FYI: ND margins NOT on target...

Easton Pa ND Fan · Nov 16, 2006

ND's margin is averaging 10.8 pnts/game. With a
certainty of 90%, teams with approx 11 pnts/game
margins win 9 out of 12 games.

At 9-1, ND is presently above the prediction line, but
19 pnts/game is the norm for teams that go 11-1. Either
defense steps up or the Irish score more points to win
at USC.

A +90% relationship was found between game margins
and % of games won for 117 team's 1360 games last
season...Tom

<pre>To win: Win by 6 games 01 pts 7 games 04 pts 8 games 08 pts 9 games 11 pts 10 game 14 pts 11 game 19 pts 12 game 33 pts </pre>

tedwick · Nov 16, 2006

wow. someone's having too much fun in stats class : D

Easton Pa ND Fan · Nov 16, 2006

Unbiased grade: A+...

Unbiased grade: A+...

tedwick said:
wow. someone's having too much fun in stats class : D

tedwick,

From just playing with spreadsheet numbers, I have a
bad feeling about ND winning both at USC and at a BCS
bowl.

The high relationship between winning points and winning
percent (of games) does not allow much wiggle room. Hope-
fully, the summarized data might not apply entirely to any
one particular game...Tom

kmoose · Nov 16, 2006

The good new? Army is durn near guaranteed to boost, rather than lower, ND's average margin of victory.

LOVEMYIRISH · Nov 16, 2006

Easton Pa ND Fan said:
tedwick,

From just playing with spreadsheet numbers, I have a
bad feeling about ND winning both at USC and at a BCS
bowl.

The high relationship between winning points and winning
percent (of games) does not allow much wiggle room. Hope-
fully, the summarized data might not apply entirely to any
one particular game...Tom

Yeah, you are spot on man...we should all be worried. Have faith!!! But be cautious...

tedwick · Nov 16, 2006

Easton Pa ND Fan said:
tedwick,

From just playing with spreadsheet numbers, I have a
bad feeling about ND winning both at USC and at a BCS
bowl.

The high relationship between winning points and winning
percent (of games) does not allow much wiggle room. Hope-
fully, the summarized data might not apply entirely to any
one particular game...Tom

if we're playing statisticians, i think i may have good news on a couple fronts.
first: there's a difference between correlation and causality. there may be a relationship between points/margin, but that does not cause the team to have a specific record. not sure about the "wiggle room" you're talking about, but statistically, ND doesn't have to have a certain average margin of victory to be 11-1. Whether Army boosts margin of victory or not has no relevance to what ND's record is, as long as it's >0.
second, and probably less obvious and more important: statistics have predictive power in a different sense than the way we're utilizing them. The statistics compiled here allow me to, after the season is over, look at a team and it's average margin of victory and predict what its record is. what this statistic is analyzing is the outcome of the season, not the outcome of a game. It doesn't work the way you're trying to use the stats because we're looking the likelihood that a team wins two specific games, two independent events if you will. What you would want there is a probability of a team winning a single game given their margin of victory. assuming each of the games in a 12 game schedule is an independent event with an equal probability, a team that goes 9/12 (i.e. statistically correlated to 10.8 point margin per game) on a season has a .75 probability of winning a single game. Now, this is clearly a flawed way of looking at things because games aren't of equal probability and there's no really accurate way of predicting the outcome of a specific game. The thing with statistics is that, if we're assuming that each game is an independent event, the outcome of the previous games aren't particularly relevant. i.e. we can't say that, since ND has a specific margin of victory right now, they're destined to be a team with a specific record in the end. the games are independent of wins and losses. I suppose if you found a statistic with margin of victory correlated with probability of winning a single game, you could make a more meaningful prediction.

dylan020 · Nov 16, 2006

ugh, i tried to read that and my eyes had to squint halfway through....do you now predict im chinese??

sorry, good post ted but i think i've just studied way too much and my eyes cant comprehend, hell, nor my mind for that matter

tedwick · Nov 16, 2006

dylan020 said:
sorry, good post ted but i think i've just studied way too much and my eyes cant comprehend, hell, nor my mind for that matter

heh. all studied out? i'm just getting started. exam tomorrow, and i've got 4 or 5 chapters in my text to read. whoooo procrastination...

