Bucks Diary

Monday, March 24, 2008

NCAA Weekend: Lessons learned

The PVOA system, applied to the NCAA tournament, had an okay showing over the four day weekend. In fact, for an untested, purely theoretical, system I can't complain. We hit on 36 of 48 games, including 15 of 16 on Thursday, and 12 of the 14 games in which I had at least one team alive over the weekend. Friday was a disaster, but we'll get to why.

Actually, if you count only games in which there was a pick available, we hit on 36 of 46. That's a little over 78%. Its far better than I've done since the days when I actually followed college basketball.

But lets talk about the misses. Here they are.

The missed pick listed first:

St. Joseph's-Oklahoma
St Mary's-Miami FL
Drake-Western Kentucky
Arizona-West Virginia
Connecticut-San Diego
Duke-(West Virginia.. I had Arizona here)

The problems, as you can see, have mainly been at the margins. Here are some things I learned:

1. Field Goal Percentages not indicative
If there was no clear blue water between two teams using PVOA, then I went with the team that had the higher effective field goal percentage. This was complete folly, based entirely upon this guy's analysis ("Even with the small sample, this is fairly strong evidence that the better shooting team has a distinct advantage in matchups of similarly seeded teams."). I lost every single one of the "better shooter" games, save for Michigan State-Pittsburgh.

2. Beware of the "Even" Mid-Majors
Moreover, and perhaps more significantly, in every one of those games the better shooting team ended up being the Mid-Major: St. Joe's, St. Mary's, Drake. Next year I will lean against the Mid-Majors in a close call. Obviously the statisical "level playing fields" are harder to achieve when comparing Mid-Majors of relatively equal strength to the more middling "Big Conference" teams such as Oklahoma and Miami. That needs fixing.

Indeed, once we got to the weekend and had a lower percentage of mid-majors, the PVOA analysis picked up steam once again. So, again, the method of doing Mid-Major evaluations needs tweaking.

3. Beware of too many trifectas
If I had it to do all over again, I would break stalemates by picking the team that relies less on the three pointer. If you go on the "Point Distribution" outline on KenPom.com, you will see that you cannot find a single tournament team among the top, 75 I believe, most 3 point reliant teams in the country. I noticed that last night. Of course, Texas and Tennessee are still rocking, and both of them rely on the three... but by and large its dangerous to do so.

4. Offense will leave you
Next year I am going to place an added weight on defense. Most every one of the surprising teams to go down were teams heavily reliant on offense (and a lot of those were also, as I stated above, heavily reliant on threes for offense), and just average on defense. Clemson, Vanderbilt, Connecticut, etc.

4. Apply some common sense
This is a fine line. Once you start letting all of your "hunches" loose, suddenly you have no system anymore. But, I suppose in cases like Indiana and Arizona, I should go with the overwhelming evidence of a trend. The trend in Indiana's case is that they quit on Dan Dakich. The trend in Arizona's case is that they played well against a very tough schedule, but they couldn't pull out wins.

Where I'm At

I feel like I'm doing alright... I'm hitting on more picks than all but two in my pool, but circumstances have to fall into place precisely for me to end up in the money.

Two games will go a long way toward deciding that:


This is where the rubber meets the road with this system. National polling and media sentiment would have you believe in Tennessee, and local experience had most around here skeptical of Wisconsin.

But the numbers are pretty clearly on the side of Louisville and Wisconsin, especially Wisconsin. And I'm really not worried about them. If they lose, it will have been because Davidson played a spectacular game. The Badgers will show up and give a good account of themselves.

Louisville and Tennessee is a little dicier. Louisville is slightly too dependent on the three pointers for my taste, but, thankfully, so is Tennessee. So it may come down to defense, and that is edge to Louisville. Plus, historically, Bruce Pearl teams sprint to the field of 16 and then run out of gas. And Tennessee is the worst team left in the tournament at preventing their opponent from getting second chance rebounds on the offensive end.

NBA talk will return tomorrow or the next day


At March 25, 2008 at 8:26 AM, Anonymous Anonymous said...

If you had just taken the favorites in every game you would have gotten 35 right.

Beware of "confirmation bias" -- just because something worked over a recent sample set doesn't mean it works. And when it comes to predicting sports results, almost any random system works about 50% of the time vs the spread and 75% straight up.


Post a Comment

Links to this post:

Create a Link

<< Home