somehow we're only 124th
|7 weeks 6 days ago||The point isn't that TOP is||
The point isn't that TOP is an offensive or defensive metric, obviously both contribute. The issue is that it has no correlation to winning and any loose connections between the two are because winning yields TOP, not vice versa.
|15 weeks 17 hours ago||I've done that in the past,||
I've done that in the past, for this one with a 30+ point spread, the line would have been at 100% the entire game.
|16 weeks 1 day ago||Notre Dame was in fact #10 in||
Notre Dame was in fact #10 in 2010.
|16 weeks 1 day ago||Yes, they were 9th last||
Yes, they were 9th last season.
|16 weeks 1 day ago||It's two different pieces. I||
It's two different pieces. I am picking Michigan to win 8 games, but based on the National Championship Secret Sauce article, each of the last 10 national champs were rated in the top 10 of Roster Talent and Michigan is #10 this year in that metric.
|45 weeks 1 day ago||Yes, Michigan starts much||
Yes, Michigan starts much higher and is close but slightly higher throughout
|1 year 4 weeks ago||4th Down Odds||
A couple points of clarification
The official PBP lists the play as 4th and 1, not 4th and 2
For those saying that the odds aren't about this team, the odds on any one single play aren't that different from the best to the worst. Do they differ, absolutely, but not that much.
Yes Michigan hadn't converted a third down prior to the play, 7 of the 10 failures where on plays of 8+ yards, not a lot of relavance.
Michigan has converted 58% of 3rd/4th and 1 this season, right on the expected average.
No team in the last 11 season has ever average less than 35% conversions on 1 yard to go situations. For Michigan to be at 30% expectation would mean that they were a standard deviation worse than the worst team in this situation of the last 11 years. And even then you are break even. I know if felt bad and that the play call was awful, but the numbers were firmly on Michigan's side unless you think this is actually the worst offense of all time, and then it was a break even decision.
|1 year 11 weeks ago||The pictures I found and even||
The pictures I found and even Google Street view show a yellow trim around the top of the stadium
|1 year 11 weeks ago||I added some contact info at||
I added some contact info at the bottom if you are interested
|1 year 14 weeks ago||I watched the game, I know||
I watched the game, I know what was happening but with MSU Defense and the USF offense, but come on. Best case scenario after that punt is to get the ball back where you had it, down 15 with a minute and a half to go. In a one possession game you can justify a punt but not when you need two possessions punting is game over.
|1 year 14 weeks ago||Thanks for the heads up. Poor||
Thanks for the heads up. Poor wording on my part, fixed it.
|1 year 14 weeks ago||There is some slight random||
There is some slight random variation but the fumble rates are pretty consistent year after year, regardless of age.
|1 year 14 weeks ago||Great question Michigan was||
Michigan was at 93% going into the play:
Incompletion: still about 93% (it was 3rd and 11, first down was unlikely)
Safety: only drops slightly to 92%, ND field position about the same as a punt, still need two TDs
Reality: dropped to 82%
The safety would have seemed devastating but in reality the interception was the only play that would have really swung the odds.
|1 year 14 weeks ago||It makes predicting a||
It makes predicting a specific game very difficult, but over many games the pluses and minuses even out. Even in one specific game there is a pretty decent chance that swing plays are neutral, it's just that we tend to remember the ones were they don't.
|1 year 14 weeks ago||The flip side is that from my||
The flip side is that from my work, teams with great defenses should go for it more, not less. They are probably going to stop the other team either way, so take advantage of a potential short field for the offense and go for it. When you are on the opponent's side of the field, trading punts is almost always a net negative for the team in opp. territory.
|1 year 16 weeks ago||I think they are going to be||
I think they are going to be about as good of a team as last year, but they trade Indiana and Penn State for Wisconsin and Ohio State. Plus, they weren't an extremely dominant 9-3 team last year, Illinois was the only FBS program they beat by more than 2 TD's. If you just adjust for schedule and assume everything else is the same, that takes them from 9 to 7. If they were a little lucky that gets you to six (about where I have them) and having the same team they did in 2012, which is about what I am predicting, quality-wise.
|1 year 16 weeks ago||I agree that red zone only||
I agree that red zone only tells part of the story. It's a stat designed to balance out total yards. Moving the ball up and down the field and scoring in the red zone can be two very different things, but I don't think it's a deceiving stat. I think most people take it for what it is. People don't look at rushing stats and say that they mean everything and neglect passing stats. Like most stats, red zone efficiency is a measurement of a team's success in a specific facet of the game, I hope I didn't give the impression that this was any more than that.
|1 year 16 weeks ago||In this study and most of||
In this study and most of them I do, I include all plays for the first half of the game but only plays in the second half if the drive starts or ends within two touchdowns. Depending on defintions could tweak the When It Mattered sentiment but this should be about what you were looking for.
|1 year 21 weeks ago||Rodgers and Newton did after||
Rodgers and Newton did after starting at Jucos. Mallet did it his first season at Arkansas. Leinart did it as USC and Tebow did his first year as a starter but saw lots of time the prior year. I think that's it for the major conference guys.
|1 year 22 weeks ago||Another MgoBlogger in Topeka?||
Another MgoBlogger in Topeka? I thought I was the only one.
|1 year 25 weeks ago||Henne||
A quick note on Henne. First, he was right below the cut off at -0.4 for his freshman season. I had him ranked at 42nd for the season and he finished tied for 19th in passer rating. There is a gap there but not a huge one. I think to understand the gap some, his opponent adjusted numbers are on par with similar quarterbacks, the gap isn't due to opponent adjustments. I think the big item that makes Henne seem worse than a quick gut check is that he wasn't great on third downs. His first and second down rating match up with his passer rating but he ranked 60th in third down PAN for the season.
|1 year 28 weeks ago||Fixed, thanks.||
|1 year 30 weeks ago||Corrected, thanks.||
|1 year 40 weeks ago||The rating is an average of||
The rating is an average of all the service's ratings. That's a good point that this could be some of the difference at both the top and the bottom for Hoke vs Carr, with Hoke working in a 4 service environment vs 2 for most of Carr's tenure.
|1 year 44 weeks ago||It's really a quantity vs||
It's really a quantity vs quality thing. The aggregate points I used to rank has a strong bias towards signing more players. The guys Michigan got were very good, I believe Michigan was top 10 in LB avg rating, there were just two of them which kep the overall number down.
|1 year 49 weeks ago||The numbers are developed||
The numbers are developed based on the last ten years of games between evenly matched FBS teams, accounting for down, distance,possession, timeouts and score.
|2 years 4 weeks ago||Five games from the last 10||
Five games from the last 10 years above 80% on defense and lost:
2005 Nebraska 92%
Leading by 11 with 10 minutes left
2005 Ohio St 88%
Leading by 9 with 8 minutes left
2009 Purdue 86%
Up 13 in the third quarter
2004 Texas 84%
Leading by 10 to start the fourth quarter
2008 Purdue 84%
Up 14 in the second quarter
|2 years 4 weeks ago||You're about right on for the||
You're about right on for the 10% number but the first offensive play was run from the 46 after a long Breaston return, increasing Michigan's chances to about 28%.
|2 years 4 weeks ago||Like I wrote, in the final||
Like I wrote, in the final seconds things get a bit hazy. The 9% vs 5% has to do with the way I am counting. The 9% was the lowest when an offensive play was run. That was used for easier tracking in the database. It was 5% at the time of the punt, but 9% was the lowest when an offensive play was run.
The 9% does seem a bit high but I looked at the 100 games with the most similar situations, trailing by 3 ball around own forty and about 18 seconds left and 6% of teams went on to win the games so it isn't dramatically over stated.
|2 years 5 weeks ago||All of their damage is being||
All of their damage is being done in the first half and they are really pulling off early, but just able to get our in front really quickly.
|2 years 5 weeks ago||Not saying it can't be done||
Not saying it can't be done before then, but just that its unlikely to, at least before 2014. There will always be outliers like KSU but planning on being an outlier is not a very good strategy.
|2 years 5 weeks ago||There is a time lag. The||
There is a time lag. The upper class players on the late Carr years were outstanding, some of the best in the country. The players he recruited those years that were upper classmen under Rodirguez, not so much.
|2 years 5 weeks ago||There is a points metric||
There is a points metric behind it but you can think of the top as best in country and the bottom as worst.
|2 years 5 weeks ago||Fixed, thanks||
|2 years 6 weeks ago||Think of the replacement||
Think of the replacement player concept, but localize it to one team. If Denard wasn't getting the plays, would we be better off? Compare Denard to AJ McCarron. Denard is +.23 and the rest of the team is -.23. If we take away plays from him the net result is -.46/play. McCarron is at 0.36 but his team is at 0.22. Even though McCarron's per play average is higher, he has more support behind him would still mean there is a good chance of similar production. Michigan doesn't have that luxury. Denard isn't the most productive this year, but losing him would be a bigger blow than any other team losing their best player.
|2 years 6 weeks ago||Despite being only 8 games||
Despite being only 8 games in, Denard is already more valuable than any previos year. In 2010 Denard was worth 99 points and last year that dropped to 48 points with a strong supporting cast. This year's weak non-Denard showing has eclipsed prior. Doesn't mean he wasn't used more previously, just that this year the difference between his production per play and everyone else's is substantially greater.
|2 years 8 weeks ago||I value 100% of my readers||
I value 100% of my readers even the blacked out ones playing Halo 3. I updated the heading to clear things up a bit.
|2 years 8 weeks ago||The issue isn't having the||
The issue isn't having the plays or the power to crunch them, it is a decision not to use them. As the chart above notes, Michigan was already 90+% chance of winning at the start of the third quarter with a three score lead, and it quickly climbed over 95%. Yes the starters stayed in the game but in my opinion, when the score reaches three possessions or more in the second, at the very least the strategy shifts from the teams' perspectives. Some positions may be subbed more liberally, play calling will likely shift down to a more conservative approach for the team with the lead.
|2 years 8 weeks ago||The numbers are all based||
The numbers are all based from Michigan's perspective so that rush defense heading was referring to Michigan's rush defense versus MSU's rush offense.
|2 years 9 weeks ago||Most of the perceived||
Most of the perceived variance is based on the fact the due to the lead, the second half is not included in these numbers. First half saw 6/8 passing and 4 big conversions while rushing had a decent 27 carries for 118 yards and 4 conversions. Second half the rushing really dominated but was excluded due to the size of the lead.
|2 years 12 weeks ago||Pulled the old database up||
Pulled the old database up and Brian Luke from Kansas gets the worst showing at -28 in a 2005 game against Oklahoma. 11/30 86 yards, 3 INT and -40 yards rushing
Juice Williams narrowly edges out Henne for worst B1G game, with a -20 in 2008 against Penn St
|2 years 13 weeks ago||+20 EV is one of the top 300||
+20 EV is one of the top 300 performances (not opponent adjusted) of the last ten years. It's only 6th for Denard overall, his best being last year against NW when he was +28, which is top 40.
On a per play basis it was his seventh best with his best being 2010 against Indiana when he went +27 on 35 plays.
|2 years 13 weeks ago||I am mostly using preseason||
I am mostly using preseason expectations, adjusted for teams that have largely under or overperformed such as Wisconsin. Usually by week 4-5 there are enough games to start to get a picture of what the true in-sesason perforamnce looks like.
|2 years 13 weeks ago||Extra Value. Essentially how||
Extra Value. Essentially how many points above average performance. A 0 is an average performance across all of the FBS. Each point above is a measure of how many extra points a player/team has performed.
|2 years 13 weeks ago||Great description, drive||
Great description, drive seasonality. I also like the saw tooth of the Air Force drive with many third and fourth downs and so the odds fluxuate down by down.
|2 years 14 weeks ago||Players are rated based on||
Players are rated based on their most recent evaluation. If a player was a 5 star out of HS and then 4 star out of JC, the 4 star is the one used for that service. If the service doesn't rate them for JC then the HS is used.
The talent rank is a baseline rating using only data before entering school. I have lots of ways to look at players once they arrive, but this is measuring just prior evaluations. Since no adjustments are made to talent rank based on post-arrival performance, the Jordan Kovacs' of the world are still non-counters in this measurement.
|2 years 14 weeks ago||Post FAIL||
Completely left off the B1G's Indiana representation on the table. Has now been updated.
|2 years 16 weeks ago||The 85% is based on receiving||
The 85% is based on receiving production from receivers/tight ends only (backs excluded). Bell's return is factored into the equation separately but losing essentially an entire receiving corps and a quality quarterback, independent of everything else is not a combination that has a great history of offensive success. I would love to factor in offensive line but at this point I haven't been able to determine a quantitiative method with sufficient history to include.
|2 years 16 weeks ago||Junk time is tossed. All||
Junk time is tossed. All plays in the first half are in and plays in the second half are in if the drives either begin or end within two TDs.
|2 years 16 weeks ago||Since 2006 13 BCS teams have||
Since 2006 13 BCS teams have lost their starting QB and over 85% of their offense. Two teams (Ole Miss 2006 and Ks St 2010) improved, 3 stayed about the same and 8 teams cratered by a substantial margin. MSU could be an exception this year but history is not on their side.
|2 years 16 weeks ago||My Top10: 1.Oklahoma 2.Oregon||
My Top10: 1.Oklahoma 2.Oregon 3.Texas 4.Alabama 5.Georgia 6.USC 7.Ohio 8.Michigan 9.ND 10.LSU
|2 years 18 weeks ago||I use (and typically lose)||
I use (and typically lose) using my data set that drives the EV calculations but one of the great things about fantasy football is its disconnect between easily measurable and valuable.
|2 years 21 weeks ago||Morris probably doesn't have||
Morris probably doesn't have a shot at passing Mallett. Shane's best overalll rank (19) was Mallet's worst. Mallet was 19/5/12 to Rivals/Scout/ESPN and Morris is currently 22/29/26/19 with 247 in the mix. Shane would have to pass half the players currently in front of him and have no one leap from him just to tie Mallett. For as highly of a scouted player as he is, that doesn't seem likely.
|2 years 23 weeks ago||Saw the Math signal in the||
Saw the Math signal in the air and got here as quickly as I could.
Don't think I have done anything specifically on time of possession (but agree with the general conclusion). You might be thinking of my Myths of Manball post where I dispell the myth that long drives rest a defense.
Other than that you must be thinking of someone else
|2 years 23 weeks ago||I almost agree with your||
I almost agree with your reasoning. In general, risk tradeoffs are easier for those attempting to rise than those already at the top. A failed fourth down attempt could certainly cost a top team a game and that game could cost them a shot at a championship. The point is that the strategy is more likely to win them a game than lose one. The risk is there and thats why this is an outsider strategy and not the norm. A more aggressive strategy does open up more risks but the point is that it opens up even more opportunities, no matter where your team falls in the rankings.
|2 years 23 weeks ago||Off Adv = Your Off- Their||
Off Adv = Your Off- Their Def
Def Adv = Your Def - Their Off
|2 years 27 weeks ago||Yes, field position is fully||
Yes, field position is fully controlled for. A baseline is established at each spot and the field and the outcome at that spot is compared only against the baseline from the same position.
|2 years 27 weeks ago||That too, he had the audicity||
That too, he had the audicity to burn the letter I sent trying to recruit him.
|2 years 30 weeks ago||You got it. Top teams move||
You got it. Top teams move up, bottom teams move down. In today's world, the levels would be D1-1AA-D2-D3-NAIA, here there are up to 12 levels per conference.
|2 years 32 weeks ago||247 only began with a partial||
247 only began with a partial 2010 class and then a full 2011, not a lot to evaluate them on yet.
