SRS system - how does it really work?

[wcugrad95]

I have read the NCAA-provided details about the SRS system, and even posted them in some other threads. I understand how they determine your WL ratio, and how they provide points based on location (home/away/neutral) and classification of opponent (FBS/FCS/DII). I also understand the SOS is really not based on any of the widely-used systems (Massey, Sagarin, etc.) but is actually based on the SRS scores of your opponents, that is based on their WL ratio and the SOS of their opponents, that is based on the WL ratio and SOS of their opponents....

So I get the overall premise, and on the surface it sounds like a pretty fair and reasonable way of at least getting the field narrowed down as best they can given the discrepancies in competition. However, they mention that Margin of Victory (MOV) is now used and capped at a max of 21 points, but they do not provide any clearly stated details about how it is used. They say your SOS is based on the teams you played and their SRS, but they don't tell you how to use (or if you are supposed to largely ignore) games against FBS teams for the SOS component.

It also sounds like several posters believe that the SRS is only important if it fits the committee's narrative on who they are including. I think it helps whittle the field down, but it is not the single point of reference for the committee. So does anybody really know how they use MOV and value FBS games (outside of just giving fractionally more for a win and taking fractionally away less for a loss for your WL ratio)?

Sorry - I have mentioned some of this in other threads, but thought I would start a topic in case there are other people out there wondering about SRS or some "experts" on the topic. There has to be an actual equation somewhere, but all the details you would need to calculate it don't seem to be provided by the NCAA.

[MR. CHICKEN]

........DIS SECRET....IS MO' HIGHLAH....GUARDED........DAN DUH RED PHONE #......IN WASHIN'TON....(Endicott 8-8449)..........AH'D GUESS....DEY'ED....PULL CARD FROM BOTTOM UH DECK....FO' BUBBLE SQWADS......WHIFF......GOOD ATTENDANCE....:D....BRAWK!

......DUH BIG BIRD....IS JES' FISHIN' FO' GIGGLES............AH KNOW SQWAT........'BOUT DUH......NASA LIKE FORMULAH.......YER TALKIN' 'BOUT.....xsmhx....AWK!

[Reign of Terrier]

I think it would be really helpful if someone spreadsheet oriented who knew how this worked could make one and post it with regard to the playoff bubble teams

[Professor Chaos]

I don't think the NCAA has ever provided an equation or algorithm for it. It seems kind of silly because there's plenty of places where you can find the general algorithm for the RPI in basketball. I think it's because the NCAA knows it's still very rough around the edges and is continually tweaking it. You're right that it's used by the committee (and by their claim the only ranking that is used) but it's just a single data point in the larger picture of a team's resume since they've shown that they'll go against it when it comes to seeds and when it comes to the last few at-larges in the field.

[wcugrad95]

If I had how they factored MOV and what they plugged in for an FBS team's SRS to use for SOS, I'd be at least pretty close. I would just be using the WL "score" of the teams WCU played rather than the official SRS score, but it would be pretty close. If it really is the SRS of each team, that is calculated on the score of each team that played, which comes from the scores of each team that team played, and so on. So to get where they say they get, you'd need the final SRS score of every school because it is somewhat dependent. Don't think I can do that even if I had the missing data points - took me a while just yo plug-in WL ratio of Western's 10 FCS opponents.

[kalm]

Lance mentioned in The Wedge Podcast Week 10 Review that under the current SRS guidelines, a w/l against a 9-1 San Diego pretty much gets rated the same as a w/l against a 9-1 NDSU.

[Lehigh Football Nation]

Here's how it works:

* Get in a room
* Pick all the teams you want in the playoffs and rank them 1-18 and stick the rest of the autobids in there

Voila SRS

Kidding. Here is the real definition (https://www.sports-reference.com/blog/2015/03/srs-calculation-details/).

[wcugrad95]

I looked at that site and several others last night, but they don't really provide the special sauce that is used. There is an actual algorithm somewhere, because there are actual numbers that come out in the and. But I can't find the full equation for the NCAA-defined SRS. The site provided above and the others I can readily find provide a lot more detail on how things are done comparing pro teams. SRS is supposed to take into account points (MOV) and end up with a ranking that shows how much better you are than the average of all teams. But if you look at SRS for say FBS teams, the top teams will be like 24 or 22. If you look at the SRS scores for the teams in the playoffs last year, NDSU was 2.00 and Samford was #24 as what I would presume was the last team in/considered at .45. I can't see how the absolute top of the FCS is only considered 2 points better than the overall average of the FCS field. So the FCS formula does something really screwy somewhere.

Or it really is the "get in a room" approach and the numbers just get made up to tell their story xeyebrowx

[Lehigh Football Nation]

Someone a few years ago tried to do a SRS of all of college football (D-III to FBS) and it came up with some ridiculous numbers because of limited data. Like Wisconsin-Whitewater being ranked higher than most mid-Top 25 FCS teams and some FBS teams, stuff like that. The reason is that there isn't enough of a dataset between W-W's opponent list and, say, Baylor's to come up with a proper-looking rating system. So right from the get-go there have to be some fudges involved.

SRS works best in a closed system like the NFL where there are a fixed set of 32 teams, and they can only play each other. In an open system the only way to make it work is what I would consider fudges where you try to simulate the "quality" of, say, a win over 0-8 Georgia Southern and how that fits in with the existing system. The SRS to its credit gets better at being predictive the more games are played.

[wcugrad95]

And now that I actually pay attention to last year's SRS, it makes me feel more and more like it is not even a very important data point. Originally I expected Samford from the SoCon was one of the last teams in for 2016. Per last year's rankings (still on the NCAA FCS site), there were others who got in that were not conference champs that were lower than Samford - UNH (#30), NC A&T (#28), and Weber State (#27) were easily identified just looking right past Samford at #24. So you probably need to be in the top 30, but then all the other "talking points" versus stats are what end up making the ultimate difference.

I'd still like to know the equation, though.

[thebootfitter]

One of the posters on Bisonville (I think -- maybe it was on here) a few years ago tried to replicate the SRS based on w/l records of all FCS teams and plugging in the scores per the published guidelines. I think they got pretty close.

If you ask me, it's probably not worth the effort given that it isn't a very solid system within the context of FCS football.