THE ODDS OF THE BCS
GETTING IT RIGHT

An Introduction from Goldilocks

Hello. Goldilocks here. In a moment I will discuss the likelihood of the BCS rating system successfully (as in, “without controversy”) choosing the top two Division I-A college football teams in any given season. But first, I should address another issue everybody always asks about.

Truth is, I like bears just fine. In fact, Chicago has always been one of my favorite NFL teams, and don’t even get me started on baseball and my Cubbies! But I know you may have heard things. Basically, there was a little misunderstanding over whether I was allowed to use their place for a week, and it wouldn’t have been a big deal except the bears came home early before I had a chance to clean up. But enough about that.

Explaining the BCS Rating System

There are lots of articles that try to explain how the BCS rating system works. They talk of calculations (this number added to that number, etc.), but miss the point. The numbers are just smoke and mirrors. From its inception, the purpose of the BCS rating system was to identify the season’s top two teams. No matter what system is used, there will either be two clear choices or there won’t.

And if history is any indication, the BCS rating system should work about 29% of the time. I’ll get to that in a minute.

First, consider a brief history of bowl games and the early incarnations of the BCS. For years, bowl games went their own merry way hoping to snag that really big pairing. But in 1992, all the major conferences except the Big 10 and the Pac 10 formed the Bowl Coalition and agreed to send the top two teams to the same bowl. Conference ties were maintained, so the 1992 championship went to the Sugar Bowl (thanks to Alabama and the SEC) and the 1993 & 1994 championships went to the Orange Bowl (thanks to Nebraska and the former Big 8 conference). The Bowl Coalition then morphed into the Bowl Alliance, which abandoned conference ties in favor of establishing the championship site in advance. The Fiesta Bowl went first with the 1995 championship, followed by the Sugar in 1996 and the Orange in 1997.

The problem, of course, was that the Big 10 and Pac 10 continued their Rose Bowl commitment. If the #1 or #2 team happened to be from one of those conferences, the Bowl Coalition/Alliance merely got the best pairing available. However, the Big 10 and Pac 10 began to feel the pressure to participate. Eventually—and reluctantly—the Big 10, Pac 10, and Rose Bowl joined the other bowls and conferences, and the Bowl Championship Series (BCS) was born.

The first BCS rating system was a curious hodgepodge of polls, computers, and schedule strength. The writers’ and coaches’ polls (which were the only sources used by the Bowl Coalition and Bowl Alliance) comprised only a quarter of the entire formula. Computer ratings were the most novel component, but at first only three systems were used. Losses were counted, and a calculation of Strength of Schedule virtually eliminated minor conferences (since a weak schedule could equal four losses or more). BCS officials heralded their new rating system as having been thoroughly tested, yet only once in the next seven years would it be left unchanged.

The computer component received a lot of attention, and as a result has seen the most adjustments. Systems have been added and dropped. Victory Margins were banned. Once considered equal to human polls, they now comprise only half their weight.

The Quality Win component was invented when Florida State edged Miami for the #2 spot in 2000, even though the Hurricanes beat the Seminoles during the regular season. It nearly created disaster the first year it was used, was seriously diminished the next, and soon disappeared entirely.

Although the primary purpose of the BCS ranking system was simply to identify the top two teams, it has unwisely been used to rectify other ills beyond its original scope. In 1998, Kansas State saw a perfect season evaporate in overtime of their conference championship game. They had only loss and finished #3 in the final BCS standings, yet did not receive a BCS bowl invitation. This incident—plus the rating system’s strong bias against teams outside BCS conferences—eventually led to new rules, such as: 1) non- BCS teams MUST receive a BCS bowl invitation with a ranking of #6 or better, and 2) a team is automatically included when ranked #4 or better.

The last item led to the lowest point in BCS history. Making up a weather-delayed game, California (ranked #4) won 26-16 on the road against Southern Miss only to see Texas overtake them on the final vote. (Some voters apparently forgot a far more anemic 27-23 Texas victory over Kansas the prior week.) Even though it had nothing whatsoever to do with finding the top two teams, it had a $14 million impact. As a result, the Associated Press sued to prevent the BCS from using their poll in future calculations.

I had to admire the sportswriters for the integrity they showed in that decision. It gave me newfound respect for the media—in spite of all the bad press I got over that incident with the bears.

