Wednesday, March 14, 2007

The "National Bracket": An Analysis
Of The Wisdom of the Fan Crowd

Click here (or scroll down) for today's Quickie ("Defining Choices")

And be sure to check out my fun pre-Tourney conversation with Peter Bean from Burnt Orange Nation: Part 1 is here and Part 2 is here. Cross-blog mash-up!

On to a very intriguing post:

The National Bracket: It's one of my NCAA Tournament obsessions. As a "Wisdom of Crowds" tool, this aggregate look at the combined picks of every fan entering ESPN.com's Tournament Challenge is a phenomenal peek into the collective wisdom of the fans.

Two years ago, I made a big deal out of this: If you used the National Bracket to make your own picks, you would have outperformed 90 percent (90 PERCENT!) of all individual brackets.

Now, you might not would have won your office pool (which will always go to the statistical outlier), but you would have finished VERY respectably. This, using the collective wisdom of fans everywhere.

So what does an analysis of THIS year's National Bracket reveal?

Of course, the vast majority of decisions are heavily for the seed favorite. This year, there are even a handful of outright upsets called by the majority:

Round 1:
10 Gonzaga over 7 Indiana (56-44)
Round 2:
5 Tennessee over 4 UVA (52-48)

In Round 1, there is one intriguing virtual tie:
10 Georgia Tech (50.1) vs. 7 UNLV (49.9)

And in the Sweet 16, there is one virtual tie:
3 Texas A&M over 2 Memphis (50.2-49.8)

But usually, the "wisdom" is more subtle: Merely a NOTABLE percentage of fans see something in an underdog -- matchups that are close enough to flag as a significant portion of the "crowd" spotting a potential upset. (For the purposes of my analysis, if the underdog is within 30 percentage points of the favorite, it's worth noting.) Ones to watch:

Round 1 Teams within 30% range:
Old Dominion (36.5) vs. Butler (63.5)
Winthrop (45.3) vs. Notre Dame (54.7)
G. Washington (38.0) vs. Vandy (62.0)
Texas Tech (47.1) vs. Boston College (52.9)

Notably outside the +/- 30% range:
Creighton (32.7) vs. Nevada (67.3)
VCU (24.4) vs. Duke (75.6)
Arkansas (33.9) vs. USC (66.1)
Illinois (32.4) vs. VA Tech (67.6)
Long Beach (21.2) vs. Tennessee (78.8)

Sweet 16: Now THIS is fascinating. Of the 8 Sweet 16 games, besides the 50/50 Memphis-Texas A&M game, only ONE has the underdog within even 20 percentage points of the favorite: Texas (40.6) vs. UNC (59.4). Do the fans know something the experts don't?

The theme continues in the Elite Eight: Of the four games, Ohio State is a massive 55 percentage point favorite over Texas A&M. Florida is a similar 55-point favorite over Wisconsin. Kansas is an 18.4-point favorite over UCLA. And...

...UNC is less than THREE percentage points ahead of Georgetown, 51.3 vs. 48.7. This indicates, as opposed to the other regional finals, that UNC-G'town is a virtual dead-even match in terms of popular perception. The crowd speaks: Watch those Hoyas.

In the Final Four, Florida is nearly a 24 percentage point favorite over Kansas; Ohio State is 22 percentage point favorite over UNC.

The final is predictably tight: Ohio State is the popular pick, which I respect as the choice of the "crowd," but by less than 7 percentage points over Florida. I think that makes this result a virtual toss-up.

As it should be.

As a P.S., it will be interesting to see if that vote total -- or any along the way -- change in the final 24 hours before the brackets lock and the games begin.

-- D.S.

PS: On the opposite end of the "wisdom of crowds," FanIQ has the breakdown of all of the various "expert" picks for your comparison. Please keep "expert" in quotes.

24 comments:

Cycledan said...

"I'm no statistician, but I'll bet that's within a margin of error that would make the true result a virtual toss-up."

