The predictable unpredictability of NCAA upsets

Before the tournament started I told you I wasn’t going to waste time this year over-analyzing the tournament because no matter how good your system you can’t pick them all right.  The first round always features around 7-9 upsets, with upsets defined as the less efficient team defeating the more efficient team.  So it wasn’t worth the time.  The best result that the best picking system in the world could consistently get right would probably be around 24.  Anything above that is pure luck, anything below that and the system is faulty.

That’s what happened again in this tournament.  Using the Ken Pomeroy efficiency ratings, the best objective based NCAA “power” rankings I know of, as our guide, there were 8 upsets in the first round of this tournament.  8 times the statistically weaker team defeated the statistically stronger team.

(SIDEBAR: You might be thinking, why couldn’t someone develop a system that identifies upsets?  The answer is such a system wouldn’t make any sense.  You would be predicting games using criteria other than the proven capacity to win basketball games.  Such a system will only fuck you up, because you will end up picking upsets where they don’t exist and thereby getting more wrong than you ought.  A terrific example of this is my Harvard Sports Analysis debacle. The HSA writer used regression analysis to identify characteristics of teams likely to pull upsets.  The problem with that is, by definition, the teams likeliest to pull upsets in any given game are still — by definition — the weaker team in the given contest.  So, like  a mental patient, you end up picking at invisible flies.  I fell for this illogic and blindly went along with the 3 teams HSA identified as the likeliest upset teams.  All of them lost badly.  And it stands to reason.  Each of them, with the possible exception of Sienna, were the proven weaker teams in their given contests.)

Back to the discussion…

Some of those “upsets” happened between statistically close teams, like the Texas A&M victory over Utah State, or the Washington victory over Marquette.

But some of the games were just simply not predictable.  Ohio over Georgetown.  Cornell over Temple.  Why would you choose either of those winners?

So you see why I gave up with trying to find a magical stats based NCAA Tournament picking system.  None can exist because of the iron rule that around 25% of the first round games will be upsets. (I keep saying the “71% rule” is applicable, but that’s the NBA.  I went back and looked and in the NCAA tournament the rule is closer to 74%, probably because of the more uneven distribution of talent in collegiate basketball).

The Pick Wars:  Obama Rules

President Obama must watch a lot of college basketball.  He outpicked the 74% rule, meaning he picked well and got a little lucky, predicting correctly 25 of 32.  That was one better than a straight Pomeroy bracket.

Scott Van Pelt picked to the Pomeroy limit, getting 24 correct.

My “15 minute method” landed me in the “okay” zone of 22 correct.  Its exactly what you would have gotten if you were one of those office lame-o’s who just pick according to seed.

You get in the “okay” zone by being conservative but trying to cherry pick a few upsets (and throwing in a few stupid “notion” picks).  This system can reap rewards, but only if you hit on most of your upsets.  On my upset ledger, I went with the Harvard boys advice and got badly burned.  Sienna — my “guarantee” — could not beat a weak Purdue.  But to be fair, I think I would have picked whomever played Purdue as an upset special no matter who it was.

Bill Simmons illustrates the hazards of trying to go for it all with wild upset picks.  Its sort of like that board game, I forget what its called, where wild guessing is harshly penalized because unless you’re very lucky, this strategy has no chance of succeeding.  You end up missing on the easy ones and also missing on the close ones.  You miss on your upset and then you absorb defeats on the real upsets.

Bill Simmons method is the one I used for years.  I routinely finished last or near the bottom in every pool.

The Fantasy Geek, I don’t know what he did.  His picks were just idiotic.

And if you’re Joe Lunardi, a guy who studies the NCAA pool like a pre-med student studies for the MCATs, don’t you have to do better than my 15 minute method?

Here are the Pick War results.  This might be the last one, because it gets kind of unfun from here because of “already eliminated” losses. Maybe I will.

And hey, how about the Big Ten?  Supposedly so weak this season, they advance every team except Minnesota, who probably should have swapped places with Illinois anyway.

The Big East?  Oooh, that hurts…

NCAA Pick Wars – First Round Final Results

1. President Obama — 25 of 32

2. Pomeroy Efficiency Picks — 24 of 32

3. Scott Van Pelt — 24 of 32

4. Picking According  to Seed — 22 of 32 (69%)

4. Ty Will — 22 of 32

4. Michele Beadle — 22 of 32

4. Colin Cowherd — 22 of 32

8. Joe Lunardi — 21 of 32

9. Dick Vitale — 20 of 32

10. Bill Simmons — 19 of 32

11. Matthew Berry — 17 of 32

Footnote:  One thing I forgot to mention.  Did you notice that on Day One there were exactly 4 “statistical” upsets (75%) and you got the exact same result on Day Two.  I bring this up because if you follow this blog you remember I did an experiment tracking upsets in the NBA and was dumbfounded how night after night exactly 30% of the games would feature statistical upsets.  That kind of “predictable randomness” if you will, is what fascinates the hell out of me and makes me interested in statistics.  I used to hate numbers now I love them because they can tell you really interesting  things about the strangely predictable outcomes  of human activity.

Real quick example.  It fascinates the fuck out of me that season after season exactly the batting average for balls put in play in Major League Baseball is right around .290, when balls are struck fair, and stay in the park, the historical average number of  them that will result in hits will be  .290%.  It never changes. Why .290%?  What does that say about human capacity?  Is that the exact limit of human defensive ability in the sport of baseball?

For some reason that kind of thing fascinates me.

2 Responses to “The predictable unpredictability of NCAA upsets”

  1. The Predictable Unpredictability of Ncaa Upsets Courtside Analyst Says:

    [...] picking system. None can exist because of the iron rule that around 25% of the first …This Blog Cancel [...]

  2. satellite direct blog Says:

    satellite direct blog…

    [...]The predictable unpredictability of NCAA upsets « COURTSIDE ANALYST[...]…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


Follow

Get every new post delivered to your Inbox.

Join 25 other followers

%d bloggers like this: