Football for Nerds: An Interview With Accuscore's Zach Rosenfield

Written by Zach Rosenfield & Gatorsfirst.com Co-Founder James Brown

With the college football season rapidly approaching, Gatorsfirst.com is presenting a different take on the 2009 season every day for the entire month of August. We call it our ''Full Month of College Football Previews''. As part of the project, we are having fellow bloggers and other college football personalities from around the SEC and the rest of the nation give us their take on the current state of college football. This approach will allow us to bring you a much more in-depth look at the 2009 season than we could possibly provide on our own.

The next in this series is an interview with AccuScore's Zach Rosenfield. AccuScore is an analytical method to predicting the most likely outcomes of sporting events. I always love any type of college football argument that is backed by data, so it seemed a natural fit for our season previews to talk with him about how his computers predict the national and conferences races to unfold this season.

James Brown, Gatorsfirst (JB):Before getting into why the Gators are this season’s favorite to win the BCS Championship, can you tell us a little about your website and how your projections work?

Zach Rosenfield, AccuScore (ZR): Sure thing. AccuScore is a simulation engine that uses players, teams and coach’s statistics and past performances to project the outcomes of games and seasons. We simulate games one-play at a time under real game conditions that factor in more than 100 variables. We then repeat the process 10,000 times and are able to get a very good indication of what we think will happen in a game.

In March, we transitioned into a premium site to meet the various needs that fans have leading up to their game. Each fan has an interest in pre-game information for diversified reasons. We are set up to meet those needs. On the site, you can get everything from speculative articles about the value a player might have, to fantasy information, as well as our probability for each game’s final score.

(JB): As you get data for the current season, say in early October, how much are you using your preseason projections? In other words, if a team has a poor stretch of games in September, are you drastically changing your model to predict how they finish the rest of the season?

(ZR): Every time a play happens, our database gets that much bigger. Games are like movies, with the players being the lead actors. You have to take into account all the variables that support that play. For instance, if Tim Tebow completes a 10-yard pass to Riley Cooper, there is a slew of information we process around that such as: what quarter was it in, what down was it, what part of the field was it on, was it at home or on the road, what was the score at the time, was it a passing situation or a run situation, was it on grass, was it at night, was it raining, etc.

Teams who consistently put up bad numbers usually have a bad stretch. It is inversely proportional to teams that are playing well. However, because we focus in the long sample, we look for sustainability with a team, its players and coaching tendencies.

Injuries also play a big role in this. If a player is hurt or questionable, it will have an immediate impact on our simulations and the team’s win probability.

(JB): Do you predict individual player statistics for each game? What about for the entire season? I’m looking for any advantage I can find in my college fantasy football league.

(ZR): We do project player stats for college football, but on a game-by-game basis. We are going to be expanding that in the near future as college fantasy leagues start to gain momentum. For the time being, our game forecast pages give a complete breakdown of how we think each player will do. We also developed a custom simulator that allows our users to tinker with the stats and see the fallout it brings. Every stat in a game has an impact on each player in the game. It’s a lot of fun.

(JB): Okay, enough with the general questions, we are all dying to know: How, exactly, did you rate Florida as having the highest probably of reaching Pasadena?

(ZR): Come on…do we really need a computer simulation engine to predict that one? When the defending champion returns all their starters on defense and their offense is led by the biggest impact player in the game, it’s a layup.

We have Florida’s record to be 11.5 wins and 0.5 losses. The only reason they have any marks in the loss column is based on the laws of compounding probabilities. The Gators will be significantly favored in every game they play this year and with good reason.
They also got a huge break in scheduling and PR. The nation has a love affair with Florida and Tim Tebow and they will not get grief for an otherwise soft non-conference schedule. On the other hand, Penn State is going to get killed in the media for what they have on the books in September. Additionally, not playing Ole Miss or Alabama is very good for the projections. Florida would be favored in both those games, but not having to play them is even better when your goal is Pasadena.

Top to bottom, the SEC is down this year. Lots of quality programs, but none of them are hitting at full speed right now. People put too much stock in bowl games and I think there is more to learn from LSU’s games against Ole Miss and Arkansas then what happened in the Chick-fil-A Bowl. We have done a preliminary simulation on the Gators' trip to Baton Rouge and Florida is winning by a comfortable double digit margin. Also, Florida has Urban Meyer and LSU has Les Miles. End of story.

