Anyone want to help take a stab at crowdsourcing the SRS?

[rokamortis]

So all of the talk of the SRS got me wondering if we could create some reasonable imitation based on the information the NCAA provides in the handbook. The first step is to get the scores in for all of the FCS teams and then we can play with how to calculate the SRS.

I started with the Big South teams but could use some help in getting the rest of the conferences as I really have no interest or time in doing all of the data entry myself. So if anyone is interested you can PM me your email and I can give you edit access so you can add the scores. Once that is done I'm sure we could use some folks eyeballing it to ensure that the data was entered correctly.

The spreadsheet is here: https://docs.google.com/spreadsheets/d/1ezYvatq37YIGS_Ey2AGlz5u-WhsMVKaA3xYi7fjCZV8/edit?usp=sharing . It should be view only unless you have been added manually.

Here is how I implemented based on my understanding of the rules.

FCS Home Win: .75
FCS Home Loss: -1.25
FCS Road Win: 1.25
FCS Road Loss: -.75
FCS Neutral Field Win: 1
FCS Neutral Field Loss: -1

FBS Road Win: 1.35
FBS Road Loss: -.65
FBS Home Win: .85 (I don't think we have any - but just for completeness)
FBS Home Loss: -1.15 (I don't think we have any - but just for completeness)
FBS Neutral Field Win: 1.1 (I don't think we have any - but just for completeness)
FBS Neutral Field Loss: -.9 (I don't think we have any - but just for completeness)

DII Home Win: .65
DII Home Loss: -1.35
DII Road Win: 1.15 (I don't think we have any - but just for completeness)
DII Road Loss: -.85 (I don't think we have any - but just for completeness)
DII Neutral Field Win: .9 (I don't think we have any - but just for completeness)
DII Neutral Field Loss: -1.1 (I don't think we have any - but just for completeness)

DIII / NAIA / Other
0 (doesn't count towards SRS)

Byes
0

My assumption is the average opponent W-L score is the base for SOS and then once that is calculated for everyone we can recalculate with the combined SRS (SOS and W-L) calculation. Any assistance from mathematicians / statisticians on the board would be helpful. I believe a couple of Furman profs created the SRS so perhaps if someone knows them and can get clarification that would be helpful. We can cross the bridge if we get the data.

The calculated fields are:
W-L Calc: average of the scores*10; done to help standardize based on different number of counter games and the multiplier will likely change
Counters: games that have non-zero values.
C Wins: stands for Counter Wins, number of wins that count towards the SRS. It isn't necessary but thought it would be helpful to see.
C Losses: stands for Counter Losses, number of losses that count towards the SRS. It isn't necessary but thought it would be helpful to see.
SRS: just a placeholder for now

[superman7515]

FBS Home Win: .85 (I don't think we have any - but just for completeness)

Yale

[rokamortis]

Yale
Nice catch.

[rokamortis]

Thanks to some awesome volunteers (I won't out them in case they want to be anonymous) we now have the Big Sky, Big South, CAA, NEC, MVFC, Pioneer, and independent Charlotte entered. Anyone want to help put in the others like the Ivy, Patriot, OVC, Southern, Southland, MEAC, and SWAC?

[BisonTru]

I'm willing to add some more when I have down time. But I think to get the SRS correct we need to figure an SRS for all of the FBS and Division II, too. I'd surely help when I can, but we could use some more volunteers if that's the case.

[rokamortis]

I'm willing to add some more when I have down time. But I think to get the SRS correct we need to figure an SRS for all of the FBS and Division II, too. I'd surely help when I can, but we could use some more volunteers if that's the case.

Good point, because it is opponent schedule and they allow FBS and DII to count towards that. Ugh ... but this is interesting and I think it helps us. Once we get the data in we could even tweak it if we wanted to create an AGSRS. I'd probably just start with average of FCS opponents though and see how close we get - because we don't really know how many points they'd give the other divisions - although we could assume it would follow a similar methodology.

Does anyone know of a good database for this data? I've seen some things but haven't seen something that has DII, FCS, and FBS so I have no idea if it exists. If we had access to the database that listed the schedule for each team with location (home, away, neutral), result (either W/L or score of team and opponent), and division of opponent (FBS, FCS, DII, other) then we could write a script / program instead of hand entering this. Or at least something pretty close and we could probably work around it.

