Tuesday, September 28, 2010

Do Computers Really Even Matter?

It's rather amusing to hear the assorted TV knuckleheads bellyache about the computer rankings in the BCS. They don't understand how the computers work. They think the BCS formula gives the computers too much influence. Blah blah blah.

So here's a hint: Don't worry about it.

That's right, don't worry about it. As in, the computers don't make a whit of difference when it comes to the only BCS standings that matter - the final one. It might as well not be part of the formula.

Since the adaptation of the current formula, in 2004 after the hue and cry over the 2003 split championship debacle, every single team that ranked either first or second in the two human polls (AP in 2004, Harris since 2005, and the Coaches poll) played in the BCS national title game.

Let's take a look: (Coaches-AP/Harris)

2004: USC (1-1) vs. Oklahoma (2-2)
2005: USC (1-1) vs. Texas (2-2)
2006: Ohio State (1-1) vs. Florida (2-2)
2007: Ohio State (1-1) vs. LSU (2-2)
2008: Oklahoma (1-2) vs. Florida (2-1)
2009: Alabama (1-1) vs. Texas (2-2)

It's a perfect 12-for-12. If the BCS formula never existed, these teams would've still met for the title anyway.

On the other hand, finishing first or second in the computers meant nothing. A few teams did that and ended up watching the BCS title game on TV:

2006: Michigan (T-2 in computers), lost out to T-2 Florida
2007: Virginia Tech (1), lost out to No. 2 LSU and No. 3 Ohio State
2008: Texas (2), lost out to No. 3 Florida
2009: Cincinnati (2), lost out to No. 3 Texas

For the context of this year, it means that all the speculation about Boise State's computer ratings is mere sound and fury. At the end of the season, it's entirely up to the 173 voters in the Coaches poll and Harris poll to decide whether they're willing to put the Broncos in the top 2 of their ballots.

Take last year for example. The Broncos went undefeated but finished sixth in the Coaches and Harris polls. They also finished sixth in the computer ratings. But if you take No. 2 Texas' poll points and apply them to the Broncos, they would've made it to the BCS title game to face Alabama, despite being sixth in the computers ratings.

See, that's the genius of the current BCS formula. It pretends to be a lot more than what it really is: A beauty pageant. In reality it's no more scientific than the Miss Universe contest or American Idol, but every bit as political. Your knowledge of the voters' biases will produce a better prediction of the final standings than your ability to comprehend Wes Colley's methodology.

So if you're a Broncos fan, curb your enthusiasm over this week's standings that placed Boise State No. 2. And don't get too dismayed when it fell to No. 6 as was the case a week ago. Stop wasting your time trying to figure out the Broncos' strength of schedule. But do put pressure on these voters (if you know how, and here's a little help) to give Boise State a fair shake.

It's just like writing your congressman, and probably just as satisfying and useful. But at least now you know better than to blame the machines.


Anonymous said...

If the computer polls don't carry that much weight, how come they lifted OU all the way up to the #1 spot in the first rankings this season?

The Guru said...

If you read this piece closely, you'll see the reference is to the final standings. The computer ratings have less and less influence as the season wears on.

Anonymous said...

How can that be? Don't the computer rankings always count for one third of the ranking?

The Guru said...

The variation in the computer rankings narrows among the top teams as the season wears on. Most of the unbeaten or 1-loss teams invariably end up in the top few slots, rendering the computer influence negligible.

tyamdaly said...

The computers don't matter because the BCS told them not to use scores. All of their algorithms need to use the score to be "somewhat" accurate. Take the scores out it is just a wash. Yes they do matter. For the big boys. It helps keep them on top. Humans are biased they will put whoever they want in the BCS game. Do you like this BCS system? What changes do you think that could be made to better it? There are 6 computers but only 4 are used. That is not an accurate computer score. The NCAA can't decide how to divide by 6? 1/6=.16666667. Try that on your "simulated" standings. I can do those also. And yes, all 3 people I know,(You, Me, and Brad Edwards for ESPN) that do that had Boise in the top spot. Mine and yours had the same # for Boise. Peter Wolfe needs to be investigated for fraud.

tyamdaly said...

