Nebraska stats - that was then and this is now

[beorach]

Rating our stats in nine categories, relative to national averages and their spread, I evaluated Bo's Huskers' performance up until the bowl game last season and compared it to what Riley's team has done to date this season (only against FBS teams and on a scale of 0 - 1 where 0 is as bad as possible and 1 is as good as possible):

 

Pass "O" - Bo 0.32 to MR 0.57

Pass "D" - Bo 0.86 to MR 0.36

Rush "O" - Bo 1.0 to MR 0.67

Rush "D" - Bo 0.5 to MR 0.83

Scoring "O" - Bo 0.83 to MR 0.83

Scoring "D" - Bo 0.75 to MR 0.33

Total "O" - Bo 0.67 to MR 0.83

Total "D" - Bo 0.58 to MR 0.33

TO Margin - Bo 0.43 to MR 0.46

 

Overall, the ranking is at 0.66 for Bo with a full season and 0.58 for MR to date. There are obviously arguments to be made about schedule strength, the inflation of that 'Rush "D"' rating, given the recent weakness against the pass, talent and commitment levels for each squad, etc., but I just thought some folks might be interested in this given the nature of much of the board discussion.

[Undone]

So, like, what exactly did you do to come up with the 0 - 1 numbers?

[beorach]

I've explained it all before in this and past seasons. The basic idea is to take all the stats I can make sense of within nine categories and assign a rating to each of 0, 0.25, 0.5, 0.75, and 1, based on what range it falls in... All I'm doing is taking averages and standard deviations from cfbstats.com to do this in a spreadsheet.

 

To give a specific example from a simple category, I'll explain Nebraska's rating for rushing offense.

 

Here's the raw data I consider worth tracking:

 

average yards per carry - 5.35

total rushing touchdowns - 10

average yards per game - 195

 

yards per carry national average - 4.28

yards per carry national standard deviation - 0.94

 

Since the value of Nebraska's statistic in this category is greater than the average plus one standard deviation, I give them a 1. If it had only been greater than the average plus half of a standard deviation, I'd give it a 0.75.

 

rushing touchdowns national average - 9

rushing touchdowns national standard deviation - 5

 

Since the value of 10 falls within the range set by the average +/- half a standard deviation, I'm giving this 0.5.

 

yards per game national average - 170.11

yards per game national standard deviation - 55.72

 

Since the value of 195 falls within the range set by the average +/- half a standard deviation, I give this a 0.5 rating as well.

 

For the whole category, I just take the average of these ratings. That's what gives Nebraska, under MR, a 'Rush "O"' rating of 0.67. This is obviously quite simplistic but the method's consistent and gets beyond just comparing basic stats...considering the evaluation of those, based on a range, also factors in the relative statistical significance of those figures.

[kchusker_chris]

So, like, what exactly did you do to come up with the 0 - 1 numbers?

 

Since he has this: Scoring "D" - Bo 0.75 to MR 0.33

 

...when MR is ranked #45 (giving up 24) to Bo's #53 (giving up 26.6) ranking last year....I'd say he pulled them out of his a$$.

[NebraskaShellback]

[MichiganDad3]

Thanks for the work. Statistics help take some of the bias out of how we see the games.

[Moiraine]

Thanks for the work. Statistics help take some of the bias out of how we see the games.

 

Not if the statistics are biased.

[beorach]

 

So, like, what exactly did you do to come up with the 0 - 1 numbers?

 

Since he has this: Scoring "D" - Bo 0.75 to MR 0.33

 

...when MR is ranked #45 (giving up 24) to Bo's #53 (giving up 26.6) ranking last year....I'd say he pulled them out of his a$$.

 

 

That's looking at one stat within one category. For scoring defense, I'm looking at these three: number of touchdowns, number of field goals, and points per game. Nebraska has posted average figures, per my rationale shared above, for the first and third categories...but they've given up 12 field goals (which is higher than the national average plus a full standard deviation). That's where the 1/3 = 0.33 comes from.

 

The bottom line is I'm trying to measure how much a team's performance differs from that of the theoretical, average one...as determined by the statistical performances of all 128 FBS teams in games between them. Weighting the differences, especially the greater ones, allows a person to differentiate between these teams to a greater degree than by just considering the rankings and stats you might find on USA Today's game preview pages, for example.