Dan, the margin of error is the expected error of a representative sample from what the actual really is. There is no actual here. These are predictions for a game, not actual probabilities of winning the game. Even if you count the ESPN pool as a representative sample of all pools, the sample is so large that assuming the ESPN crowd is no different than the rest of the population, it will be within 0.01% of all pools in the US combined from a statistics point of view.

Dan Mega said...

The UCLA star Collison sprained his ankle. And I have UCLA in the final versus OSU. Let the panic begin =(

RevScottDeMangeMD said...

I hate these ESPN polls. Americans are stupid and don't make any sense when I see them. It makes me so angry! For example: Last year before the bowl games began, there was a question that asked "Who is the best college football team?" and every state except Michigan picked Ohio State. I mean, come on!

Long story short, it's the same with this national bracket...stupidity. Not because it's the majority of people that think no upsets will happen...but the fact that 1.) every #1 seed is in the Final 4 (never has happened and won't this year) and 2.) every #1 and #2 seed in the elite 8 except for memphis, who barely lost out to A&M (again, never has happened and won't happen this year).

I'm shocked Shanoff likes this bracket so much. I figured he would bash Americans for being dumb.

DanShanoff.com said...

good explanation, cycledan. I'll edit it out. your comment will look weird, but you've helped. Credited here!

jhawkjjm said...

As of 1:17 pm cst: Florida 57.7% Kansas 42.3%

Interesting to see if this one (my personal interest) keeps moving or not.

The one that surprises me the most is UNC's large advantage of Texas. Both teams are super young but the best player on the court will be wearing burnt orange. And Augustine is an amazing point guard who I believe went through the entire Big XII tourney with 0 turnovers. He's the key to beating Texas. Durant's always going to get his.

Dan Mega said...

revscottdemangemd said Long story short, it's the same with this national bracket...stupidity. Not because it's the majority of people that think no upsets will happen...but the fact that 1.) every #1 seed is in the Final 4 (never has happened and won't this year) and 2.) every #1 and #2 seed in the elite 8 except for memphis, who barely lost out to A&M (again, never has happened and won't happen this year).

What did you expect from the MAJORITY of people putting in their predicitions? Just because the MAJORITY has the #1's in the Final Four doesn't mean that every one of those people have ALL FOUR in the Final Four, it just means that the majority of the people picked a #1 from each bracket to be in the final four. You can try to twist the facts to justify your hatred for America, but you won't get past this stupid American.

Brian in Oxford said...

Is it too early to start a sweet sixteen, second chance pool?

durbin said...

any more news on collison's ankle? this is all espn has:

"Darren Collison sets the pace for UCLA, so the Bruins have to hope that his injury doesn't trip up their NCAA Tournament journey.

Collison suffered a grade one sprain of his left ankle on Tuesday during practice, ESPN.com's Andy Katz has learned. The guard is listed as probable for UCLA's first-round game against Weber State in Sacramento on Thursday. UCLA earned a No. 2 seed in the West Regional of the tournament."

should i be changing my picks? is a grade one sprain the worst scenario or the best?

jhawkjjm said...

Collison should be fine for the second round. As a 2 seed they should be fine (says the KU fan...)

The injury I wasn't aware of was Marquette's McNeal.

jhawkjjm said...

grade 1 sprain = mild sprain

Cycledan said...

Using the average pick is very similar to simply picking all the favorites. I used to run a pool where I wrote a Windows program to allow people to use their mouse to simply click on teams to enter their brackets - imagine that?

Now with the advent of websites such as ESPN and CBS Sportsline that allow you to run pools, my program sits in a directory of my hard drive collecting virtual dust.

However I performed a similar experiment as Dan did except we only had about 200-300 entries. Just by picking all favorites, you would do very well each year. Top 10% seems about right. It varies based upon the number of upsets. However you will almost never win by picking all favorites. The winners typically picked nearly all favorites and a few key upsets. For example if you picked George Mason and Florida to go far last year, you probably would win most office pools.

durbin said...