(JB): How accurate are these predictions from year to year? How did you rate Florida a year ago? What team had the highest probability of reaching Miami before the 08 season?

(ZR): We had a good year last year where we had 79% accuracy with picking the winner, which was up 2% from 2007. Admittedly, college football is the easiest sport to pick winners, but we still performed over 9% better than what the notable handicapper’s average was. We are proud of this because while it is easy to project Florida over Hawaii, it is not so easy when forecasting the MAC.

Last year was a lot more challenging in projecting our national championship finalists. There were a lot of good horses in the race and the season broke perfectly for the teams that made it. I don’t have last year’s numbers in front of me, but Florida owes a great deal to Alabama for posting a 12-0 regular season. If the SEC winner was 9-3, it would have been harder to take Texas in the BCS standings in the last week of the year.

(JB): What other points would you like to make about the national race?

(ZR): The number one point I want to make about this year’s race is it is not wide open. It’s actually very top heavy and it’s not hard to decipher who has the inside track. I will say that we are projecting it to be another SEC vs. Big 12 BCS Championship Game. We don’t see how Florida is going to lose during the regular season and we are expecting big things from Ole Miss in the West.

USC is going to be down this year and will struggle with 8 new starters on defense and a quarterback in Aaron Corp who is not the real McCoy. Also, they have a horrific schedule with road games at Ohio State, Notre Dame, Cal and Oregon.

The ACC will once again be the fulcrum of mediocrity and we feel Virginia Tech is very overrated. We have them at 8-4. Miami is most likely going to start 0-4 and the slew of average teams will most likely beat themselves up.

The Big 10 is poised to pick up the pieces if we see a season like 2007 because they have two slugs in that conference. Ohio State will be good, but replacing Hartline and Robiskie is going to be a much bigger deal than replacing Beanie Wells. Penn State is a good team with a gift of a schedule that could work out for them. If they beat Ohio State in Happy Valley, they very easily could go undefeated. I want to note that if there are three undefeated teams and any of them are from the Big 10, the Big 10’s entry will get shut out.

You have the identical situation in the Big 12 South this year as you did last, only replace Texas Tech with Oklahoma State. Whoever wins between Texas and OU in Dallas (we have UT as a 54% favorite) has an inside track at Pasadena. However, it can all once again be undone if Texas beats OU and loses to OSU on October 31. OSU would most likely be undefeated when they go to Norman. That can once again set the chaos of last year back in motion.

It should be noted that USC is very vulnerable this year and we think a Big 10 Champion with one-loss would not get the nod over a Big 12 Champion with one loss. We have penciled in Florida to go undefeated and if they should lose in the SEC Championship, whoever beats them most likely would be in.

(JB): What other teams stand out in the SEC? Is there a sleeper team in the conference that you think pundits are overlooking, but your computers really like? Why is this the case?

(ZR): I might ruffle feathers here, but I am ok. The SEC is down this year. The best matchup you will see is Alabama at Ole Miss and that just sounds weird. I know Gator fans will say “what about LSU?” and my answer is “yeah, what about them?” The best team LSU had under Les Miles lost two games and needed miracles against Alabama and Auburn and some home cooking to beat your Gators. So like I said, we have Florida by double digits in that one.

To be honest, nothing jumped off the page when looking at the SEC when it comes to sleepers. We have LSU at 8-4 and Georgia at 7-5. I will say the one that caught our eye was Alabama. We think they are not going to be as good as their projections are at 10-2.

Not to beat a dead horse, but we are high on Ole Miss. Our preliminary simulations show they will beat Alabama and Nutt will get the three-peat against Les Miles. As weird as it is to say, it is highly conceivable that Ole Miss and Florida can both be undefeated and play in the National Semifinals by way of the SEC Championship Game.

But just to be clear, Ole Miss is projected to go 11-1 or 10-2, it varies. Ironically, they are winning preliminary simulations against all the teams they play, but the laws of compounding probability put dashes in their loss column when looking ahead at the entire season.

(JB): Can I nail you down on a predicted finish for Tennessee, South Carolina, Arkansas, and Auburn?