 

Following the loss to Illinois, I posted my formula's results for the remaining games on the Husker schedule. It predicted a loss at home to Wisconsin and a win in Minneapolis. I shared then that I didn't trust it to bet on alone...and there is absolutely nothing in it that speaks to a team's potential (only measured performance). Calculating spreads again today, it only has us losing to Iowa at home the rest of the way. All that means is only Iowa has posted significantly better overall stats, relative to national averages yadda yadda, such that they outweigh a 3.5-point home advantage. Michigan State is slightly better than Nebraska, statistically, but not to the extent Iowa is... Their turnover margin, relative to Nebraska's, would be enough for me to stay away from betting us.

 

Setting up the first spreadsheet was kind of a PITA (for someone who's not great with such things) but I just copy, paste, and sort raw data now before looking at matchups. I couldn't do it if it wasn't football! Anyway, Northwestern's numbers against the run are just down to average now, thanks to Michigan and Iowa. Their numbers in the Pass "O" category are lower than the national average minus one standard deviation for all but two categories: total number of completions, still low enough to not reach the average range; and yards per attempt, which is higher than the average plus one standard deviation.

 

I'm just going to quit writing now before I irritate anyone further with the contents of my a$$ - . Here's the list, rankings overall statistical performances in those categories I listed above, down to good old Nebraska U (without any SOS consideration):

 

Team overall Houston 0.759 Appalachian State 0.758 Oklahoma 0.755 Baylor 0.751 Michigan 0.745 Wisconsin 0.745 Florida State 0.743 Ohio State 0.743 Iowa 0.731 Boise State 0.731 Alabama 0.729 LSU 0.728 Clemson 0.717 North Carolina State 0.706 Notre Dame 0.698 Stanford 0.694 Toledo 0.690 Duke 0.688 Temple 0.679 North Carolina 0.667 Florida 0.665 Western Kentucky 0.663 Utah 0.655 USC 0.652 Oklahoma State 0.649 TCU 0.642 Utah State 0.638 Mississippi 0.635 Georgia 0.627 Georgia Southern 0.623 Penn State 0.623 Michigan State 0.602 Navy 0.601 Pittsburgh 0.591 UCLA 0.591 West Virginia 0.587 Texas A&M 0.586 Bowling Green 0.582 Memphis 0.582 Nebraska 0.581

[beorach]

 

Thanks for the work. Statistics help take some of the bias out of how we see the games.

 

Not if the statistics are biased.

 

I appreciate it, MD3. I'm not sure how numbers can be biased, though, Moi? As I shared above, the method by which I'm assigning rankings has been consistent going back to last year. If I ever really figure something out, as I've shared before, I probably won't be spending time on message boards...

[Moiraine]

 

 

Thanks for the work. Statistics help take some of the bias out of how we see the games.

 

Not if the statistics are biased.

 

I appreciate it, MD3. I'm not sure how numbers can be biased, though, Moi? As I shared above, the method by which I'm assigning rankings has been consistent going back to last year. If I ever really figure something out, as I've shared before, I probably won't be spending time on message boards...

 

 

Statistics can be biased. I wasn't really speaking about anything you did. It's just a fact. You're not really dealing with statistics (since you have all of the data from the population) until you start attempting to predict things, which is what you just posted about.

 

Anyhow, I don't think field goals allowed should be weighted at all. It's too highly correlated with points allowed. At most you should have TDs and points or FGs and points. Having FGs and TDs and points is redundant. Field goals blocked or missed might be good if there's a team that's especially good at that.

[beorach]

Thanks for the response. I agree to an extent about the field goals but I'm keeping things very simple and that outlier is part of a picture that I'm trying to see. Giving up field goals beats giving up TD's but not counting them at all means you're ignoring the fact that teams have been able to get into range and capitalize to a statistically significant degree.

[Moiraine]

Thanks for the response. I agree to an extent about the field goals but I'm keeping things very simple and that outlier is part of a picture that I'm trying to see. Giving up field goals beats giving up TD's but not counting them at all means you're ignoring the fact that teams have been able to get into range and capitalize to a statistically significant degree.

 

But you are counting them. FGs are included if you only have points allowed and TDs allowed. Having FGs allowed in addition to that is giving it too much weight and over penalizing teams that allow a higher proportion of FGs/points than other teams, even if it's actually a good thing. Which shows in the huge discrepancy between Pelini and Riley in that category. Just including the two variables gives the whole picture (unless you want to start getting picky and think about safeties and points scored by the opposing team's defense).