Wednesday, March 15, 2017

1-68 seed list analysis

NCAA rank listed first, then my ranking in parentheses.

1 Villanova (Villanova)
2 Kansas (Kansas)
3 North Carolina (Gonzaga)
4 Gonzaga (Duke)
5 Kentucky (Arizona)
6 Arizona (North Carolina)
7 Duke (Kentucky)
8 Louisville (Baylor)
9 Oregon (Oregon)
10 Florida St (Louisville)
11 UCLA (UCLA)
12 Baylor (Butler)
13 Butler (Florida St)
14 Florida (Florida)
15 West Virginia (West Virginia)
16 Purdue (Notre Dame)
17 Virginia (Iowa St)
18 Minnesota (Purdue)
19 Notre Dame (Virginia)
20 Iowa St (SMU)
21 SMU (Cincinnati)
22 Cincinnati (Minnesota)
23 Maryland (Creighton)
24 Creighton (Michigan)
25 St Mary's (Wisconsin)
26 South Carolina (Maryland)
27 Michigan (Miami)
28 Dayton (Northwestern)
29 Wisconsin (St Mary's)
30 Miami (Virginia Tech)
31 Arkansas (Wichita St)
32 Northwestern (Seton Hall)
33 Vanderbilt (Arkansas)
34 Seton Hall (VCU)
35 Michigan St (Dayton)
36 Virginia Tech (Rhode Island)
37 Oklahoma St (Marquette)
38 Wichita St (Middle Tennessee)
39 Marquette (Providence)
40 VCU (Oklahoma St)
41 Xavier (South Carolina)
42 Providence (Michigan St)
43 Wake Forest (Xavier)
44 Rhode Island (Kansas St)
45 USC (Vanderbilt)
46 Kansas St (Wake Forest)
47 Nevada (USC)
48 Middle Tennessee (Nevada)
49 UNC-Wilmington (UNC-Wilmington)
50 Princeton (Princeton)
51 Bucknell (Vermont)
52 East Tennessee St (New Mexico St)
53 Vermont (East Tennessee St)
54 Winthrop (Bucknell)
55 New Mexico St (Winthrop)
56 FGCU (Iona)
57 Kent St (FGCU)
58 Iona (Northern Kentucky)
59 Northern Kentucky (Texas Southern)
60 Troy (Kent St)
61 Jacksonville St (Troy)
62 North Dakota (North Dakota)
63 Texas Southern (South Dakota St)
64 South Dakota St (Jacksonville St)
65 UC-Davis (New Orleans)
66 North Carolina Central (Mt St Mary's)
67 New Orleans (UC-Davis)
68 Mt St Mary's (North Carolina Central)

Biggest differences:
South Carolina 15 (NCAA 26, me 41) - this ranking was nonsensical
Vanderbilt 13 (NCAA 33, me 45) - you know, losses are supposed to count too
Middle Tennessee 10 (NCAA 48, me 38) - this is part of the general disrespect for mid-majors
Rhode Island 8 (NCAA 44, me 36)
Dayton 7 (NCAA 28, me 35)
Michigan St 7 (NCAA 35, me 42)
Wichita St 7 (NCAA 38, me 31)
Virginia Tech 6 (NCAA 36, me 30)
VCU 6 (NCAA 40, me 34)

Takeaways:
- The SEC's efforts to schedule up did pay off more than I thought it would.
- I don't know what to do with the A-10
- Losses don't matter as much to the NCAA given what Vandy and MSU got

I got 68/68 teams...I missed a team's overall rank by an average of 2.838 spots per team.

I wonder how 2.838 compares to other bracketologists.  If you want, give me your 1-68s and I can calculate.

3 comments:

Bryan Wilson said...

I commented before about my computer seedings - I'll give you my numbers if you're interested.

I scored 333 on the bracket matrix. I would have scored 341 if I just used Saturday night's seeds and ignored Sunday's games entirely. That was one of my main takeaways.

I got 68/68 teams correct also. My average overall number difference was 2.41. My average difference would have been 2.29 using Saturday night's seeds.

My auto-calibration system says the biggest changes between the 2016 and 2017 resume weights were the following:
- 2017 had higher emphasis on % vs. top 50, top 100, etc. rather than raw # of wins
- Number of top 100 wins matters more in 2017
- Road/neutral % is a nearly meaningless metric in 2017
- Marquee wins / bad losses meant a little more in 2017
- Raw Win % did not matter as much in 2017
- Slightly lower SOS weight in 2017 but still quite high

Thanks again for a year of entertaining bubble talk.

Bryan Wilson said...

And to comment on your main misses -

South Carolina: this one perplexes me too. My guess is there was pressure to seed them higher than the other SEC teams (Vandy, Arkansas) and they were also given pretty generous seeds.

Vanderbilt: My system actually had them at the highest 8 seed. I knew that was too high but I also suspected it would be closer to correct than most people's predictions. The historic precedent was Georgia getting in many years ago at 16-14 with a ridiculous schedule. Not only did they get in, they were given a pretty high seed. I expected a repeat.

Mid Tennessee / Wichita / etc. The last few years there has been an enormous emphasis on SOS and # of top 50 wins. The over-reliance on these two metrics just cripples any mid major's seed and at large hopes. My system's biggest miss was Wichita St (i had them at 30 overall vs. 38 actual). I was actually surprised Gonzaga got a 1 because they lag in both of these metrics. Maybe the rules do not apply the same for teams close enough to the top seed lines.

Andrew said...

"Road/neutral % is a nearly meaningless metric"

That is frightening.

I think everyone got burned with Sunday changes. It's even worse that they said they applied the B1G results but not the AAC ones. It's a crapshoot on what they actually care about.

With no math to support this, I've always thought 2.5 would be a good benchmark for my imaginary bracketology metric.