kjones · Nov 16, 2006

Tedwick is right on in his analysis of statistics, which many many people use incorrectly. Using past statistics to predict the future is usually a dicey business, even with things with far far less independent variables than a football game, but might I also add the danger of relying too much on mere numbers. There IS a reason we actually PLAY the games also.

tedwick · Nov 16, 2006

kjones said:
Tedwick is right on in his analysis of statistics, which many many people use incorrectly. Using past statistics to predict the future is usually a dicey business, even with things with far far less independent variables than a football game, but might I also add the danger of relying too much on mere numbers. There IS a reason we actually PLAY the games also.

well, i mean, using statistics to predict the future is generally ok, so long as you're using them correctly and you are prepared to be wrong occasionally. Predictive statistics are probably more useful (and definitely more interesting) than descriptive statistics.

On a further note, something I just realized is that the stats put forth here are descriptive and not predictive. i.e. i can look at them and say "wow. the teams that have a larger MOV typically win more games." however, I can't say "since this team has a specific MOV, they will probably win (or lose) this specific game."

However, you are dead on with the "football games are very complex events" bit (using "event" in a statistical sense). The error associated with using compiled statistics to predict outcomes of individual games is tremendous, and it's near-impossible to make such a prediction simply because you can't quantify the enormous majority of factors that play into the outcome of a football game.

kjones · Nov 16, 2006

Well, what I was really meaning was that using the sort of statistics commonly used by sports-casters and the like to predict outcomes is pretty haphazard and generally bad science. Things such as "he's 1 for 6 during night games when the wheather is below 50 degrees and its November." Many sports stats are really just descriptive, like you said, and not really useful at all when saying what is going to happen THIS time. Stats as a whole are very useful (working on my PhD in Engineering), but people often use them inappropriately.

Clotho · Nov 16, 2006

What are SC's margins of victory, and how do those factor in?

Here they are-

-36
-18
-17
-6
-6
-7
+2
-42
-25

-155 overall in 9 games

Avg margin of 17.2 PPG

ND this year
-4
-24
+26
-3
-14
-21
-3
-24
-19
-22

-108 in 10 games

Avg margin of 10.8 PPG

Now! Here is ND last year prior to the BCS bowl

Avg margin of 14.5 PPG

Better yet, here's ND last year prior to Stanford

Avg margin of 15.3

How much did that mean when we could only beat mediocre Stanford by a late TD?

You can't use statistical analysis to predict a ballgame this way- only useful in looking back over a season gone past and seeking patterns, and seeing what was meaningful. There are garbage time TDs, there are fluke plays (both for and against a team), there are things like momentum shifts and "catching a team at the right/wrong" time. There are, of course, quality of opponents. Does SC's huge avg margin of victory consider that their streak of close wins, and one loss, was against mediocre teams? Or that their blowout over Arkansas was against an essentially McFaddenless offense? Does our margin of victory factor in that Michigan State was playing lights out and full of confidence, and that our team was still reeling from the Michigan loss? Or that the UNC game was never competitive, and those were garbage time scores? Ditto with Navy, Purdue, and a few other games. Does it factor in Charlie's unwillingness to pour it on, so that many games end with Aldridge running into the pile play after play, not even trying to score? And so on. I'm not sure how much these stats apply to this team, because Weis is not a typical college coach. He simply refuses to play to impress pollsters and score superfluous touchdowns that his team surely could. That margin of victory could be much higher than it is, if what he cared about was calming those fans of his team who happen to double as statisticians.

Clotho · Nov 16, 2006

Or, in other words- the Chicago Bears had been blowing people out by an avg of something like 24 PPG over the first 7 games, including an average margin of over 30 at home. Was there any possible way, using statistics, to predict that they would lose, by 18 points, to 1-6 Miami, at home?

jerseyborn1971 · Nov 16, 2006

If we went by statistics ND's record this year would be considerably worse. What were the odds ND would come back against MSU? What were the odds ND would come back against UCLA? What were the odds ND would come back against GT?