"grade 1 sprain = mild sprain"

great. that's what i wanted to hear. i've had mild sprains before and it doesn't last more than a couple days. still scares me a little for the likely gonzaga second-round matchup. i've got UCLA winning the whole thing on a couple of my brackets.....

rafael said...

tsk, i saw no hatred for america from revscott...
just hatred for stupid people in america, of which there are an abundance. lol

Still, you're right. the poll is odd with the ones all making it to the FF because most people have 1-2 1seeds making it to the FF, but it rotates. Overall, it makes it look like all 1seeds made it.

That last paragraph makes little sense.

Anyways, my 1seeds are Ohio State and Kansas. Florida makes 1 (of 3) of my brackets, UNC makes none.

TBender said...

@Brian in Oxford:

I am so ready for the 2nd chance pool.

GuyInTheCorner said...

Did anybody hear Doug Gottleib's theory that Marquette held back on releasing the information on Jerel McNeal till after the bracket came out so as to not damage their seeding?

I know he didn't play in their last three games, but apparently the rumor was that he'd be fine on thursday and now the team told ESPN that there was no chance for him to play in rounds 1 or 2 and little chance that he isn't done for the year.

Obviously Marquette would have motivation for this story to be true and no obligation, other than moral, to inform the NCAA of the injury, but if they did to it, are you OK with that?

shpiders said...

@ guyinthecorner:
Tom Crean ... the Bill Belichick of College Hoops?

Brian in Oxford said...

As an example, cycledan,

my pool I run gives bonus points for upsets. I won with 295 points, over about 20 poeple. I had 0 final four teams right. Only 3 people actually got 1 of the final four games right, those who had UCLA getting to the final (they finished 2nd, 3rd, and 5th).

As a contrast, picking all favorites would have given you zero bonus points, and a total of either 232 or 235 (if you picked Monmouth in the play-in game). Of my 295, I had 32 bonus plus 3 for the play-in game.

My point here is that picking the chalk was actually worse than thinking it over.

rafael said...

Unrelated: Damnit, Pete. Just shut up already! You're making it harder for me to put you in the HoF when I'm commissioner.

Craig R. said...

I can't believe I picked Florida, no matter how many times I tried not to. I guess it must be the hometown boy Billy Donovan that I have to root for.

TF said...

Did anyone watch the playin game? There were a lot of people there last night, including Niagra students who took a 7 HOUR bus trip to the game, was leaving immediately after to go back to school, then driving over to Chicago from there. Seemed counter productive, but I guess classes couldn't be missed.

My friend tried to get a couple to stay behind by offering to drive them back to school in the morning.

Anyway, entertaining game, I thought FAMU was gonna get destroyed, but they held theirs. Neither team has a chance in Hades of even hanging with Kansas, but it should be fun for them.

Joe (Dayton)

GuyInTheCorner said...
This comment has been removed by the author.
GuyInTheCorner said...

Note on the play-in game:

There was one play in the second half where that wild guy with the crazy hair for FAMU got an offensive rebound and did a pump fake / butt ram move right under the basket. He caught one of the Niagara players rigt between the legs with his hip, leading to this great exchange by the announcers...

#1: Well that's an offensive foul under the glass and it looks as if the player is still down.
#2: Yeah, he's holding his... midsection. Let's take a look at the replay and maybe we can figure out what happend.
(replay plays and you clearly see the guy get hit right between the legs and go down)
#1 & #2: Oh!
(5 second dead silence as camera shows guy on floor)
#2: Well. There's no reason to show that again.

That was absolutly priceless.

John said...

The problem with a national bracket is that everyone knows that an upset or two or five is going to happen. What we don't know is what teams are going to pull those upsets. On Friday, or Saturday morning everyone across the nation is going to wake up and think, "Wait, that team won?" Until that moment though, anyone's guess is just as good as any "expert's".

Natsfan74 said...

I absolutely agree on the Pete Rose thing. But at least he said he bet "for" the Reds to win every night. Could you imagine if he was betting against the Reds?

But given Pete's track record of I didn't bet, I did bet - but not on the reds, I bet on the Reds to win... How long until he says he bet against his Reds when the pitching matchups weren't favorable or something?