[robsnotes4u]

Dont know if you saw this. I posted it on another thread. Has, MOV, but you can adjust the formulas.

http://www.screencast.com/users/Sports-Reference/folders/PFR/media/cea8d1e3-ed91-431f-a0df-1336c03b1268

[thebootfitter]

Rok - I just saw this thread. It fits nicely with a bunch of data I've been collecting over the past couple years. I just hopped on a plane, but if the wifi works, I can take a peek when I'm in the air.

FWIW - I have scores of all D-I (FBS and FCS) games for 2013 & 2014, including division and home/away, etc. There are a number of questions I'm trying to answer with it all, but I don't think it would be difficult to add in the SRS points. (Not to take away from anything you've already done... But hopefully I can add to it.)

Edited to add: I think I can easily get all of the D-II game scores too.

[centennial]

The document clarifies how they get the SOS and normalize for the amount of games played. However, the clarification hinges on the SRS in some kind of weird cross. Minor in math, physics and degree in ECE and I have hard time figuring out their unique take.

[rokamortis]

The document clarifies how they get the SOS and normalize for the amount of games played. However, the clarification hinges on the SRS in some kind of weird cross. Minor in math, physics and degree in ECE and I have hard time figuring out their unique take.

I have a comp sci background (good bit of math but certainly not a mathematician or statistician) and I agree, it just doesn't spell out all of the secret sauce. But I think we could get some estimation. Hell, after just seeing some of the results it is pretty clear what the W-L score does for a team and that is the main component of the SRS.

There is a good article about SRS that names the Furman profs that helped create it. I'd probably just reach out to them and see if they'd give some advice on the calculation.
http://edge.furman.edu/furman-professors-help-ncaa-crunch-some-numbers.html

[rokamortis]

Rok - I just saw this thread. It fits nicely with a bunch of data I've been collecting over the past couple years. I just hopped on a plane, but if the wifi works, I can take a peek when I'm in the air.

FWIW - I have scores of all D-I (FBS and FCS) games for 2013 & 2014, including division and home/away, etc. There are a number of questions I'm trying to answer with it all, but I don't think it would be difficult to add in the SRS points. (Not to take away from anything you've already done... But hopefully I can add to it.)

Edited to add: I think I can easily get all of the D-II game scores too.

That's awesome. I don't feel like I need to recreate the wheel here or have 2 different systems. So if you already have things at a certain point (seems like you are much further ahead) I'd be happy to let you run with it. Take a look and we can chat.

[thebootfitter]

That's awesome. I don't feel like I need to recreate the wheel here or have 2 different systems. So if you already have things at a certain point (seems like you are much further ahead) I'd be happy to let you run with it. Take a look and we can chat.
I sent you a PM too, but just to reiterate here... I am happy to contribute the data that I've gathered (which now includes all DII and FBS games), which I think will be helpful, but my primary objective isn't really to reverse engineer the SRS. I'll contribute my data and try to add to the conversation where I can, but I'd rather have this be in your hands and have others contribute as well.

My primary objectives for gathering the data are to explore the biases associated with computer models for cross-divisional match ups (specifically Sagarin and Massey). I'm also exploring the impact of Sagarin ignoring the relative measures of sub-DI teams. Many folks (myself included) tend to discount the value of computer models for cross-division match ups, and some people discount Sagarin entirely for use with FCS teams because of the exclusion of sub-DI teams. I'd like to know how much impact both of those items really have on the figures.

Edited to add: Can someone re-post the link to the SRS description in this thread? I know it was posted somewhere else, but I couldn't find it. Would be nice to have here for reference.

[rokamortis]

I sent you a PM too, but just to reiterate here... I am happy to contribute the data that I've gathered (which now includes all DII and FBS games), which I think will be helpful, but my primary objective isn't really to reverse engineer the SRS. I'll contribute my data and try to add to the conversation where I can, but I'd rather have this be in your hands and have others contribute as well.