|2 years 34 weeks ago||Something I derived. ESPN||
Something I derived. ESPN ranked the top 150, which included the top 14 DT's. That left 9 unranked four star DT's of which Pipkins was the second highest ranked. Those 9 players were evenly allocated among the rankings between #151 and #249 (there were 249 4 or 5 stars). The remaining ranked 3 stars were spread between #250 and #1196 based on their position ranking.
|2 years 34 weeks ago||In terms of incentives to||
In terms of incentives to give more stars to stack the deck, that probably wasn't clear enough in my explanation. Everything is based on each site's national player ranking. After the Top X for a site are listed, I take the remaining four stars and rank them based on position ranks and then take the three stars and do the same. Each service has each of its players (sometimes as many as 1500) with a position rank, given a national rating. For example last year Willie Henry was ranked #484 on Scout (#38 DT), 1008 on 247 (#75 DT) and 1153 on ESPN (#97 DT). With no position rank from Rivals he was considered unranked. Even though Ondre Pipkins didn't make the ESPN Top150, he was ranked 166 because he was the second highest rated DT that didn't make the Top 150. Hope this helps clear it up.
|2 years 34 weeks ago||The crux of this method is||
The crux of this method is "How highly did you rate the successes?" The higher you rated the successes, the better your service. I am looking at 1000+ players per service per class so the pool is substantially large.
|2 years 34 weeks ago||Since this was based mostly||
Since this was based mostly on all-conference ND is laregely absent. I did check the average ranking for each class and Scout is right in line with Rivals for ND recruits over the last 11 classes. Nothing stands out for either one. It's possible they are both overestimating, but they are generally in line with each other.
|2 years 36 weeks ago||Beyond the obvious reasons||
Beyond the obvious reasons for avoiding, I don't track games against FCS teams.
|2 years 37 weeks ago||I agree on the importance of||
I agree on the importance of the offensive line but I haven't been able to find a quantifiable metric that correlates offensive line to offensive success. Not saying that it doesn't exist, just that if there is no way to quantify it and measure its affect on the variance then there is no effective way to include it.
|2 years 38 weeks ago||For 2003 its a bit rough and||
For 2003 its a bit rough and you might be right. I only have recruiting rankings back to 2002 so most of the upperclassmen weren't counted in that total.
|2 years 38 weeks ago||D'oh. Thought there were 12.||
D'oh. Thought there were 12.
|2 years 38 weeks ago||Trying to figure out what I||
Trying to figure out what I am missing on the scholarship situation. How can we fit two more in the 2012 class and still get our 3 in 2013. Is Brian assuming Biefeldt slides over to a walk-on?
I have a 2013 Roster of
SR: Hardaway, Morgan
JR: Burke, Horford
SO: Stauskas, Robinson, McGary, Biefeldt
That's 8, two more in 2012 is 10 and not enough room for 3 in 2013. What am I missing?
|2 years 39 weeks ago||Pant size||
|2 years 39 weeks ago||You're spot on. That was a||
You're spot on. That was a pretty lazy statement I made. Miles success may have changed but to say that it was an Ok St issue is counter to the data.
|2 years 39 weeks ago||Time will tell how Miles||
Time will tell how Miles turns out but because of the time from 3/5 years for him were at Ok State not LSU.
|2 years 40 weeks ago||This is just one piece in a||
This is just one piece in a larger database. I have nfl draft and early entries tracked and academy guys are listed for all signing periods since they can ign with multiple schools.
|2 years 40 weeks ago||Just looking for the FBS||
Just looking for the FBS portions. The JC years are essentially left blank.
|2 years 40 weeks ago||Just for football. Kenpom is||
Just for football. Kenpom is the king for BBall don't feel like there is a lot to add there.
|2 years 40 weeks ago||I am working on it. Only||
I am working on it. Only Scout and Rivals have data far enough back to do a good analysis with the NFL draft because of the lag time. Getting it all uploaded and managed is a bear. For any of you NOTY fans, you can understand how hard it is to link up all the creative spellings and name changes (like Maruce (Jones-)Drew or Patrick Johnson (Peterson). At some point I am hoping to get somewhere with it but don't count on it any time soon.
|2 years 41 weeks ago||None of the 5 star player||
None of the 5 star player counts are that high, but guards were 3/6 when it came to 5 stars getting drafted and tackles were 2/10. It's not just a more tackles are rated question. But there is definitely uncertainty tied to the small sample size. I didn't bring it out as an issue in the article because even with the small sample size the results were very consistent.
|2 years 43 weeks ago||Never played a down in my||
Never played a down in my life. Skinny and slow from a young age! My dad was a basketball coach so I didn't fall in love with football until college were I ended up spending five years on the support/coaching staff spending every day with the team and coaches at an NAIA school.
|2 years 48 weeks ago||If you note the y-axis' the||
If you note the y-axis' the only time recruiting is compared against team success is on the overall team recruiting totals. Offense, defense and position recruiting are only compared against offense or defense success. I agree with your point that they are tied, but I am not comparing OL recruiting to team success, just offensive.
|2 years 49 weeks ago||This will all be tied into my||
This will all be tied into my play by play databse which has everything else. These coaches will be tied into team seasons for W/L and team performance.
|2 years 50 weeks ago||Correct, from 2007 to the||
Correct, from 2007 to the 2011 regular season end, there were 37 OT's where the first team in an overtime period failed to score. Only three times did that team survive to see another period, Arkansas Ms St last year, Fresno St Wyoming in 2009 and Tennessee Kentucky in 2007. That's an 8% chance of surviving to the next period plus about 50/50 to win in the next.
|2 years 50 weeks ago||Forgot my sarcasm font on||
Forgot my sarcasm font on that one. As the chart above shows, that area yards are critical to improving your odds. Wouldn't take a coin flip shot at winning without trying to something more.
All of the Sparty 2 pt attempts were questionable. The first two were way to early but at least the argument could be made at that point that MSU offense was stalled and you have to try, even if its still the third quarter. By the time the try in the fourth came MSU had started moving the ball on offense. Even though time was getting tighter, there were several possessions left and obviously scenarios where the foregone points would have cost them. If Sparty kicks to go up 21-19, the Georgia TD gives them 25-27 depending on what they decide to do for PAT (I would kick, go up 5), but no matter what, the final MSU TD seals the game. With 14 seconds left I might even take a knee instead of try a point (assuming you were trailing by 4 or 5, not 6), just to avoid what happened to Nebraska.
|2 years 52 weeks ago||Since 2003 I can't find a||
Since 2003 I can't find a team that has intentionally done it. Three times a team has either faked it or botched a snap and run the fire play, but no straight up attempts.
|3 years 15 hours ago||I agree, take them as soon as||
I agree, take them as soon as you can. Taking them on 1st and 2nd down instead of 2nd and 3rd guarantees you get the full value.
|3 years 15 hours ago||1. This is some weird coach||
1. This is some weird coach deal where there are certain places that you go for the home run. Don't really have anything to support it or reject it.