Also, how may "shared" national championship have there been? Yea, that is what I thought. How can an organization such as the NCAA have a "shared" champion? Do they have that in basketball, golf, tennis, baseball, volleyball, lacross, or any other sport? No!!! Because its all about keeping the money in certain hands. Maybe the NFL, NBA, NHL, and WNBA can have 2 or 3 champions. More parades = more money for their cities. I love College football but their so called national championship is only "mythical". That is why there are shared championships.

Anonymous said...

Obviously your facts on the last 6 years make a good case, but isn't possible the computer portion of the formula has too much weight. Here's why. Each of the parts of the formula (two human polls and the computer rankings) count 1/3 of the final ratings. However, if any one of the four computer rankings that are used is skewed, it has a far greater effect on the final ratings than any one of 75 voters screwing up their vote in the human polls. One counts 1/4 of its portion and one counts 1/75 of its portion. This doesn't make sense to me and could greatly affect the ratings in the end. I propose counting each of the human polls twice and the computer polls once so they only represent 1/5 of the total. What do you think? Although I know things will change, this week is a great example of how the computer polls have far too much weight with Oregon being #1 in both human polls (which should be 2/3 of the total) yet still end up #2 in the total ratings.

Southern Duck said...

@Guru - Using a small sample size and a single year example (Boise St in 2009) is poor analysis. Using your 2009 example logic, Texas would have mad the CG in 2008 if they and Oklahoma would have swapped computer scores, despite being the voters #3. It is easy to cherry pick examples to show what you want to prove.

This year will be interesting since BSU is unlikely to get much more than .8 to .84 from the computers at the end of the season. So while Boise St seems entrenched now in the voters top 2, if they do run the tables, the story will be the computers this year.

The Guru said...

@Southern Duck - Yes, small sample size (1), but the fact remains that the way the formula is tilted, the computer rankings will make very little difference - this year will be no exception.

I have a full post on this that will go up shortly.

Southern Duck said...

Guru, I was referring to the sample size of six years of current BCS formula to draw a definitive conclusion that because computers haven't played a role, they never will.

To date, the lowest computer points for a top 2 BCS finisher was .89 (Florida in 2008), and they barely squeeked by Texas. Now take Boise State. In their 4 undefeated seasons, Boise State's final computer average has varied between .73 and .78. I wouldn't expect much better this season.

So realistic hypothetical. Boise State is a fairly solid #1 and averages .98 from both voting polls, but pulls in a .8 from the computers (very realistic). For an average of .92. I hopefully don't need to explain to you that a .92 is unlikely to get into the NCG. In six years, lowest CG participant so far has been 2007 LSU at .9394 and 50% of the time the 3rd place team has scored better than .92. While not impossible, Boise State, even if #1 in the polls, will need the cards to fall just right with the #2 and #3 teams in the polls in order to get in.

Anonymous said...

"the computer rankings will make very little difference - this year will be no exception."

Sorry guru, not so this season. The computers have it right, Oregon v. Auburn is the right match.

Anonymous said...

what if the top team is the same in both polls but the #2 and #3 teams are swapped in the polls?

Anonymous said...

Question: IF Auburn loses and pollsters face the prospect of TCU or not in the championship game then they may decide to take another look at Wisconsin's November.

Considering no one would want a rematch of Oregon and Stanford, and voters may worry about a repeat of '03 OU (with Auburn not winning the SEC), I think there is outside shot that UW gets moved up all the way to #2 in the human polls. HOWEVER, look at those computer scores (could you please explain UW's computer rankings by the way)...might we have the computers finally matter?

Anonymous said...

There is a simple ranking system proposed at www.acroterica.com that accomplishes most of what the BCS is intended to do without the need for a playoff system.

Take a look