Stats Aren't Helping


Robert
Sep 18, 2018
[]
[]
17 Comments

I wrote a post during the 2011 season - we were 6-1 and ranked #21 at the time (yes, Illinois football, ranked). I had developed all these fears that the 6-0 start was a mirage (and I wrote about it), and then I looked up a bunch of stats and felt even worse. So I wrote a post I called Nerdstats Aren't Helping to talk about how everything pointed towards a disastrous finish. We then lost the rest of our games. The stats were right.

I've now been in a funk since Saturday. Yes, we nearly won, and yes, that's a huge improvement over last year, but I've had that same "mirage" feeling. I hate that feeling - I could have found Illini positivity in the middle of the Slush Fund Scandal - but it's been hanging over me for 48 hours now. We're 2-1, the Big Ten West is a train wreck, meaning wins are available, and I'm... bummed?

So, like 2011, I turned to the stats. And, like 2011, they delivered brutal news. As in, so far we're basically Rutgers. Meaning that if we traveled to Kansas this weekend, we'd likely lose 55-14. I don't want that to be true, and my eyes don't tell me that true, but that's what the statistics say so far.

Most of the stats I follow are S&P+ stats from Bill Connelly. You can find all kinds of stats on his Google Docs page or on Football Outsiders. He'd tell you that it's still too early to look at season statistics, so that's not really where I'm going with this. We're currently "ranked" 99th nationally per the S&P+ statistics, but that ranking won't be very reliable until about halfway through the season. It's scary, but it's only three games so it's not fully reliable yet (see: LSU dropping nine spots after winning on the road at Auburn). Mostly, I just want to look at these first three games and our statistical performance on the field.

And he has a way of doing that: Postgame Win Expectancy. Meaning, if you played this game again, with the same statistics, what are the odds you'd win? Here's how he defines the statistic:

Presented in the team stat profiles, this makes the following statement: "Based on the key stats from this game -- success rate, big plays, field position components, turnovers, etc. -- you could have expected to win it X percent of the time." Luck and randomness play a major role in the game of football, and this is an attempt to look at just how random a given outcome may have been.

Our postgame win expectancy against Ohio State last year? 0%. There aren't enough crazy bounces in the world to overcome that kind of statistical domination. Our postgame win expectancy against Purdue in 2015? 100%. We dominated the game so thoroughly that no amount of luck could help Purdue win that game. Ke'Shawn Vaughn just ran for 75 more yards on Purdue as I was typing that.

The postgame win expectancy this year is... alarming. Against Kent State it was 92% - our domination in the second half was legit. Against Western Illinois it was, let's see here.... 49%.

~record scratches off~

49%? Meaning, statistically, Western Illinois should have won that game? Yes, statistically, with total yards being even + the field position Western had + their big play ability on offense, if we play that same game 100 times, they win 51 and we win 49.

South Florida? Now it gets really ugly: 4%. Meaning, we play THAT game 100 times - same yards, same big plays, same field position, same turnovers - South Florida wins 96 times. That "close" game we saw was us nearly getting once-every-25-games lucky.

I mean, one only had to look at the total yards (626-380) to realize how lucky we were that the game was close, but 4% is fairly stunning. That was a 51-20 South Florida win... that turned into 25-19 because of missed field goals, 15 USF penalties, and long drives for South Florida that ended in zero points time after time after time. They shot themselves in the foot so many times that they almost let us win.

That doesn't help my mood when trying to evaluate where we stand so far this season. You just can't count on teams to shoot themselves in the foot 23 times. You have to put up yards and make big plays (and you have to limit yards and prevent big plays) if you want to win consistently. And we're giving up 6.4 yards per play so far this season, which is 104th out of 130 teams. (We are putting up 5.4 YPP on offense, which isn't bad - 67th nationally.)

So while a game like Saturday might make you think the stats will love us (we hung close with a fringe top-25 team!), we actually dropped 22 spots in the S&P+ rankings because of all the frightening statistics. Which means that after three weeks, statistically, we're 64th out of 65 Power Five teams (with only Rutgers lower than us). That's... not good.

And we also dropped in the S&P+ Expected Wins chart. Connelly's system said last week that our expected win total was 4.9 (offense was better! defense wasn't bad! Big Ten opponents on the schedule are surprisingly weak!). After South Florida, that dropped to 4.0. Basically, that's the statistics saying "yeah, after what we saw South Florida do, this defense probably isn't going to stop anyone from moving the ball at will, so Illinois isn't going to win many games". Giving up 624 yards can just kill a statistical profile this early in the season. You'd need ridiculous turnover luck.