2. Brian put my thoughts about this in a mailbag at one point. From the 1 or 2 coaches should be very conservative, once you get even a couple yards out, return to play book as normal.
|3 years 16 hours ago||Delete||
|3 years 16 hours ago||I wouldn't be surprised if it||
I wouldn't be surprised if it was gut. The math helps quanitfy and in some cases clarify but the problem with coaches making dumb decisions is that they are often working off an old situational framework (3 yards and a cloud of dust era) that no longer exists. The framework used to be true but the game has changed, forcing decisions to change. Some coaches have evolved with the game and can trust their gut, others are still in a bygone era. A few are just idiots.
|3 years 16 hours ago||While I agree with you in||
While I agree with you in theory, the 60% is too high of a number. The difference between good, average and bad on a specific play is much less than you think. Teams with a significant offensive advantage still only score on a third or fourth and goal from the 3 a hair over 50%. If you could get to 60% I agree that an aggressive always go two point strategy could be effective, just don't think 60% is possible.
|3 years 16 hours ago||Great question. Have thought||
Great question. Have thought the same thing a time or two. I quickly ran the numbers on teams that are trailing and start a possession with 2-4 minutes left in the game. Teams trailing by 3 (n=161) go on to win 21% of the time. Teams trailing by 4-6 (n=253) go on to win 25% of the time, a 20% improvement. While I can't dig to the specifics, it certainly appears that settling for the field goal trailing by 3 is a real concern.
|3 years 18 hours ago||At the end of the first||
At the end of the first quarter, the team that just had the wind is likely to be in a positive field position situation. If its on offense, they have some time in the second quarter to continue that drive or possibly punt from a better spot on the field. If they are defense, the other team is likely to be backed up and will have to either punt deep in their own territory or spend the first part of the second quarter moving into a more neutral field position.
|3 years 2 weeks ago||I probably didn't make this||
I probably didn't make this clear enough when I wrote up the article but the defensive numbers are season cumulative, where the offensive numbers are per game averages. Plus the defensive players are just noted for their good plays, I don't have any way to assign blame for the bad plays based on PBP data. A tackle after a 20 yard gain may be saving a TD or it could be covering a blown assignment.
|3 years 2 weeks ago||I upload every game's play by||
I upload every game's play by play into a database and push a couple of magic buttons.
|3 years 3 weeks ago||The breakeven value is a||
The breakeven value is a return that you gets you to the red zone. An average return of a post-safety free kick is to the 40, worth 2.28. That plus the 2 points is 4.48 points total. An interception giving you a first and ten at the opponent 20 is worth 4.5 points on average. Not a very likely scenario but a good return could justify giving back the 2 points.
One caveat is that this is for average offenses. The better your offense is, the less sense it makes to give back the points, because you can probably score anyway. For it to make sense for Michigan it would almost need to be a touchdown. For Penn St it might make more sense to take interception from the 40 or so in because even a sixty yard drive is a stretch.
|3 years 3 weeks ago||I have a big post on this||
I have a big post on this coming up in the next month or two and I don't want to spoil too much of it, but in general, running backs are highly over valued. Passing the ball is a much more efficient use of downs. Fitz had an excellent game by normal standards and its nothing against him, just that he used a lot of downs that could have gotten more value from passing.
|3 years 3 weeks ago||It's hard for RB's to put up||
It's hard for RB's to put up big numbers in general, and with nearly 30 carries it takes a lot of yards to put up a big number. Plus, his final TD run plus several other carries were omitted due to the score/time situation.
|3 years 3 weeks ago||The penalty was worth 2.2%||
The penalty was worth 2.2%
|3 years 4 weeks ago||A little clarification. Plays||
A little clarification. Plays themselves aren't given values directly, but indirectly. The value for Fitz's run is calculated based on the difference between future winning percent of 2nd and 10 at the 20 and 1st and 10 at the opponent's 15 in the first minute of the game. The difference between those situations is 12.5%. Whether it 1 65 yard play or thirteen 5 yard plays, the change in expected winning percent between those situations is 12.5%.
There is no fitting the data to meet expectations. I have over 1 million plays over nine seasons and have calculated winning percentages using two steps. Calculate actual winning percentages in real situations and then smooth the data so there aren't odd variances.
For those that are really interested, the general equation for calculating win probability is the following logistic function:
Where L is the lead for the offense, adjusted for field position and down and distance
Where T is the time in minutes elapsed
I started out with something much more complicated but after some thorough reviews of the data, I was able to fit the data to the above equation with a high level of confidence.
|3 years 4 weeks ago||I have cost of failure a||
I have cost of failure a little higher (86% WP), with 92% on a TD and 87% on a made FG. That would point to going for it even if your odds of success were sub 20%. Plus, in the second quarter it's stil all about maximizing points, no point in counting possessions until the fourth quarter.
|3 years 4 weeks ago||The defense was consistently||
The defense was consistently bad and the offense had to be great. This year the defense is consistently solid and the offense needs to be productive, but not necessarily great.
|3 years 4 weeks ago||All of the data is based on||
All of the data is based on nearly 10 years of actual results. I agree there is a lot of time to play, but a 7 point lead translates to a win over 70% of the time even in the opening minutes of the game.
|3 years 4 weeks ago||The pick is based on the||
The pick is based on the numerical projections with a slight adjustment to make the score match "normal" football scores.
Nebraska is calculated at a 62% chance, Ohio at 74%. About a 10% chance we lose both and above 45/45 that we win 1 or 2.
|3 years 6 weeks ago||Each week game scores are||
Each week game scores are re-adjusted for all prior games based on each teams work for the season in total. In week 7 Illinois was a +11. If they played this week they would be a +5. I don't forecast a trend, but their recent results have brought them down by nearly a touchdown in the last month.
|3 years 6 weeks ago||Didn't post them over the bye||
Didn't post them over the bye week and the last two weeks have seen Iowa slighly increase and Michigan hold. Change is pretty minimal.
|3 years 6 weeks ago||Even though the rank is low||
Even though the rank is low the effect is minimal. Special teams really only matter if you are really good or really bad at them which Michigan is neither. Kickoff is a strength, FG kicking is slighly above average. Punting has been a disappointment and both return games are the biggest source of the ranking.
|3 years 6 weeks ago||Going has a 38% chance of||
Going has a 38% chance of scoring a TD and the 1st and 10 has a 62% chance of scoring a TD. However, to get back to the same spot, Florida had to get a couple breaks. Most punts from the 42 don't get downed at the 4, the average is the 12. Second break was getting a stop. Third was a weak punt by Georgia that gave the Gators back the ball in good shape. Usually trading punts in this territory is a net loss for the driving team.
|3 years 6 weeks ago||Far and away his worst. He||
Far and away his worst. He only has one other negative day on the ground in the last two years and that's last year against Illinois. Most of his negative against Purdue was on one play, the failed 4th and goal from the 1.
|3 years 6 weeks ago||This is written, like Brian's||
This is written, like Brian's preview, in terms of Michigan. Rush Defense means Michigan defending Iowa's rush, so the ratings given for Iowa under this heading are for Iowa's Rush Offense (Michigan's Rush Defense).
|3 years 7 weeks ago||Don't have a formal way of||
Don't have a formal way of doing it but doing a quick check shows that Ohio was -6 on offense through week 5 and +0 since then. A six point swing is worth about 15-20% increase on every game which would swing all the odds, will try and add this in next week. It would also move Michigan from a 7 point favorite to a 1 point favorite.
|3 years 8 weeks ago||The time left is the key. I||
The time left is the key. I assumed that to score on the first try you would take off more time, giving less time for the second score vs the first.
|3 years 8 weeks ago||So to win if you go for it||
So to win if you go for it you have to:
#1 Convert 4th and 3 from the 17 (50%)
#2 Score a TD (74%)
#3 Recover an onsides kick (23%)
#4 Kick a FG and win in OT or win outright in regulation, likely with less than a minute remaining (15%)
About 1.3% chance of winning if you go for it like Zook did.
If you kick to win you have to:
#1 Make the FG (75%)
#2 Recover an onsides kick (23%)
#3 Score a TD (23%)
#4 Win in OT (50%)
About a 2.0% chance of winning.
The odds are slim either way but if you have at least an average kicker your odds are 55% better if you kick.
|3 years 9 weeks ago||I have K-St as a pick-em this||
I have K-St as a pick-em this weekend vs Texas Tech and with 3-4 losses when the regular season is over.
You know I always hate on NW, Iowa with the cover.
|3 years 9 weeks ago||
5 and 7
|3 years 10 weeks ago||1. Start at 59% for the home||
1. Start at 59% for the home team, add/subtract 3 percentage points for every point difference in PAN.
2. Like the commenter noted, MSU plays a more difficult schedule and Michigan is slightly higher rated.
3. Home field is the reason, every team's ratings are adjusted for the strength of opponents but the adjustment is still not fully developed this early in the season.
