For this year's NCAA Men's College Basketball Tournament I decided to conduct a little experiment aimed at predicting the outcome of the tournament. Over the years I've noticed that upsets happen, and trying to see where these upsets will occur just by looking at the brackets can be extremely difficult. Winning the bracket pool is hard enough when you have every relevant piece of information handy and often memorized and spouted by experts on national tv sports shows. But what information is relevant is the question I was trying to answer this year.
Here's how it started:
The day before the field of 64 was set, and mere hours before play-in games between Cal Poly and Texas Southern, and Iowa versus Tennessee, I started looking at each team's RPI (the number given to each division I school relative to all other schools, involving a formula so complex I don't feel like getting into here.), points scored per game (PPG), points allowed per game (PAG), and ranking in each category among all D1 schools nationwide.
After toying with several formulas to attain a quantifiable number I could assign each team, I came across one which produced an end that I deemed plausible. A red flag should have gone off immediately when I had Kansas at the top of the list. The Jayhawks won the Big 12 going 14-4 in their conference, 25-10 overall. Not a dominant team, but ranked 3rd in the RPI. Following KU were Florida, Arizona, Iowa State, Villanova, Louisville, Wichita State, Creighton, UCLA, Duke, Wisconsin, and Michigan (last year's runner-up). Plausible? Yes. Likely? No. I knew a flaw existed but I couldn't immediately identify it, until the games started to be played.
Solely looking at raw data is never a sound way to predict outcomes of human behavior, much less sporting events and anything as complicated as a 63 game basketball tournament played by amateur athletes from a pool of over 300 schools. Nevertheless I was looking to do so foolishly.
So Wednesday's play-in between Cal Pol and Texas Southern was the first test. Officially not a part of the 64 team circus that was beginning Thursday, but a chance for one of these two to be a part. Here's how they matched up:
RPI PPG Rk PAG Rk
Cal Poly 208 63.2 148 63.4 26
Texas Southern 239 76.2 36 73.7 115
So Cal Poly allowed more points than they scored all year which is reflected in their 13-19 record. Texas Southern was 3rd from the bottom of my quantifiable list. Cal Poly won 81-69. One for one. Cal Poly ranked 15 spots higher than TSU, therefore, I succeeded. For now.
Game 2 that night was the last play-in, this one between Iowa and Tennessee. The winner was to face 6-seed Massachussetts on Friday, and a lot of noise was made about Tennessee during the selection Sunday shows.
I had Iowa ranked 21. Tennessee was below Cal Poly.
RPI PPG Rk PAG Rk
Iowa 56 82.0 28 70.1 81
Tennessee 42 71.3 81 61.1 14
Tennessee routed Iowa 78-65. They say defense wins championships. I wasn't listening.
The next day I entered my picks according to my rankings, had a final four of Kansas, Arizona, Iowa State, and Louisville. Bold picks for sure, but I wanted to put my money on my experiment, just to see how many of these games would be won with it.
Thursday March 20th, Game One: 11-seed Dayton (47) vs. 6 Ohio State (36). From here on out my ranks will be in parentheses unless otherwise denoted. The match-up:
RPI PPG Rk PAG Rk
Dayton 43 73.4 60 67.6 59
Ohio State 24 69.8 95 59.8 11
This one's a no-brainer right? A last-second buzzer beater gave Dayton the win 60-59. But, how can this be? Defense wins, right? OSU looks like Tennessee on paper. My guess says that because the Buckeyes were better defensively than offensively, they under-performed when they had the ball. Iowa was shut down by UT's defense the night before, but Dayton's defense allowed fewer points per game than Iowa's perhaps allowing them a better chance of winning a tightly contested low-scoring struggle. One game, one loss.
Then I got on a mini hot streak. 2 Wisconsin (11) easily beat 15 American (61), 9 Pittsburgh (43) dumped 8 Colorado (53) by 29 points, and 3 Syracuse (30) bested 14 Western Michigan (55), and all of the winners were better at defense than offense during the regular season.
Then, another speed bump.
Everyone knows about the famous 12 vs. 5 matchups where upsets occur with such frequency that teams hate being a 5-seed. The first one this year held true to form, though I missed it.
12 Harvard (42) vs. 5 Cincinnati (35) had upset written all over it. But I was holding my position. I had invented a quantifiable statistical number that can't be beaten! I can be stubborn, sometimes.
RPI PPG Rk PAG Rk
Harvard 46 73.6 58 60.5 12
Cincinnati 21 68.7 104 58.3 6
So how is a team 104th in the nation in scoring supposed to be team who's 12th in defense? They don't. My system and my bracket was in serious trouble at this point. It's flaws have been exposed. Harvard 61, Cincinnati 57.
At the same time, the RPI was being exposed as flawed. Granted my formula incorporates the RPI for each team, it's not given much weight. I think strength of schedule may be too important in the RPI as is average margin of victory. When looking at a team's statistical ranks among other Division 1 schools, both of those measure can go out the window when trying to predict the outcome of a single game.
Here's the rest of Thursday's matches:
RPI PPG Rk PAG Rk
13 Manhattan (38) 60 77.4 29 70.0 80
4 Louisville (6) 19 82.1 7 61.0 13
Final Score: Louisville 71 Manhattan 64
RPI PPG Rk PAG Rk
10 Brigham Young (18) 31 84.2 3 77.0 143
7 Oregon (16) 28 81.8 9 74..0 118
An intriguing matchup that on paper goes either way, Oregon won 87-68.