And, uh, we're been wildly lucky there so far. There's a stat called "Adjusted Turnover Margin". Here's the definition:

What a team's turnover margin would have been if it had recovered exactly 50 percent of all the fumbles that occurred in its games, and if the INTs-to-PDs for both teams was equal to the national average, which is generally around 21-22 percent.

If there is a huge difference between TO Margin and Adj. TO Margin (in other words, if fumbles, dropped interceptions, or other lucky/unlucky bounces were the main source of a good/bad TO margin), that suggests that a team's luck was particularly good or bad and might even out either in the next season or in the rest of the current one.

Our expected turnover margin based on those stats? -2.4, 114th in the country.
Our actual turnover margin? +5, 10th in the country.

What does that mean? We've been once-every-ten-years-or-so lucky. I've written about this stat so many times - it's random, and it has nothing to do with skill, and one year Alabama will be 117th and the next year 22nd just because it's an oblong ball that bounces weird - and we finally have a three game span where we're lucky as hell... and we didn't really capitalize (unless you consider Western Illinois as "capitalizing" given that the statistics suggest we should have lost).

Again - and I always emphasize this because it's a polarizing statistic - this is just luck. When Dawson DeGroot blocks a punt (great play) it's completely up to chance whether we get 2 points or 7 points. Completely, 100% on the shape of the ball. If it hits on the point, it probably bounces out of the endzone for a safety. If it hits on the fat part of the ball, it bounces into the air and softly falls in Stanley Green's hands. It's completely up to chance at that point. Remember Terry Hawthorne batting down the pass at Michigan that softly fell into the Michigan receiver's hands for a touchdown to send it to the next overtime? 4 times out of 5 that's an incompletion and Illinois wins. It just so happened that the point of the ball hit Hawthorne's hand a just the right angle. Total luck.

And we've been totally lucky so far. One of the luckiest teams in the country. And we kind of have nothing to show for it. Blurgh.

Shall I get to some good news? Statistically, the Big Ten West looks as bad as it does when using the eye test. Our 4.0 Expected Wins doesn't look the greatest, but here's the expected wins for the rest of the Big Ten:

Ohio State 10.8
Penn State 9.7
Wisconsin 8.5
Indiana 8.2
Michigan 8.1
Iowa 8.1
Michigan State 7.5
Minnesota 7.2
Maryland 4.3
Northwestern 4.3
Illinois 4.0
Nebraska 3.7
Purdue 2.9
Rutgers 2.2

Again, it's still early, and Purdue's offensive performances suggest they'll win way more than 3 games (although that defense is a disaster). It's too early to be saying that these stats "mean" everything. Indiana might tumble. Nebraska might surge. There is a lot that needs to play out. But so far, the stats put us right at 4 wins (which, I think, is what many of us expected from Youngest Team In College Football - The Sequel.

So all is not lost. We need to acknowledge some things.

  • We still don't have the suspended players back. We're putting up these (poor) statistics while shorthanded. Bennett Williams and Nate Hobbs will be a big boost to the defense. Lou Dorsey and Carmoni Green will be a big boost to the offense. Help is (hopefully) on the way.
  • Injured players are hopefully returning soon. Like, you know, the starting quarterback (AJ Bush) and perhaps our best defensive lineman (Jamal Milan). That might improve our chances just a tiny bit. Take away South Florida's starting quarterback and best defensive linemen and I bet their statistics go down 10 spots in every category.
  • The Big Ten West is awful-to-horrific. Has anyone mentioned that? Oh, it's all they're talking about on every college football show, radio, RV, and podcast? Good. Because the Big Ten West is awful-to-horrific.
  • It's still a young team that will hopefully be significantly better in November than in September. We might not have an opponent who will shoot themselves in the foot like South Florida, but we might turn up the pressure and force some mistakes against, say, Nebraska in November.
  • Special Teams stats so far: #4 in the country. The stats say what our eyes see: we probably have the best kicking/punting game in the country. That's the kind of thing that can win every close game. Please, defense, keep 'em close.

In that sense, this should be a statistical low point for the season. We get players back, we play 3-expected-wins Nebraska instead of 9-expected-wins South Florida, and we start moving up in all these categories. I'm still clinging to November being solid.

But for the first three games? We're lucky to not be 1-2 right now, we're coming off what should have been a 51-20 drubbing, and the schedule only gets harder from here. This is not the encouragement I was looking for.