BCS members seek the same thing from college football’s regular season that I look for in food and lodging (which is why I consider myself an expert on this topic.) They claim their approach preserves the importance of the regular season, but that only holds true when the season’s final results are “Just Right” (with only two major undefeated teams). The rating system appears to fail when the results are “Too Hot” (three or more undefeated teams) or “Too Cold” (less than two undefeated teams). So I think the BCS rating system is mainly intended as a diversion. For all its hocus-pocus, BCS members are simply hoping things work out by themselves. And if they don’t, blame the rating system (and not those who said the system would work).

Like I said before, the BCS rating system WILL work—about twice every seven years.

Examining Results from Past Seasons—the BCS Years

If you look back at the BCS years (1998 to present), controversy ensued whenever the results were not “Just Right.” Here’s a quick review:

1998—Tennessee was a clear #1, but several teams with one loss had a claim for the second spot. Controversy was muted the first year in part by the newness of the BCS rating system, and perhaps a little by the popularity of #2 Florida State and coach Bobby Bowden. (Curiously, there were actually two unbeaten teams. Tulane played a considerably easier schedule than Tennessee and could do no better than 10th in the final BCS standings.)

1999—JUST RIGHT. Only Florida State and Virginia Tech completed their seasons without a loss, although the Hokies narrowly edged Nebraska for the second spot. (Note that this is considered a “Just Right” year in spite of #12 Marshall with their 12-0 record.)

2000—Oklahoma was a clear #1, but controversy ensued over the second selection. Florida State was #2, but lost to #3 Miami during the regular season.

2001—Even more controversy. Miami was easily #1, but Nebraska edged Colorado for the second spot, even though Colorado waxed the Cornhuskers 62-36 on their way to the Big 12 title. Lost in all this was Pac 10 champ Oregon, second in both polls but #4 in the final BCS standings. (After the bowl games, it appeared the Ducks had been the proper choice, since they whacked Colorado while Miami gave Nebraska its second straight spanking.)

There was another area in which the BCS was fortunate in its first four years. Except for 1999 (a “Just Right” year), each season ended with one unbeaten team and questions over the #2 spot. In all three years the unbeaten team went on to win, thus avoiding a second round of controversy.

2002—JUST RIGHT. Only Miami and Ohio State made it through their seasons unscathed. And of course, it took a “Just Right” result for the BCS to finally get through a winter without revising their rating formula.

2003—BIG controversy. Nobody goes unbeaten, but Southern Cal tops both polls only to find themselves #3 in the BCS standings and out of the championship. They would beat #4 Michigan in the Rose while #2 LSU beat #1 Oklahoma in the Sugar. Even more absurd was that Oklahoma got drubbed 35-7 by Kansas State in the Big 12 title game, yet remained #1. And for the second time in three years, the Big 12 sent a team that didn’t even win their own conference to the BCS title game. (Apparently not a good practice, since both lost.)

2004—The biggest controversy yet. It was the BCS’s first experience of a “Too Hot” year, and gave proof to the fallacy of the BCS argument that their approach preserves the importance of the regular season. Auburn won every game they played, but in the end it meant nothing.

2005—JUST RIGHT. And this (Southern Cal vs. Texas) was probably the most “Just Right” result in the history of college football. But before the members of the BCS strain a muscle trying to pat themselves on the back, they should realize that in the last eight years they’ve already beaten the odds.

Examining the Results from Past Seasons—Pre-BCS

To see how well a one-game (BCS-style) format will work in the future, it helps to examine the final regular season results from previous years. While a few computer rating systems were around prior to 1998, ranking teams was (and still is) primarily the result of human polls. Most noteworthy are the Associated Press writers’ poll (since 1936) and the coaches’ poll (since 1950). An excellent place to view AP poll results is at SoonerStats.com (http://www.soonerstats.com/fb/polls/index.cfm), where all previous weekly tallies can be found. However, note that the Final poll has not always included bowl game results. From 1936-1964 as well as the 1966 & 1967 seasons, the Final poll did not include bowl games. For the 1965 season and from 1968 onward, you must use the next-to-last poll in order to see the final regular season results.

This study will begin with results from 1950, which coincides with the creation of the coaches’ poll. (I have not located a similar database for the coaches’ poll, but results are basically the same as far as this study is concerned.)

As noted previously, the regular season results of any given season will either be TOO HOT (three or more undefeated teams), TOO COLD (no more than one undefeated team), or JUST RIGHT (two undefeated teams). However, overtime was not adopted until the 1996 season. Because this study examines past results and applies them to the current BCS format, it is necessary to resolve tie games from prior years. Mathematically, a team would have a 50-50 chance of winning (or losing) any game that ended in a tie.