4. Odds of going 1-1 in toss-up games are about 50%, 25% each to win both or lose both.
5. Winning and losing is irrelevant to how the game grades out. A losing team can grade out higher than a winning team (like ND did against us). Handy wins should at least keep our PAN afloat assuming our prior opponents don't fall apart.
|3 years 10 weeks ago||Starting the game at only the||
Starting the game at only the 20 is a slight negative (one that was outside of Michigan's control) and sitting at 3rd and 5 after the loss on the first Fritz play pushed us slightly below 50%.
|3 years 10 weeks ago||Ohio St has always been very||
Ohio St has always been very good at the field position game. Their offense is just so terrible this year they can't do anything with it.
|3 years 10 weeks ago||Will add individual games||
Will add individual games from here on. For now, I'll leave it here in the comments:
@Northwestern: 92% (probably lower because I don't have a Persa factor in)
@Michigan St: 48%
Ohio St: 95%
|3 years 10 weeks ago||You have to get to 100% one||
You have to get to 100% one way or another and just like the TD early is worth 7 just like the one late, all points count the same. The plays with the most leverage come when the score is close, not far apart. Which is more valuable a play that doubles your chances of winning from 1% to 2% or a play that doubles your winning percent from 50% to 100%?
|3 years 10 weeks ago||End of half situations are||
End of half situations are removed from my numbers. As noted above, the blemish is from getting 3 instead of 7 in the first half last Saturday.
|3 years 10 weeks ago||Ebert and Persa are not +18||
Ebert and Persa are not +18 together. Receivers rarely get a negative play so their numbers are always highly positive. Plus adding a QB and a receiver's values together is like adding the yards together. A QB who throws for 300 and a receiver who catches 100 yards don't combine for 400 yards. Persa had a strong game against what to date has been a very good defense. +11 in one game is very good. +11 is top 5 level and well above Persa's historical success. He is good but probably not +11 over a whole year good.
|3 years 11 weeks ago||It's all situational.||
It's all situational. Vincent's run pushed the game from tie to 7 points (almost, I know he didn't actually score himself). Where Denard's TD took the lead from 7-14 which isn't as valuable as breaking open the tie.
|3 years 11 weeks ago||16.7% Chance of a touchdown,||
16.7% Chance of a touchdown, 73% half time odds.
7.8% Chance of a field goal, 85.5% half time odds.
|3 years 14 weeks ago||The Value added is inedepent||
The Value added is inedepent of score or time, it is based only on down, distance and field position. He would have added 9.1 whether it was in the first quarter or in the fourth up by 4 TDs. The value would always be 9.1. The effect on WPA is dependent on score and time. The big return while tied in the second quarter was worth a lot. If he would have done that in the fourth with Michigan up 3-4 TDs, the WPA would be effectively 0 because the game was already in the bag.
|3 years 14 weeks ago||As far as the model is||
As far as the model is concerned, all teams are perceived even, it only evaluates based on in game factors.
|3 years 14 weeks ago||The X axis is plays, forgot||
The X axis is plays, forgot to label it. The plays are split if there is a return. i.e. a kickoff and a kickoff return are two separate plays.
|3 years 15 weeks ago||Last Year||
Last year was the first year I did it.
|3 years 17 weeks ago||Thanks for the||
Thanks for the feedback.
#1 is absolutely an issue. I have done a double strength of opponent adjustment to try and adjust for this issue and you are right the SEC teams are the biggest gainers and the lower conferences are the biggest losers in the process. Unfortunately due to time, resource and coding ability constraints I haven't been able to do it in a manner that is fast and flexible enough to use in a productive way, at least not enough to justify what it is still a minor tweak. The first pass of opponent adjustment probably is 95% accurate for 75% of teams and about 75% accurate for 95% of teams.
#2 Don't know if I completely understand your question but I'll give it a go. My ultimate goal is to be able to compare teams against a universal standard. To do so, whether as a replacement value or average value, requires making some broad assumptions. Those assumptions won't be valid across all situations for all teams but in the aggregate should be reasonable, and at the very least they are highly consistent.
The other challenge with a more team specific approach is the sample size quickly reduces. Because college teams turn over faster than pro teams and the there are fewer games in a season, getting an adequate sample on a specific situation is nearly impossible.
|3 years 17 weeks ago||Fumbles||
Every way I have tried to look at it shows no correlation from year to year on fumbles forced or recovered. I know there is a lot of teaching coaching behind it, but I haven't been able to find any data to suggest it's possible to be consistently good or bad at it.
|3 years 28 weeks ago||Ran the numbers||
This should play out pretty similar to onsides kicks in terms of recovery, your bigger change is in field position.
In my DB, there are 21,084 kickoffs from the 30, the average start for the ensueing possession is the 29. In 873 punts from the 30, the average start for the ensueing possession is the 33. Not a huge difference, but definitely some difference.
In terms of recovery, I don't have great info for college, but in the NFL a non-surprise onsides is about 20%. In 2,195 3rd and 15 (as a proxy for 4th and 15) situations, the conversion for college teams in competitve game is about 16%. This would probably be a bit understated since worse offenses are more likely to get into 3rd and 15 situations.
This is an intriguing idea and it seems like 4th and 15 from the 30 (and maybe the 35) is about the right alternative to kicking off from the 30 in terms of recovery success and resulting field position. The only major difference is with a successful conversion the offense would be in better position than most onsides kick recoveries.
|3 years 37 weeks ago||Co-sign 100%||
For all the great things he did on offense, Rodriguez could not have handled the defense any worse. Casteel spoiled him at WVU and he obviously didn't know what he didn't know. Even with all the youth and peronnel issues, a marginal handling of the DC situation probably has him still in AA.
|3 years 42 weeks ago||I think Brian touches on the||
I think Brian touches on the fundamental conflict of any advanced metrics or methods. Are they about predicting or rewarding, which can be two very different ideals and by definition will on occasion be at odds. If the goal is predicting then the free at the end of the are relatively meaningless. They are two potential points in a pool of 120 for the game (this is Wisconsin basketball we're talking about). If we are talking about about rewarding then the free throws make all of the difference. We are taught that a team that wins by a point is clutch and the team that loses by a point can't take care of business when it counts. In reality most teams fall on both sides of the equation at different times.
Although the work I have done on college football is about predicting, I am with Brian that the NCAA tournament should be all about rewarding. The wins and losses do matter, because when you look back on a season, it is all about the wins and the losses, not about the PAN or the Kenpom rating. 4 points on Jan 27 in East Lansing made all the difference.
Where I think college basketabll can use some help is on a more wonkish look at rewarding/ranking. The predicting and grading performance is well covered by Kenpom and the like but there is a serious gap in the selection process that needs to be acknowledged. Currently the tools used to reward/rank are very loose and at best proxies for accomplishment. Record vs. Top 50 RPI is good, but games versus teams in the 40s are very different than games versus teams in the top 10. RPI is directionally good, but for a tournament calibur team, playing a team ranked 100 should be an easy win, just like a win versus team #300. For RPI sake they are very different. The easy win vs. #100 keeps your RPI afloat whereas the blowout versus #300 drops you.