Comments

smadeck on September 18 @ 11:53 AM CDT

I understand that the Illini aren't exactly lighting up the stats department and that clearly this is a bottom feeder in the B1G right now. I am just so confused as to how you can write fairly positively all summer long and now after 3 games into this season and still in your "We won't learn anything" phase how you are SO overwhelmingly negative. I came out of the USF game thinking, "Well thats about as positive as you can be with this team." I generally agree with you, but your posts after the USF game are so negative, AND contradictory to what you've been writing all summer it is hard to read. Especially when referencing a stat that is self proclaimed LUCK. Luck, as you say, is just that luck so why worry about a stat trying to take the Luck out of Luck? The stat and you admit there is no controlling it so take it for what it is....luck.

Robert on September 18 @ 02:40 PM CDT

I bring up luck because I've been writing about it for years, lamenting the fact that we can't ever seem to have a lucky season. It's finally happening... but it wasn't enough for us to start 3-0, and that makes me sad.

As for the "we won't learn anything for 25 games and that runs through the Purdue game - HERE'S WHAT I LEARNED FROM THESE STATISTICS" criticism, that's absolutely a fair point. I'm breaking from my previous statements and forging ahead with a premature stance. I blame standing in the northwest corner of Soldier Field and having my heart ripped out right in front of me with 2:30 to go.

smadeck on September 18 @ 05:44 PM CDT

Fair enough! Look forward to reading more from you! On a side note, I was driving around Lake Tahoe listening to the game and I heard a sideline with Juice....brought me back. Here's to hoping we can get that from Juice Deuce next season.

1970 John on September 18 @ 11:54 AM CDT

Again from someone who doesn't know what he's talking about other than to say that fumble recovery isn't completely luck of the draw. It also depends on how many players you have around the ball, etc. Ditto interceptions. Is it all roll of the dice? Maybe, if you're playing a board game.

Yes, I know that our pass coverage has been lousy, but on the other hand (ha!) we've been very good on tipping balls by linemen. If that were happening against us, we'd be talking about how lousy our quarterback is or something.

I'm not discounting luck...but a well kicked ball is going to bounce your way more often than not. Fortune favors the prepared.

We may be--OK, we are--doing a lot of things poorly, but the turnovers, I'm not leaving that all to luck.

Is that even a little encouragement?

Robert on September 18 @ 02:57 PM CDT

The luck stat doesn't look at the quantity of turnovers. A clean interception is a clean interception (and give me 20 just like Marchese against Western Illinois, please). Force 10 fumbles and you'll recover 5; force only 2 fumbles and you'll recover only 1. 5 > 1.

What it looks at is the turnovers that rely on luck (tipped passes, fumbles) and tracks the departure from the national average. If you only recover 1 of those 10 fumbles, you were just flat-out unlucky.

AHSIllini32 on September 18 @ 01:27 PM CDT

I know TO's have a luck component but when 1) TO's have been a hallmark of your head coaches coaching career and 2) his current program has forced one in 17 straight games I have a hard time saying something along the lines of "our luck will run out".

Yes they'll be a game where we don't force a TO but I'd be shocked if we all of a sudden went through a dry spell.

Robert on September 18 @ 02:35 PM CDT

That's not what this stat is saying. A team can have 6 turnovers and have bad turnover luck or a team can have 28 turnovers and bad turnover luck.

The stat says this:

a. Teams recover right around 50% of all fumbles. Sometimes it bounces your way, other times it bounces their way. Defenses recover more fumbles than offenses (a function of more guys standing around the ballcarrier), but when looking at teams, you look for the departure from 50%. Any variation from that is luck.

I always quote 2009 Oklahoma (at least I think it was 2009). They went to the BCS title game that year with some of the craziest fumble luck you'll ever see. IIRC it was 18 of 20 - they somehow recovered 18 of the 20 fumbles in their games. That's not skill - that's luck. Every coach will tell you that he has the most aware players in history and they'll dive on more fumbles than the opponent, but that's really never the case. Look at 20 seasons and 400 fumbles and most every team will be right around 200 recovered, 200 opponent recovered.

b. Teams tend to intercept around 21-22% of every tipped pass. Any variation from that is luck. If you intercept 38% of all tipped passes in a season, you were wildly lucky. The ball hit the hand at just the right angle that it deflected away at 24 degrees instead of 47 degrees, and because of the 24 degrees, it deflected directly to Jartavius Martin. If it had deflected at 47 degrees, it would have fallen harmlessly to the turf.