(ZR): Terrible. Stagnant. Improved. Gross.

I think it was time for Phil Fulmer to move on from Tennessee, but a quick coaching change is not necessarily the answer. Like it or not, Lane Kiffin is an unproven commodity despite his resume. I can’t give him credit for the success USC had because it didn’t break stride when he left and his tenure in Oakland didn’t show enough. But the real question in Knoxville is do they have talent? We expect them to rebound at 8-4, but those 8 wins will not come against the brass of the SEC.

The Spurrier experiment is going about the same as the Holtz experiment for South Carolina. After playing the Gators tough back to back years, they really laid an egg last year and did nothing against LSU in a marquee game the year before. USC is one of those middle of the road teams that should have shown something by now, and the fact they haven’t leads us to believe it’s because there is nothing to show. We have them at the bottom the East.

Even if Houston Nutt stayed in Fayetteville, Arkansas would have had a forgettable 2008 campaign. You cannot ignore how Arkansas finished down the stretch and being a “tough out” finally paid off with very nice comeback against LSU on Thanksgiving weekend. Arkansas is a hard team to project because we really don’t know what we are getting with Ryan Mallet under center. We do project them at 5-7 and potentially 6-6.

Ever since being shut out of the BCS Championship Game, it has been one disaster after another for the Tigers and are looking at 5-7 this season. We are in agreement that getting rid of Tuberville was a terrible decision only to be outdone by the hiring of Gene Chizik. Auburn has always been stout on defense, but it’s their lack of patience with an offensive system that has really set them back. It’s going to be a long few years for Auburn.

(JB): I’m afraid to ask the question, for many different reasons, but… are the Gators going to go undefeated?

(ZR): I really can’t stress how hard it is for a proper simulation engine to project a team to go undefeated. With that in mind, it should be noted that Florida is as close to the hole as any team can get. Their projected ½ losses is a full game better than last year’s highest probability team (USC) who we had going 10.5 and 1.5.

I am not going to sugar coat it. If Florida loses a game in this year’s regular season, it would not only shock me, but it would be a true embarrassment. You side step Ole Miss, don’t see Alabama, LSU is well…LSU and Georgia will assume their familiar position in Jacksonville. Sprinkle in some softies while touting the best defense in the land and tell me if you feel any different about the potential for a Gator loss. And we didn’t even talk about the offense! 

Tim Tebow sure does get a lot of PR and some of it can be a bit annoying at times (cue Thom Brennaman), but PR isn’t what lead the Gators past Alabama and PR isn’t want lead Florida on a soul crushing drive last in the 4th quarter against OU. Tim Tebow is what got it done and you can’t take that away from him.

(JB): So, can you make a pick for the National Championship?

(ZR): Texas/Oklahoma vs. Florida. Although our preliminary projections show Texas beating both OU and OSU, our computer is actually choosing Oklahoma to come out of the Big 12. Let me explain.

Texas is a 54% favorite against OU and 51% against OSU. Oklahoma is a 46% favorite against Texas, but a staggering 81% favorite against OSU. Texas is on the right side of two virtual coin flips. Despite being only 46% against Texas, that game is a virtual coin flip, but the fact the Sooners are a 4 to 1 favorite against Oklahoma State is what makes them our pre-season favorite in the Big 12.

I want to thank you guys for taking the time to chat with me. I really like what you guys do and think it is great that you bring great content to your fans. As an avid college football fan, I understand the value of sites such as yours.

Moreover, I am a die-hard Sooner fan who was at last year’s BCS game. I have nothing but great things to say about Florida and your unrelenting support for your Gators. I hope OU gets at redemption and if not, my message is this to the Florida Gators… “Beat Texas!”

Zach Rosenfield is an AccuScore analyst who focuses on NCAA Football. In addition to overseeing AccuScore.com's daily editorial content, his opinions and analysis has been published by Wall Street Journal, Yahoo!, USA Today, Los Angeles Times and more. In addition to his published analysis, Rosenfield has appeared on more than 40 radio stations nationwide, including being a regular on ESPN Radio, Sporting News Radio and numerous shows on Sirius/XM.

You can view the original interview here: http://gatorsfirst.com/index.php/florida-gators-sports-news/college-football/predicting-the-2009-college-football-season.html