I have put together a work in progress solution to this that does an initial RPI pass and then awards points for each win versus teams that are of a certain threshold. The better the team or the bigger the margin the more points awarded. Game location and time of year matter. Losses work the same way but with no threshold. The worse the team the bigger the loss of points. I think there is a lot of work to do on the weights (Georgetown is my #2 team) but I think this directionally where things need to go.
|4 years 4 weeks ago||1AA games are completely||
1AA games are completely excluded and not considered. All other games are adjusted for strength of opponent so its not just an effect of teams loading up on cupcakes at home and having to play real games on the road. The Non-AQ thing jumped out at me but the MAC is one of the lowest teams. The other idea I had was that it might be more about travel distance. The non-AQ conferences, other than the MAC and Sun Belt have wider footprints where most of the AQ conferences have more concise footprints. Certainly not the end all answer, but without digging in deep, seems like it could be a major contributor.
|4 years 5 weeks ago||Each play is assigned a value||
Each play is assigned a value on how it impacts the expected points of the team and how the field position affects the defense. The play is then adjusted based on the opposing defense does on an average run or pass play. The value is both assigned to the team and the individual players getting the stat. For running plays it is all assigned to the ball carrier and for pass players its assigned to the QB always and receiver if it is completed. Sacks are only assigned to the team and count as passing plays, not rushes. It's not perfect but it does a good enough job.
|4 years 5 weeks ago||My numbers have Penn St's||
My numbers have Penn St's opponent adjusted offensive numbers at +8 vs Michigan and +13 against NW. Even though neither defense is the 85 Bears, those number indicate that Penn St with McGloin have done substantially better than average teams have against bad defenses. Under Bolden, Penn St averaged -5 in six games and -11 in three Big Ten games. With only two data points the magnitude could be a fluke but there is no doubt that the Nittany Lion offense has been substantially better under McGloin. Can that hold up against an elite Buckeye defense is the question.
For the prediction I used Penn St's season average of -2 opponent adjusted on offense. If the can sustain the +10 level of the last two games thats a 12 point swing a makes the game a straight up pick em. I am not going to go so far as to say that my money is on pick em, but if you believe that McGloin can hold up at all against the Buckeye D then the Nittany Lions and the points looks like a great play.
|4 years 5 weeks ago||Yes, all the numbers and||
Yes, all the numbers and ranks are based on PAN. Denard is the top rusher and #2 QB behind Cam Newton. The offense is #2 overall and in rushing, behind Auburn for both.
|4 years 5 weeks ago||Wow that was an absolutely||
Wow that was an absolutely terrible couple of sentences from me there. I have edited it. Normal increase by defense = ~9 wins, above normal increase by defense to average level likely means 10+ and an outside shot to run the table.
Thanks for the catch.
|4 years 6 weeks ago||Thanks for the correction, I||
Thanks for the correction, I have fixed it.
|4 years 7 weeks ago||I typically go with the model||
I typically go with the model forecast, but with Penn St's injuries being difficult to account for (especially at QB) and the big advantage at special teams that I believe will be negated by Michigan offense I made the call to fudge a bit. The actual model forecast is 31-28 Michigan.
|4 years 7 weeks ago||Punt Stats||
My special teams numbers do the same thing and have Michigan 5th on the season. Unfortunately 2-3 points a game seems high. I have Michigan's punt team worth 5.6 points above average total on the season. The national leader is at +7.0. About a point a game seems to be as good as it gets by my calculation.
|4 years 7 weeks ago||Michigan can't and shouldn't||
Said perfectly. The classic player who can look good on a bad defense but not the kind of player you can build a good defense around.
|4 years 10 weeks ago||Was actually at this game||
Was actually at this game last night and after watching Martinez live and Denard on TV, they are very different players. Robinson is light years ahead in passing the ball. As one of the previous comments noted, their big runs have been very different. K St was woefully out of position all night on defense and Martinez had to get past one level and to the sideline to break it open. On Robinson's big runs he has had to beat the second level and have featured a higher degree of difficulty. Martinez is definitely fast enough to make a lot of teams pay. I think Robinson is clearly the better QB with the ability to change directions, be patient as the play develops and have a highly competent passing game. Martinez is very good though. His touchdown on the opening drive of the second half when K St was only down 14 and had Nebraska in 3rd and 10 was the classic, incredibly surprising QB draw and he had no one around from the time he hit the line of scrimmage.
|4 years 10 weeks ago||I have adjusted numbers but||
I have adjusted numbers but the Indiana game is really wack right now. With all of their other opponents so terrible, our adjusted performance for that game actually looks really good and throws everything off. That's why I put it down as incomplete, adjusted numbers with a great number against Indiana weren't reliable enough at this time.
|4 years 10 weeks ago||The touchdown percentage||
The touchdown percentage obviously varies based on starting field position. Drives starting the 30 yard line score 1.9 points per drive. If all of those points are on touchdowns, that equals a 27% chance of TD from a drive starting at the 30. Each yard on the offense's side of the field is worth about 0.5%.
|4 years 10 weeks ago||The ranking doesn't translate||
The ranking doesn't translate to win percent, the PAN does. The simplest way to look at is the home team starts at 59% chance of winning and adds or subtracts 3 percentage points for every point difference in PAN. This ratio has been very consistent over the years.
|4 years 10 weeks ago||Don't have specifics on hand||
Don't have specifics on hand from last year but I do know that the team grades out significantly higher this year than last. Without pulling the specifics, the defense is probably going to be marginally worse, as are special teams. But this is currently the #2 rated offense in the country, last year's group couldn't touch these guys. I will dig and see how this year compares with last and maybe put a post up tomorrow.
|4 years 10 weeks ago||For the most part their||
For the most part their numbers are pretty similar to mine in terms of expected wins. I have Michigan and MSU virtually tied in terms of team strength. Home field is worth 9% and so on a neutral field I would have it 51/49 Michigan.
|4 years 10 weeks ago||Agree that some of the||
Agree that some of the numbers, especially Wisconsin look off. I haven't looked at how Sagarin or anyone else does their projections, but my numbers are a pure power poll (regardless of wins or losses) based on how teams do on a down by down basis, including special teams. Each team's rating is based on they did against their opponent in comparison to how other teams have done against the same opponent.
For Wisconsin, their game scores are +13 vs UNLV, +1 vs MSU, -6 vs Arizona St and -9 vs San Jose St. On the season, they only have one game were they have played a team significantly better than the rest of that team's opponents. For Michigan, the game scores are +32 vs Indiana, +13 vs BG, +11 vs UConn and +1 vs ND. The Indiana game is inflated because everyone else the Hoosiers have played was without a pulse but overall, Michigan has played a tougher schedule than Wisconsin and played better against that schedule. I fully expect the Mich-Indiana game score to come down as the season progresses and Wisconsin's numbers to improve. The Indiana game is a big example of why 50% of the opponent adjustment is done based on 2009 performance still. Comparing Michigan to W Kentucky and Akron is going to always look good but as the season goes on the numbers begin to clear up more.
|4 years 10 weeks ago||National rank||
|4 years 11 weeks ago||My predicted spread is in||
My predicted spread is in line with my overall numbers but I did depress the overall point totals based on two factors. #1 I think one or both of the teams are going to look to slow this game down and reduce the overall number of possessions, thus reducing the overall point totals. #2 I have no numbers to back this one up, but it seems like matchups that are destined to be absolute shootouts often fall short of the sky high expectations.
|4 years 11 weeks ago||The calculations in your edit||
The calculations in your edit are correct. The pass offense probably didn't add up because I had Indiana's adjusted pass def as -2 allowed, when it should have been +2. This is now corrected.
|4 years 11 weeks ago||I have no doubts that our||
I have no doubts that our offense will roll against Indiana. The 35 point target is more because I anticipate us slowing the game down like we did against UConn to limit total possessions and give our defense a better chance to adjust and rest.