So the stat looks at the departure from that 21-22%. If you're grabbing a lot more than that, you're getting lucky. If you're grabbing less than that, you're unlucky. It's simply looking at a departure from the average. Yes, you might tip 23 passes in a season instead of 4, and that increases the odds that one or two of those falls to you, but that's not what this stat is reading. It's saying "of all tipped passes, did you grab more or less than the average?"

Add all of that up and we've been incredibly lucky this season.

alotofpeoplegotoschoolfor7years on September 19 @ 07:23 PM CDT

I understand the rate stat component of your analysis, but it seems the talent and scheme might have a say in the aggregate number of balls that are, in fact, tipped. This is where the skill is, no?

BexleyIllini on September 18 @ 02:40 PM CDT

I realize that the stats don't mean much this early, but I would guess our "expected wins" after 3 games in 2017 was lower than the current number (4.0), maybe close to 2.0?

Robert on September 18 @ 02:52 PM CDT

As I recall, his stats had Western Kentucky as a very solid team last year (no one knew they weren't good yet). So our victory over them gave us a boost. I want to say that after three weeks, we were somewhere in the range of 4 Expected Wins.

illiniranger on September 18 @ 03:41 PM CDT

good read. appreciate the sober by the numbers analysis. it's still early maybe we can find something. DL play needs to pick up steam.

IBFan on September 18 @ 06:15 PM CDT

Stat garbage, results matter.

Dr. Chim Richalds on September 18 @ 08:43 PM CDT

I know it doesn't speak to a lot of fans Robert, but this is a fantastic statistical deep dive. This type of stuff has been proven to be much more predictive than won-loss records across every sport. I think a lot of fans are also loathe to credit anything but the most obvious luck-driven play to luck itself (there's all sorts of cognitive bias research behind that), when in reality it's an enormous factor in game-to-game variation of performance.

Sweetchuck13 on September 19 @ 06:19 AM CDT

I totally agree. I'd love to see this type of "nerdstat" breakdown on a weekly basis, where Robert takes a few interesting pieces and goes deep.

Having said that, I don't think it's time to panic just yet. This year's B1G is awful. We'll get smoked this week, and perhaps Wisconsin. Besides that, we should be able to play with every other team on our schedule, and I like our chances to at least get a couple more wins.

ATOillini on September 18 @ 09:18 PM CDT

“Lies, damned lies and statistics.”

I often agree with that one, but in this case I think the stats you’re referencing are spot on. However, I am hoping the return of suspensions, injuries makes a big difference (at least eventually). When teams like Bama, tOSU lose a guy the replacement is often right there as to talent. Teams like us? Not so fast my friend. We often have a huge gap given the same situation. Maybe wishful thinking on my part, but I’m going to run with it for now.

Norcal Illini on September 19 @ 04:29 PM CDT

This is a great analysis. One thing you didn't mention is that the coaches probably also know these stats. Since we're only three games into the season, it can give them things to improve on. It doesn't mean they have to be this way the entire season.

The lack of a pass rush is a big concern, but I don't remember seeing many blitzes. Is it possible they have been keeping the LBs in pass coverage to help protect the freshmen in the secondary? Or maybe they don't want to show much prior to the B1G season?

IBFan on September 19 @ 07:49 PM CDT

There is so much more to the game than these stats "point out". Im not sure how many former athletes are out there in IB land, but you don't care if your pitcher gives 20 hits a game all year but his era is 2.50. It doesn't matter what could happen, what most likely is to happen, but what happens. Trying to say we should have lost 51-20 or whatever is a gross misunderstanding of the game. I'm butt-hurt as the next guy but cmon man. The defense caused many of those penalties, including the holding that was picked up on the touchdown. And objectively, that was holding. Did the defensive player gain an advantage when he pulled on the shoulder and jersey? Yes, so much so he immediately threw his hands after the ball was thrown as if to say "I wasn't holding". The defense being called almost the entire game was predicated on not getting beat deep, protecting the youngsters. The lineman calls were about contain and occupying blockers so the backers are free to make plays. It is also very noticeable that the dline has/is being taught the effectiveness of a batted ball. Watch the game again, watch the games again, these teams don't win 96%, 51% against this improved squad. No way, you can watch the game and tell that wouldn't happen. Maybe too many nerds doing nerd stats talking about luck, which doesn't exist, simply doesn't exist. End of rant....for now.

Speak Your Mind

Please login or register to post comments on the IlliniBoard.