My primary objectives for gathering the data are to explore the biases associated with computer models for cross-divisional match ups (specifically Sagarin and Massey). I'm also exploring the impact of Sagarin ignoring the relative measures of sub-DI teams. Many folks (myself included) tend to discount the value of computer models for cross-division match ups, and some people discount Sagarin entirely for use with FCS teams because of the exclusion of sub-DI teams. I'd like to know how much impact both of those items really have on the figures.

Edited to add: Can someone re-post the link to the SRS description in this thread? I know it was posted somewhere else, but I couldn't find it. Would be nice to have here for reference.

First of all - thank you. I took a peak at the data and it is glorious. I'll for your reply to a rather lengthy email before posting. My my goodness - it is exactly what I wanted and more. Just can't thank you enough. Really impressive. Like looking into the briefcase in Pulp Fiction.

Your interest in the data and biases are very interesting. I'd be interested in learning more.

Here is the 2014 FCS manual: http://www.ncaa.org/sites/default/fi...5_Revised2.pdf (http://www.ncaa.org/sites/default/files/PreChamp_D1_Football_2014-15_Revised2.pdf)

Specifically, the SRS information is:

NCAA Simple Rating System

The NCAA Football Championship Committee will use the NCAA Simple Rating System (NCAA SRS) as a tool for evaluating teams for selection into the 24-team championship.
The NCAA SRS is a ranking system used to gauge team quality. Within the NCAA SRS, the rating of a team will be calculated largely by two components: a strength-of-schedule measure (SOS) and a win-loss differential (WL).
A team’s SOS measure is simply the average NCAA SRS rating of that team’s opponents for the season.
A team’s WL measure factors whether or not a game was won or lost; the location of the game (home/away/neutral site); and the NCAA (sub)division of the opponent.


In a game between two FCS teams played at a neutral site, the winning team gets one point and the losing team loses one point.
In a game between two FCS teams played at a non-neutral site, a home win counts 0.75 for the home team and -0.75 for the losing team. A road team win counts 1.25 for the visiting team and -1.25 for the home team. In other words, there is a bonus/penalty of +/- 0.25 for the home/away teams depending on the outcome.
An additional bonus/penalty of +/- 0.1 is added/deducted for games that FCS teams play against FBS/non-Division I opponents. For instance, a home FCS win against a Division II opponent counts only as +0.65 points (0.75 for a home win - 0.1 penalty for playing a Division II opponent). In addition, a road FCS win against an FBS opponent counts +1.35 points for the FCS team (1.25 points for a road win + 0.1 bonus for beating an FBS opponent).
The WL measure is then doubled to equalize the contributions of the WL factor and SOS factor toward a team’s NCAA SRS rating.

The teams’ ratings are independent in that one team’s rating depends on its opponents’ ratings, which depend on their opponents’ ratings, etc., based on the “network” of college football games played each week during the football season. The NCAA SRS will also standardize the data in order to equalize the variance in total number of games played (i.e., 11 or 12 regular-season games).
Margin of victory is not factored into any team’s NCAA SRS rating.
Games against Division III or non-NCAA opponents are also not factored into any team’s NCAA SRS rating.
The time of the season in which a game is played (early-season versus late-season games) is not factored into a team’s NCAA SRS rating.
The Football Championship Committee will begin reviewing NCAA SRS data during the later portion of the football season as it prepares for championship selections. At that time, NCAA SRS rankings may also be posted on the NCAA website (www.ncaa.com/fcs (http://www.ncaa.com/fcs)).
The NCAA SRS will not “force select” any teams into the championship; rather, it is one of several resources that the Football Championship Committee will have at its disposal when debating the merits of teams under consideration for championship selection.

[MR. CHICKEN]

Anyone want to help take a stab at crowdsourcing the SRS?.....xsmhx....AH'LL WAIT FO'.....SELECTION SUNDAE.....xnodx.....BRAWK!

[FargoBison]

Wow UNI AD, Troy Dannon ripping the SRS rating pretty hard right now on Fargo radio, says it is biased towards the bottom FCS leagues and doesn't pass the eye test. He thinks the SRS is so bad that he doesn't even want it to be released.

He also said the bottom leagues feel the same about the GPI, he said GPI wasn't even discussed by the committee for one second last year.

