask.

One of the things that bugs me about most sports reporters is their inability to change their prediction styles as they move on. They're very absolutist in their decision making (as most sports people are; this alone makes a lot of argument occur) but they're not particularly self-documenting - and they don't appear to be able to say much about being wrong or why they were wrong. I've tried to personally do a lot more of that in the 4 years I've been doing these articles but I will confess that I don't always get it right.

I do, however, always try and figure out why I didn't get it right and incorporate that into predictions later. That sometimes works out well...and sometimes works out really badly, like it did with this year's Stanford prediction. (I've been 0-3 with Stanford so far).

So here's what I've tried to do. I've gone back and looked at every single article I've written, including the fanposts that started this all, and see whether I did the following:

  • Got the winner right
  • Got the spread right
  • Got the general feel of the game right
  • Pointed out any special details that might not have been obvious
  • Blew it on special details
I'm not going to grade myself or anything; that seems silly. I will aggregate the results so you can see what is more or less reliable. There were a total of 26 games analyzed and predicted over that time. Let's see how it went.

Onto the results!

Who won: 20/26 (76.9%)

As expected this is fairly high; it's hard to get a lot of the Oregon games wrong. Interesting to me I've had as many wrong predictions this year (3) as I had the prior 2 years combined. I blew it on Arizona, UCLA and Stanford this year. The others I got wrong were Stanford 2013, Stanford 2012, and Arizona 2013.

First conclusion is clearly: don't trust me to pick the Stanford games. I'm positively Ted Kellyian in my predictive ability there.

There doesn't appear to be any correlation between early in the season and late in the season either.

Against the spread: 15/26 (57.7%)

Well, I guess I am beating Vegas and beating the F+ standard picking, so...that's good, I guess. Almost every game I picked wrong as to who would win I also blew the spread on. Other games I got wrong were games like MSU (thought it'd be way closer), Texas (ditto), Oregon State in 2013 (thought it'd be a crushing win for Oregon) and a couple other games that I thought would be closer than they were. I was oddly decent at predicting close games or blowout games correctly, but the 12-20 point range was tougher.

Game flow: 13/26 (50%)

I was honestly expecting to do a lot better on this, but I also graded myself fairly harshly. Some games I got right on the nose:

  • Rose Bowl 2011 (this is still probably the best job I've done on predicting the overall flow of a game, down to which players would have breakout successes and how the drives would look)
  • Kansas State 2012: the special teams, difficulty with a good Wildcat defense, Oregon's ability to shut KState down and a tough day passing the ball were all good highlights here.
  • UW 2014: advanced stats didn't think highly of UW and still don't, and Oregon rolled on all fronts, especially being able to do methodical, grindy drives
  • Utah 2014: Utah would have some success but Oregon would pull away late
And some big ol' failures of failitude:
  • USC 2012: I didn't think USC would have that much success passing, and I really didn't think Mariota would have that much success passing. I didn't see it as being an absurd offensive race at all.
  • Stanford 2013: I saw Oregon as being able to hold Stanford some of the time and Stanford not having a great running attack. Ugh. UGH.
  • 2013 Texas and Oregon State and Arizona: All of these said 'Oregon should be able to run all over these guys'. All of them didn't account for Mariota's injury effects. Also - no one thought Oregon State would be able to run for 200 yards on Oregon.
  • UCLA 2014 and Stanford 2014: Jake Fisher's stunning return made the numbers look stupid when predicting a hard time running the ball
In general the game flow goes south when injuries are an issue. I blew every game between Stanford 2013 and UCLA 2014, and only one game (MSU) would be one without substantial injuries. Either Mariota was hurt or the line was in shambles, and both of those things throw the numbers somewhat out of wack - especially when either the majority of the data is without injury or there just isn't that much data.