So now it’s time to assign each year to one of those three categories of Season Type. (Years marked with an asterisk require explanations that appear at the end of this list.)

TOO HOT—1951, 1954, 1958*, 1962*, 1965, 1968*, 1969*, 1970*, 1971, 1973, 1979, 1987, 1992, 2004.

TOO COLD—1950*, 1953*, 1957, 1959, 1960, 1961*, 1963, 1967, 1972, 1974*, 1975*, 1977, 1978, 1980, 1981, 1982*, 1984, 1985, 1989, 1990, 1998, 2000, 2001, 2003.

JUST RIGHT—1952, 1956, 1964, 1966*, 1983, 1986, 1991, 1994, 1995, 1996, 1997, 1999, 2002, 2005.

And here’s yet another category that doesn’t quite fit the others:

JUST RIGHT yet somehow STILL WRONG*—1955, 1976, 1988, 1993.

*Notes and Explanations, Part 1—Certain years require specific explanations that can be found in Part 2. But first, we should examine how it’s possible for a result to be JUST RIGHT and yet STILL WRONG. And for some reason, the JUST RIGHT/STILL WRONG phenomenon seems to emanate from the border between Maryland and West Virginia.

The first occurrence was 1955 when Oklahoma and Maryland were the only undefeated teams. Each had 10-0 records, but Michigan State (8-1) held the #2 spot ahead of Maryland. In 1976 only Pittsburgh and Maryland had perfect 11-0 records, but Maryland was #4 behind Michigan and Southern Cal (both 10-1 and facing each other in the Rose Bowl). In 1988 the phenomenon shifted to West Virginia, where only #1 Notre Dame and the #3 Mountaineers were undefeated, with Miami holding the #2 spot. West Virginia was again one of only two teams with perfect records in 1993 (Auburn was on probation and not eligible). The coaches’ poll had Nebraska (11-0) on top followed by West Virginia (11-0) and Florida State (11-1), but the AP had Florida State in the top spot followed by Nebraska and West Virginia. That year the Bowl Coalition combined the two polls, leaving Nebraska on top but dropping West Virginia to the #3 position behind Florida State.

The two teams’ experiences have curious parallels. In their first instance (1955) Maryland was #3 but got to face the #1 unbeaten team in a bowl game, while the second time (1976) Maryland did not get to play #1 (and unbeaten) Pittsburgh in a bowl game and thus had no shot at the national title. In 1988 #3 West Virginia faced #1 Notre Dame for the national championship, but in 1993 got no chance to play #1 (and unbeaten) Nebraska. So Maryland and West Virginia have both—twice—been one of two undefeated teams. The first time they got to face the other undefeated team, and the second time they did not. In none of those four seasons were the results JUST RIGHT. (Also, both teams lost both games, perhaps justifying their lower rankings.)

Sometimes having two unbeaten teams is not enough. (The 1993 season provides the best evidence, since it falls under the first incarnation of the BCS.) Curiously, this category does not even include examples such as Tulane in 1998, Marshall in 1999, or Utah in 2004—teams with perfect seasons that were not even remotely considered for the BCS championship game.

*Notes and Explanations, Part 2—The following notes are for individual years. All ties are considered unresolved overtime games with a 50% chance of being won or lost.

1950—The Final AP poll shows Army at #2 with an 8-0 record, but the vote does not reflect their subsequent 14-2 loss to Navy.

1953—This season is JUST RIGHT if Notre Dame wins in overtime, TOO COLD if they lose. (Add 0.5 to the JUST RIGHT total.)

1958—Because of a large number of tie games among the top teams, there are a variety of possible outcomes. In both polls the #1 team is LSU (10-0) followed by Iowa (7-1-1), Army (8-0-1), Auburn (9-0-1), and Oklahoma (9-1), with Air Force (9-0-1) #6 in the AP but only #8 in the coaches’ poll. For results to be JUST RIGHT, only one of three teams (Army, Auburn, and Air Force) can remain undefeated. At this point, note that Iowa and Air Force played to a tie. Air Force is ranked low because of their weaker schedule (including powerhouse universities such as Denver and Detroit). If Air Force is the only other undefeated team (other than LSU), they must also rise to the #2 spot—very unlikely. If Iowa wins, either Army or Auburn must win and the other lose. Yet Iowa (7-1-1) already ranks ahead of both and would improve to 8-1. That means the JUST RIGHT/STILL WRONG outcome would likely occur even if only one of the three undefeated but tied teams (Army, Auburn, and Air Force) remains unbeaten.