[tingly]

The committee didn't seem to do much with SRS last year, either. Accurate ratings are tough in football cuz there are so few games and very little interconnecting of schedules. In sports where they do get reasonably accurate, they're only used for indirect things like 'record vs. top 50' instead of a direct 'the rating has A above B.'

[clenz]

Wow UNI AD, Troy Dannon ripping the SRS rating pretty hard right now on Fargo radio, says it is biased towards the bottom FCS leagues and doesn't pass the eye test. He thinks the SRS is so bad that he doesn't even want it to be released.

He also said the bottom leagues feel the same about the GPI, he said GPI wasn't even discussed by the committee for one second last year.
Strong words from someone currently sitting in the committee.


Is there a podcast link?

Sent from my SCH-I545 using Tapatalk

[FargoBison]

Strong words from someone currently sitting in the committee.


Is there a podcast link?

Sent from my SCH-I545 using Tapatalk

They don't usually podcast but the interview was really good, so maybe they will post it online at some point. If you tweet or email Dom Izzo he will probably put it online.

[MR. CHICKEN]

20009........LET US EXAMINE.......WENT FROM 16 TEAMS.......TA....TWENT-AH-FOUR.......AN' MO' CRYIN'.....DEN EVERAH........xsmhx...BROCK!!


sent from mah computer.........HAL

[tingly]

another Troy Dannen interview http://www.youtube.com/watch?v=eCotwOXYY44

some things I pulled out of it:

The number of Division I wins is the main factor. An FBS win helps, but won't overcome a lower finish in conference standings. FBS wins may make a difference, but they question wins over lowly FBS teams. 8 D-I wins in a 12-game year, 7 in an 11-game year are the absolute minimums. 6 gets considered by rule, but they have no hope. "Wins over FBS" sounds like it isn't used until the very end like a seed or bubble tiebreaker.

You need to be playing well at the end of the year, rising in the polls/ratings. Explanations can make up for the lack of that.

The seeding meeting's work is mainly determining the last 4 at-large spots and seedings.

There is a high value placed upon who you beat. (It's like I always say, if you want a spot, you have to prove it to the committee. They won't guess. Weak schedules are death.)

[robsnotes4u]

another Troy Dannen interview http://www.youtube.com/watch?v=eCotwOXYY44

some things I pulled out of it:

The number of Division I wins is the main factor. An FBS win helps, but won't overcome a lower finish in conference standings. FBS wins may make a difference, but they question wins over lowly FBS teams. 8 D-I wins in a 12-game year, 7 in an 11-game year are the absolute minimums. 6 gets considered by rule, but they have no hope. "Wins over FBS" sounds like it isn't used until the very end like a seed or bubble tiebreaker.

You need to be playing well at the end of the year, rising in the polls/ratings. Explanations can make up for the lack of that.

The seeding meeting's work is mainly determining the last 4 at-large spots and seedings.

There is a high value placed upon who you beat. (It's like I always say, if you want a spot, you have to prove it to the committee. They won't guess. Weak schedules are death.)

solid xthumbsupx

[centennial]

Wow UNI AD, Troy Dannon ripping the SRS rating pretty hard right now on Fargo radio, says it is biased towards the bottom FCS leagues and doesn't pass the eye test. He thinks the SRS is so bad that he doesn't even want it to be released.

He also said the bottom leagues feel the same about the GPI, he said GPI wasn't even discussed by the committee for one second last year.
Of course. SRS ignores for the most part how strong a conference is, the most important component is how many wins you have. Why would any league want to discuss a fair system like GPI when SRS can be politically correct. I actually think a couple of weak teams in the MVFC would help, problem is we don't have any.

[robsnotes4u]

Of course. SRS ignores for the most part how strong a conference is, the most important component is how many wins you have. Why would any league want to discuss a fair system like GPI when SRS can be politically correct. I actually think a couple of weak teams in the MVFC would help, problem is we don't have any.

The problem of SRS, or the SOS part of SRS is simple. It all depends how far you go. The BCS only went as far as your opponents opponents, three levels. Go all the way for a better result.

USD?
W. Illinois?

[thebootfitter]

USD?
W. Illinois?
It's all relative. Definitely the weakest teams in the MVFC right now, but still pretty solid teams that would probably be at least 0.500 teams in most other conferences.