The discrepancy between Indiana pass off and our rush off is based on two things. It is much easier to put up huge passing numbers than huge rushing numbers. Michigan's rushing numbers right now are astronomical. The other main reason is that Indiana only has two games against terrible competition in the system. Those numbers will certainly come way down as the season progresses, even if the Hoosiers passing game stays very good. This is why I didn't want to put up a final number for Indiana passing. Anything I put would be overstated but there isn't enough data to tell by how much.
|4 years 11 weeks ago||Yes, home field is worth||
Yes, home field is worth about 9%, the home team will win about 59% of games against an equally matched opponent.
|4 years 12 weeks ago||When I was at a small NAIA||
When I was at a small NAIA school we had signed a big-time JUCO recruit who couldn't get ineligible for the NCAA but could for the NAIA. I asked our HC what happened when he didn't show and he told it was an old-fashioned SEC Miracle and he was now eligible for the NCAA.
|4 years 12 weeks ago||Assuming Denard plays the||
Assuming Denard plays the whole season and we make it to a bowl game, Denard needs "just" 94.1 yards per game over the next 10 games to set the NCAA QB single season rushing record and if he throws for 132.9 a game (easy) he can become the first 2000/1500 QB ever. 232.9 a game through the air gets him to 3000/1500. Only Vince Young and Dan LeFevour have done 3000/1000.
|4 years 12 weeks ago||They are removed from all||
They are removed from all teams.
|4 years 12 weeks ago||http://rivals.yahoo.com/ncaa/||
Michigan by 25.5
|4 years 12 weeks ago||Agree that that the MSU||
Agree that that the MSU decision making was very odd. However, assuming that he did not know Crist was coming back in the second half, I don't think the last play was the wrong call for ND. They didn't look like they could stop Denard and the backups hadn't been able to move the ball. A field goal at that point would have been a moral victory at best, but with stops and quality drives looking scarce for ND, I think he had to take the chance. Now if he thought Crist might be coming back it's not as much of a no-brainer in my book but still a very reasonable choice.
|4 years 12 weeks ago||I am far from a database||
I am far from a database expert but what I have works for what I need it to. My data source is the NCAA official play by plays from their website. The text is copied over from the net into an Excel file that "translates" the text into preset fields. Those fields are then uploaded into Access and I have a host of queries to manipulate the data from there. Each play has about 45 fields and a typical game has 180 or so plays. I have over 880,000 plays right now and will be close to one million by the end of this season.
|4 years 12 weeks ago||Even with over 7 years worth||
Even with over 7 years worth of data there are minor inconsistencies and variations in the data. The points where two colors intersect should be considered essentially equal. The lines between two decisions aren't hard and fast, but a general approximation for where the decision should begin to change.
|4 years 12 weeks ago||From the 1 the success on||
From the 1 the success on third down and goal is about 63%, far from fall forward for two points. This seems like a pretty good ratio to make the risk, uncertainty worth it for coaches but still adding an element of drama. Plus you could still kick late in the game if the score was appropriate.
|4 years 12 weeks ago||I would love to see the NCAA||
I would love to see the NCAA and or the NFL move the extra point to the 1 yard line. I think if the 2 PT conversion was solid option it would make the game much more interesting than kicking a bunch of relatively pointless PATs.
|4 years 12 weeks ago||There is a slight effect on||
There is a slight effect on yards gained changing field position, independent of points allowed, for great vs bad defenses, but the effect is very small and does not change the overall dynamics. Ultimately, it is about the offense. The special teams matter more than the defense, because they have situational value. If a kicker is really consistent, an ability to guarantee 3 from longer range is very valuable, as is a repeated ability to pin an opponent deep without allowing touchbacks. Good defenses are good everywhere and bad defenses are bad everywhere, there is no increase or decrease of relative value (that I have found) depending on where the offense begins a drive.
|4 years 12 weeks ago||On the graph, it bothers me||
On the graph, it bothers me too. Just been too lazy to figure out how to make it look right in Excel. The calculations match the correct yardline/distance combinations, just not the graph.
On the goalline. Good catch, I did not factor in diminished success around the goalline and a quick check shows that conversion rates are about 5-8% lower around the goalline for the same distance. Factoring this in does change things, kicking has a higher expected value with 7+ to go with a bad kicker and 4+ with an average one.
|4 years 13 weeks ago||UConn is 66 but started the||
UConn is 66 but started the season at 46 and saw a big drop after getting smoked by Michigan. Their second game was against a FCS team so it did not count.
|4 years 13 weeks ago||Homefield is accounted for in||
Homefield is accounted for in the win projections. Homefield is worth about 3 points per game. If otherwise equal teams meet, the home team will win just under 60% of the time. I do have numbers for all teams and a website, unfortunately I haven't had time to get the website up and going so it will probably be sometime in the offseason before I have anything publicly available.
|4 years 13 weeks ago||Michigan's projection is up||
Michigan's projection is up 1.2 wins vs preseason (7.5 to 8.7). Of the 1.2 increase 0.8 comes from starting 2-0. Bowling Green is up from 78% to 100%, Penn St is up from 32% to 49% and Purdue is up to 59% from 49%. Most of the other match-ups are within a couple points of their preseason marks.
|4 years 14 weeks ago||Denard was a +21 unadjusted||
Denard was a +21 unadjusted and a +18 adjusted based on last year's UConn averages. Against Western last year, Tate was a +8 unadjusted and a +6 adjusted. Tate's passing value was higher than Denard's but he obviously didn't have the ground performance. Western was Tate's third highest rated game behind Notre Dame and Purdue.
|4 years 15 weeks ago||Did a quick comparison of the||
Did a quick comparison of the true freshman QB's listed in the chart above using by Points Above Normal. In general true freshman QB's improve slightly in Year 2 but regress to the mean. As freshman they average just over 1.0 PAN, about 50th in the country. As sophomores the same players increased to about 1.7 PAN, about 10 spots nationally. The biggest benefit though, is when you look at overall team offensive performance. The increase in overall team offensive performance is about 3-4 times higher than the increase in individual QB performance. If Tate is the baseline, even if his numbers don't increase dramatically, the overall offensive performance should take a step forward.
|4 years 15 weeks ago||The PAN only measures direct||
The PAN only measures direct value. It's hard to add too much value if the other team doesn't even try. Most of the reason 2006 looks low is that there were very few attempts against Michigan in 2006.
|4 years 16 weeks ago||I currently have it at about||
I currently have it at about a 1 in 1,145 chance of going 12-0.
|4 years 16 weeks ago||Admittedly Michigan is an||
Admittedly Michigan is an outlier that is greatly favored using this weighting. With that said, every way I have looked at it indicates that the weighting should be no lower than 50% historic and even at that level Michigan is higher than the Projection (-) scenario above.
I didn't even mention the fact the 2009 team whose PAN history very closely mirrors the 2010 Michigan team was Miami (Yeah that Miami) who went 9-3 last year.
|4 years 16 weeks ago||Mathlete not so gud at||
Mathlete not so gud at Englush
|4 years 16 weeks ago||Minnesota has a tougher||
Minnesota has a tougher schedule with they don't play either of the two worst teams in the conference, Indiana or Minnesota.
|4 years 16 weeks ago||The methodology behind the||
The methodology behind the split can be found here:
|4 years 16 weeks ago||The numbers used for the team||
The numbers used for the team projections are based on team performance, not individual and do include special teams. Although the variance on special teams is much lower than that of offense and defense, they are included in the team projection.
|4 years 16 weeks ago||Those things are definitely||
Those things are definitely factors but they are difficult to model, especially given a limited data set. To effectively factor a coach you have to have an extended period at an old position plus time with a different coach there to compare program baseline (Winning half your games gets you a new job at Eastern Michigan, it gets you fired at Alabama). Plus, this method works extremely well without coaching adjustments. Last year with less data I predicted a final record for all 120 FBS teams and was within 1 game for 58 teams and within two games for 90/120.
|4 years 16 weeks ago||I am a bit limited on my||
I am a bit limited on my dataset, but using what I had, here is what I found. As you increase the years used for the historic portion, the total Mean Absolute Deviation decreases. In other words, the more years back you go, the more accurate the prediction is. This obviously has limits, but with the data from 2003-2009, for each year you add back, you reduce your error by 4-5%. At five years you have reduced the error about 20% from just repredicting the previous season.