1961—This season is JUST RIGHT if Ohio State wins in overtime, TOO COLD if they lose. (Add 0.5 to the JUST RIGHT total.)

1962—If #4 Texas loses in overtime there would only be two unbeaten teams, BUT Wisconsin (8-1) already ranks ahead of Mississippi (9-0) in both polls. The result is either TOO HOT or JUST RIGHT/STILL WRONG.

1966—Even though there are three undefeated teams, this year is counted as JUST RIGHT because the top two teams (Notre Dame and Michigan State) tied each other. With overtime, the winner would claim the top spot and Alabama (10-0) would move ahead of the loser.

1968—There are two teams with perfect records (#1 Ohio State and #3 Penn State) and two with ties. Southern Cal must lose their overtime game AND Georgia must lose either of theirs for results to be JUST RIGHT, which occurs randomly three times out of eight. (Add 0.375 to the JUST RIGHT total.)

1969—This season is JUST RIGHT if Southern Cal loses in overtime, TOO HOT if they win. (Add 0.5 to the JUST RIGHT total.)

1970—This season is JUST RIGHT if Nebraska loses in overtime, TOO HOT if they win. (Add 0.5 to the JUST RIGHT total.)

1974—Although there are two undefeated teams listed, #1 Oklahoma was on probation and not eligible.

1975—The Final AP poll shows Texas A&M at #2 with a 10-0 record, but the vote does not reflect their subsequent 31-6 loss to Arkansas.

1982—SMU is ranked #4 with a 10-0-1 record. An overtime win would have to move them past #2 Penn State (10-1) and #3 Nebraska (11-1), which seems unlikely.

Summary and Observations

According to the original list, there have been 14 JUST RIGHT seasons since 1950. There were also four seasons with a 50% chance of producing a JUST RIGHT result depending on the outcome of overtime periods that were never played, which raises our total to 16. In addition, the 1968 season had a 3-in-8 chance for a JUST RIGHT result. That gives us a final total of 16.375. This number, divided by the TOTAL number of seasons in our sample (56), equals the odds of a controversy-free BCS season.

16.375 / 56 = 29.2%

By the way, there were no—ZERO—seasons that could be considered JUST RIGHT from 1936-1949. Including those results would bring the odds down to less than 1-in-4, but the seasons are not representative. The war years in particular were unusual, as some schools cancelled football while other programs only existed during this time (like the Iowa Pre-Flight team that finished #2 in 1943).

It is interesting to note that results have been JUST RIGHT in 7 of the past 12 years, and 8 of the past 15. In fact, taking a closer look at those years can help us better understand where we are now.

In 1991, Miami and Washington split the championship. (They could not play each other in the postseason, as Pac 10 member Washington was committed to the Rose Bowl.) The Bowl Coalition began the following season. During the Bowl Coalition/Bowl Alliance years of 1992-1997, the top two teams outside the Big 10 and Pac 10 conferences met in the same bowl game. Incredibly, four straight seasons (1994-1997) had JUST RIGHT results. Yet three of those years (1994, 1996, & 1997) did NOT pair the two unbeaten teams together because one was committed to the Rose Bowl (Penn State in 1994, Arizona State in 1996, and Michigan in 1997). It must have seemed like every year there were two unbeaten teams, and only the Rose Bowl stood in the way of pairing them up.

Truth is, those years were anomalies.

If you think earlier years are not representative due to shorter schedules, think again. The 11-game schedule began in 1971. Using only the 35-year period of 1971-2005, we find exactly 10 JUST RIGHT results (28.6%). In fact, there were NONE in the first 12 years (1971-1982). One can only imagine what pressures the BCS would experience after 12 years of never getting a single JUST RIGHT result.

Conclusion

Hopefully you now understand the BCS a little better. And please, don’t be too hard on the boys and girls of the BCS. They are simply drifting along, hoping every year to find a season with results that are JUST RIGHT.

But take it from me. Sometimes you just need to get your own place, buy your own food, and maybe even arrange a payment plan for a mattress and box springs. It’s all well and good to freeload for a while, but sooner or later you’ve got to show a little maturity and deal with the problem. Otherwise, if you’re not careful a bunch of bears will come along and catch you napping. And sometimes, the bears are pissed.

Goldilocks (February 2006)

To contact Goldilocks: g-locks@hotmail.com