I'll go with blind faith when it comes to the Irish. Plus, I'm not smart enough to decipher all of your fancy numbering and all.

tedwick · Nov 16, 2006

Clotho said:
You can't use statistical analysis to predict a ballgame this way- only useful in looking back over a season gone past and seeking patterns, and seeing what was meaningful. There are garbage time TDs, there are fluke plays (both for and against a team), there are things like momentum shifts and "catching a team at the right/wrong" time. There are, of course, quality of opponents. Does SC's huge avg margin of victory consider that their streak of close wins, and one loss, was against mediocre teams? Or that their blowout over Arkansas was against an essentially McFaddenless offense? Does our margin of victory factor in that Michigan State was playing lights out and full of confidence, and that our team was still reeling from the Michigan loss? Or that the UNC game was never competitive, and those were garbage time scores? Ditto with Navy, Purdue, and a few other games. Does it factor in Charlie's unwillingness to pour it on, so that many games end with Aldridge running into the pile play after play, not even trying to score? And so on. I'm not sure how much these stats apply to this team, because Weis is not a typical college coach. He simply refuses to play to impress pollsters and score superfluous touchdowns that his team surely could. That margin of victory could be much higher than it is, if what he cared about was calming those fans of his team who happen to double as statisticians.

Well, in theory, you could make a prediction based on MOV if the correlation between a high margin of victory and winning a specific game (an independent event, in stat speak) was high enough. I think you may be confusing descriptive with predictive statistics. The nice thing about predictive statistics is that there's an error built in simply because it assumes that things are part of a distribution. Statistics ASSUMES that there are going to be teams that have a high MOV and lose and teams with a low MOV that win. It just looks at how likely that is. So it accounts for what you say about ND. We could use an appropriate statistic in this way.

The problem that came up in this thread (and i'm still struggling to explain it in a way that makes it easy to understand) is twofold.
1st, the statistic put forth was descriptive, not predictive in the scope of a single game, which is what we were looking at. If, at the beginning of the season, you told me that ND would have a MOV of 10.8, I could tell you that they'd probably end up with a record of 9-3 based on past events. However, the thing is that now that games are played and we're looking at individual events, that statistic is nothing more than an "oh, isn't that interesting" tidbit. It doesn't have any predictive value for what we're looking at, the outcome of a game.
2nd, and related to the first, is that the statistic is being misused. People were (and still are) using the stat as a deterministic model rather than a predictive one. It was "Oh, ND should be a team with a 9-3 record, therefore they're likely to lose the next 2 games". Take this example: You flip a coin 10 times, and it comes up heads 8 times. Since, on average, the coin should have a p(H)=.5, is the coin more likely to come up tails on the 11th flip? no. Same thing here.

njuneardave · Nov 16, 2006

right.

your initial table of stats is pretty. however, you failed to incorporate level of competition. Texas has a huge MOV..... against teams like Baylor and Sam Houston State. MOV is pointless without considering the competition and actually watching a game to see how close it is (for example: ND vs OSU fiesta bowl.... OSU had overwhelming stats but the game was close until that last few minutes sealed by the Pittman 500 yard touchdown run)

FleaFlicker · Nov 16, 2006

Yeah, it is flawed statistics...

As tedwick said, but in simpler language...

If a team has a 50% chance of winning every game in a 10 schedule season, if they win the first 9 games, it doesn't make the chances of winning that 10th game minute (like 5%). The chances are still 50% for that game. It was just that when looking at the whole season, the chances of Notre Dame going 10-0 overall are like less than 5%

FYI: ND margins NOT on target...

Easton Pa ND Fan

New member

tedwick

GBL WRMNG ISA LIE

Easton Pa ND Fan

New member

kmoose

Banned

LOVEMYIRISH

old timer

tedwick

GBL WRMNG ISA LIE

dylan020

the next 007

tedwick

GBL WRMNG ISA LIE

kjones

Zahm Hall Football Coach

tedwick

GBL WRMNG ISA LIE

kjones

Zahm Hall Football Coach

Clotho

Banned

Clotho

Banned

jerseyborn1971

Guest

tedwick

GBL WRMNG ISA LIE

njuneardave

Member

FleaFlicker

Guest