So if you have justified using more years (and through 2009 results, using all years possible has been justified) then the question becomes what is the right mix between new and old. For the completed 2009 season, the answer was 65% old, 35% new which I just rounded to 2/3 and 1/3 because the error isn't very sensitive to small changes in splits. This does make a big difference for team's like Michigan. Depending on which way you move the splits can make a big difference for teams whose recent performance significantly varies from historic performance.
|4 years 16 weeks ago||It's not really about the||
It's not really about the players returning, it's about the program's historical performance. Because of the nature of college football, team success is reasonably consistent over the years as classes come and go. Some obviously do better than others but this is an attempt to establish a baseline of where the program is at. The previous year is used to account for the most current snapshot of program performance.
|4 years 16 weeks ago||The Irish are actually down||
The Irish are actually down at #46. They were +5.2 last year but their historical is a brutal -0.3 thanks to a poor 2003 and brutal 2007-08.
|4 years 16 weeks ago||1. 2003 is as far back as I||
1. 2003 is as far back as I can get data, so it can't go any further than that. A 5-6 year window seems to be a good representation of the calibur of a team. How good a team was 8-10 years prior doesn't have a whole lot of relevance to present day but the period is long enough that one fluky year won't move the needle too much.
2. Strength of schedule is factored into the historical numbers. The numbers used to determine the rankings are all adjusted for strength of opponent in the given year. What is not accounted for is how this season's schedule will affect final records. Boise St and LSU are both essentially +17 but you would be crazy to think that they would have the same record. Boise St's average opponent is a -4.6 rating and LSU's average opponent is +2.0. The two teams could play equally as well but LSU will likely have a worse record.
|4 years 17 weeks ago||I dug into this one a bit||
I dug into this one a bit more and on first and second down, Henne was the #4 QB in the Big 10. On third down plays, which are have very big expected value swings due to their binary nature, Henne dropped to #14 over his career. That's the data. Beyond that I think Michigan's pass game was more suited to big plays set up by play action and the running game. I don't know that their offense and thus Henne's performance were as good as other systems when they couldn't lean on a running game to convert third downs. Again, the system thing is just a theory but there is a huge swing in his peformance ranking on 3rd down vs 1st and second.
|4 years 17 weeks ago||Things can't be inflated that||
Things can't be inflated that much. Plays in the second half are only considered if the game is within 2 scores, no 1AA games are included and all games are adjusted for strength of opponent.
|4 years 25 weeks ago||Bornstein||
Grant Wahl just tweeted that Bornstein is starting. Not good.
|4 years 27 weeks ago||Very intriguing comparison.||
Very intriguing comparison. I pulled up Oregon from that era and found something interesting. Even though Oregon passed it more often and for more yards per game, West Virginia under RR got more value for their plays. Numbers are from my expected value model and opponent adjusted. POA = Points per game Over Average
WV06: 15.5 POA, 10.8 Rush, 4.7 Pass
WV07: 12.4 POA, 9.3 Rush, 3.2 Pass
Oregon07: 8.5 POA, 6.8 Rush, 1.8 Pass
Oregon08: 10.9 POA, 9.5 Rush, 1.4 Pass
Oregon09: 7.9 POA, 5.7 rush, 2.1 Pass
WV averaged about 4 points per game through the air vs Oregon at less than 2, despite the increased attempts and yards. Not only was WV's production more efficient in fewer attempts, but the absolute production (in value, not yards) was greater even with fewer plays.
|4 years 27 weeks ago||Glad to see a fellow Wichitan||
Glad to see a fellow Wichitan out here!
|4 years 29 weeks ago||My calculations put the||
My calculations put the seasons in this order: 41352
|4 years 30 weeks ago||All games between 2 FBS teams||
All games between 2 FBS teams from 2007-2009.
|4 years 30 weeks ago||Failure to note on my part.||
Failure to note on my part. The data is only first 3 quarters and games within 2 TDs.
|4 years 30 weeks ago||Actually sat in front of||
Actually sat in front of Coach Hoke on a flight to San Diego yesterday while wearing a Michigan shirt.
|4 years 31 weeks ago||Big Mathlete follow up post||
Big Mathlete follow up post coming next week on 4th down choices based on these discussions and trying to refute some of the objections from the Romer paper.
|4 years 31 weeks ago||http://web1.ncaa.org/mfb/2009||
Plus a lot of work to get all the play by play data organized.
|4 years 31 weeks ago||In the last 3 years there||
In the last 3 years there have been 13 punts from inside the 30 yard line, Virginia Tech and Temple have both done it twice. Last year Southern Miss was the worst offender, punting from the 29 on 4th and 6. The punt was of course a touchback.
|4 years 31 weeks ago||The NCAA play by play data||
The NCAA play by play data classifies everything at the goalline or in the end zone the same way so I have no way of directly pulling it out. With that said, the relationship is pretty linear for the first 5-10 yards so you could pretty safely assume that it would hold into the endzone as well. This would but the break even point of kneel or return at 3-4 yards deep in the endzone.
|4 years 31 weeks ago||Using the data from the 4th||
Using the data from the 4th down decision chart, I assumed that the incremental times the team went for it on 4th above the 15-20% that would do it anyway would be missed FGs. This kept the number of FG makes the same but increased the number of attempts, lowering the FG%.
|4 years 33 weeks ago||You are right in that the||
You are right in that the second graph tells nothing as to whether a team was succesful on offense or not, just whether their playcalling was producing a balanced output. The primary objective I had with the second chart was to show that even teams that are the worst at traditional metrics of playcalling balance, have actually crafted very balanced outcomes because all plays yield roughly the same result, even if they are calling a lot more of one than the other.
You are correct in that all the data is based on down and distance. I think what you are looking for is all in the first chart. Teams in the top right box are teams that are effective at both the run and the pass. The teams that are in the right box and close to the diagonal line are the teams that are good at both and have optimized their playcalling selection (when accounting for down and distance). Hope this answered your questions.
|4 years 33 weeks ago||I am still working out the||
I am still working out the kinks, but my general approach is similar to regressing to the mean, but instead of regressing to a historical average - which would make Miami Ohio and Central Michigan top 20 programs and Oregon out of the top 50 - I am regressing to a weighted recruiting history. The better your recruiting over the last 4 years, the better your mean is. The outcomes are not all that dissimilar, but the distinction is important.
|4 years 33 weeks ago||You are right that based on||
You are right that based on the tradition comparison, yards per play, there is an unstated advantage to running because of the interception risk in passing. With this analysis, however, it is not based strictly on yards per play and interceptions are included, so the bias is removed. This is the best attempt I can do to take out all of the extraneous factors and get a true apples to apples comparison.
|4 years 33 weeks ago||All games versus FBS||
All games versus FBS opponents are always included in my analysis and no games versus FCS opponents are ever included.
|4 years 34 weeks ago||A special teams primer is now||
A special teams primer is now in the works. I am thinking of going into further depth in terms of how to truly account for special teams, both in the decision making (punting from the opp34) and the evaluation (punters shouldn't be penalized for punting from the opp34). Post will probably be a hybrid of how to appropriately evaluate and what those evaluations look like.
|4 years 35 weeks ago||The adjusted value||
The adjusted value measurement really only has value for full-time players. It compares their positive production versus the production of other full-time players at their position. Comparing Kenny Demens production on limited plays, isn't really valuable. What Kenny did do, was make 2 plays that put the opposing offense in a worse position. For his example, he was credited with a tackle on 1st and goal from the 1 against Wisconsin, that play had a value of .14, Wisconsin's expected points went from 6.34 on 1st and goal from the 1 to 6.20 on 2nd and goal from the 1. His second play was on a punt return. He combined with Kevin Leach to limit the returner to a 2 yard return after a 47 yard Mesko punt. The shorter return was worth .17 in field position for Michigan that was split between Leach and Demens.
|4 years 37 weeks ago||The comparison is between||
The comparison is between returning starts and change in value instead of absolute value. The point is to answer the question, if you have a lot (or a few) returning starts, do you get better or worse and to what magnitude.