Special details caught:

This is one of those ones that I think is really neat in general, and often does come up correctly. Some cool things that advanced stats predicted that you would be surprised about:

  • Oregon just having explosive drives galore against Wisconsin but Wisconsin having absurdly great methodical drives and the game being an offensive showcase (the general consensus was either Oregon was going to struggle to move the ball or Oregon was going to destroy Wisconsin)
  • Oregon's defense being good and getting a lot of stops against Arizona in 2012, as well as Oregon having some difficulties sustaining drives
  • Mariota running for more yards than he threw for against ASU in 2012 and Oregon stopping drives via turnovers
  • Mariota having a breakout game (4 TDs) vs. the Huskies in 2012
  • USC's receivers having big, explosive catches against Oregon in 2012
  • Oregon having a lot of field position success against Cal in 2012
  • Hogan being able to pass against Oregon in 2012 and Stanford getting a lot of valuable return yards
  • Oregon getting a kick return for a TD against Kansas State and a good defense getting stops against Kansas state on long 3rd downs
  • Oregon having problems with WSU passing but not having problems with stopping WSU in general
  • being able to sack the heck out of Hundley in UCLA 2013, Oregon having success with methodical drives and Anthony Barr as the spy on Mariota
  • Teams that are great against the run are death to Oregon (Stanford 2013)
  • Oregon succeeding passing against Utah in 2013 and not running well
  • Oregon not being able to finish drives well against Texas 2013, Oregon's 'good' D showing up and dominating the game like the stats thought they might (which to be fair I didn't think likely)
  • Oregon getting big explosive plays against MSU as well as a high conversion rate
  • Washington State having a good pass rush against Oregon
  • Arizona not getting any 3 and outs and Oregon doing better on passing than rushing
  • Oregon getting big long drives against UCLA and no huge explosive plays
  • More methodical drives from Oregon against UW 2014 along with Freeman having a great game
  • A big game for Mariota against Cal (shocking, I know)
  • Erick Dargan having a great game against Stanford, Oregon having methodical drive success
  • Oregon coming down to earth somewhat against Utah, the defense getting turnovers and stops
As LosPatos said - these aren't super meaningful unless you also look at what was predicted that I got wrong. Here are some of the many doozies.


Special details that were totally wrong:

  • I didn't think Oregon's D would be that good against Arizona 2012. Just good.
  • WSU was better on defense against Oregon than I thought they'd be in 2012
  • I called out for a punt block or FG block against UW in 2012. Oops.
  • Mariota being a passing machine against USC in 2012
  • Cal being able to stop Oregon's run game in 2012 (we passed instead)
  • Stanford's running attack and Oregon's inability to get much going in 2012
  • I expected OSU to run a lot in 2012. They didnt'. They got me back in 2013 instead.
  • Mariota ran kind of meh in the Fiesta bowl - I said he'd do well
  • Oregon had some major penalty issues against UW in 2013 I didn't think would happen, nor did I think Sankey would run as well as he did
  • I expected Bralon Addison to have a breakout day returning punts against WSU; just didn't happen.
  • We got way more explosive plays against UCLA in 2013 than I thought, and I had thought Hundley would be able to pass the ball (he got like 90 yards). I also thought we'd have to go the distance a lot, but special teams were awesome
  • I heavily overvalued the injury situation for Stanford, I thought we would be able to stop their run and I thought we would have some kind of running game.
  • I didnt' think Arizona would be able to pass the ball at all in 2013, and I thought we wouldn't turn it over that much.
  • Oregon State was not remotely supposed to run the ball. And Mariota was supposed to be good passing the ball.
  • I didn't think the defense was going to do well at all against Texas, much less dominate.
  • I didn't think Oregon would be able to run the ball well against MSU, and thought it might be more of a defensive struggle. Oops.
  • Oregon had a hard time running against WSU and WSU ran the ball well. And Oregon got a ton of 3 and outs when I specifically said that they would get something on every drive
  • I expected that we'd get explosive drives against Arizona and Byron Marshall would have a great game. Oops.
  • Oregon ran way better than I thought they would against UCLA, and i expected UCLA to get much bigger plays.
  • I expected Washington to have some success running. They were pretty much horrible at it.
  • Cal was also supposed to be bad running the ball, and they did well
  • Oregon wasn't supposed to be able to do anything - but especially not run the ball against Stanford. I also expected Stanford to run better than they did.
  • I expected Oregon to have a lot of methodical drives against Utah, but they were mostly explosive. Utah passing was great (I thought it'd suck) and utah running was meh (I thought it'd be dominant)

So what do we make of all of this? Well, the general flow is good. Some details are good and others aren't. Who wins is pretty decent. The spread is surprisingly decent. But there's a lot of specific details that go out of wack, and I should probably invest more time in trying to figure out specific matchups that will make sense and why beyond the advanced stats.