RPI PPG Rk PAG Rk
16 Albany (65) 185 66.2 126 63.8 28
1 Florida (2) 1 70.7 87 57.9 3
Albany gave them a game, but fell 67-55. 28th ranked defense vs. 87th ranked offense.
RPI PPG Rk PAG Rk
13 Delaware (50) 65 79.5 18 75.1 128
4 Michigan State (20) 18 75.9 39 65.3 41
Sparty 93, Blue Hens 78
RPI PPG Rk PAG Rk
10 St. Joseph's (45) 32 71.3 81 66.9 53
7 Connecticut (31) 22 71.9 75 63.1 24
A lot of pundits like upset in this one. I wasn't buying it. UConn wins 89-81
RPI PPG Rk PAG Rk
15 Wofford (63) 153 67.7 114 62.4 19
2 Michigan (12) 11 74.4 51 65.4 42
Michigan 57, Wofford 40
RPI PPG Rk PAG Rk
12 North Carolina St. (59) 55 70.9 85 69.2 73
5 St. Louis (37) 27 70.0 93 61.2 15
Almost a 12-5 shocker, SLU takes it in OT, 83-80.
RPI PPG Rk PAG Rk
12 North Dakota St. (27) 35 76.4 34 65.4 42
5 Oklahoma (17) 26 82.2 6 75.9 134
I bit on Oklahoma. Had I noticed the glaring 134 above, no way in heck would I have picked them over the number 34 offense in the country. If I knew Wednesday what I was seeing Thursday, things would have been that much more apparent. NDSU 80, OU 75. Also, nationally this was treated as some David and Goliath epic upset the likes of which haven't been seen in a hundred years. I'll give the RPI people credit here. Dakota played at a high level and deserved every bit of that win, it just wasn't that big of an upset. Oklahoma lost 12 games during the regular season. ND State just 6.
RPI PPG Rk PAG Rk
10 Arizona State (39) 44 75.0 46 68.9 70
7 Texas (41) 36 74.1 54 70.0 80
This is the first matchup where the RPI and I disagree over the higher ranked team. Arizona State was better in scoring and defense and lost 87-85. Touche RPI.
RPI PPG Rk PAG Rk
15 Milwaukee (66) 131 70.9 85 70.5 85
2 Villanova (5) 5 78.5 23 66.6 51
Milwaukee was the lowest ranked on my list, succumbing 73-53.
RPI PPG Rk PAG Rk
13 New Mexico St. (40) 72 77.2 30 66.9 53
4 San Diego St. (26) 15 70.6 88 56.6 2
In OT, San Diego ekes out the victory, 73-69.
Not the best of first days, although hindsight and a time machine would have been extremely helpful. My record was 12-4, and often times 75% of correct picks can earn money in a good-sized pool. Besides, I wasn't the only one losing these games. Most of the entries for Warren Buffet's $1 Billion perfect bracked contest were done after Dayton beat Ohio State. That re-assurance didn't help my second-guessing on Friday. I felt like I had rushed into finding the right formula, even though I had spent (or wasted) 2 hours concocting it. But, for what it's worth all I lost were 3 second round losers and 1 Sweet 16 team. 31 of my teams were still alive...until
PPG Rk PAG Rk
79.3 19 66.4 50
78.6 22 67.1 55
I've omitted the names and RPI to focus on the numbers I'm discussing. Friday's first game was between RPI numbers 81 and 9. 81 is a 14 seed, and 9 is a 3 seed, a household basketball name with its entire basketball program on constant display courtesy of the fine folks and ESPN. I think it's easy to see who wins this game, and I wish I would have seen it on Wednesday. The top line is Mercer (33), the bottom Duke (10). Mercer was better both offensively and defensively and we all bought into the familiar Duke name without hesitating. Unfortunately for me, Duke is in my Elite 8, so now I'm down to 14 of my Sweet 16 and 7 of 8. My margin for error is narrowing quickly.
RPI PPG Rk PAG Rk
11 Nebraska (58) 48 67.0 118 64.9 37
6 Baylor (29) 30 75.2 44 68.4 66
As you can see, I rated Nebraska one of the worst teams in the tournament. In fact the program has never won a tournament game. And they didn't start this year. Baylor 74-60.
RPI PPG Rk PAG Rk
10 Stanford (46) 41 73.5 59 68.6 68
7 New Mexico (14) 12 74.6 49 66.4 50
True upset.. Stanford 58 New Mexico 53. Like I said, numbers can't predict everything. Both teams were better offensively than defensively, yet was extremely low-scoring. Never would have picked Stanford.
RPI PPG Rk PAG Rk
16 Weber State (60) 149 73.2 62 66.4 50
1 Arizona (3) 2 73.1 63 58.1 4
Zona 68, Weber 59
The other 1 seeds, Wichita State and Virginia cruised to victories also.
RPI PPG Rk PAG Rk
11 Tennessee (52) 42 71.3 81 61.1 14
6 Massachussetts (25) 23 76.1 87 70.9 88
I went with UMass. And was not surprised when the better defensive team won. I was shocked by the score, 86-67.
RPI PPG Rk PAG Rk
14 Louisiana-Lafayette (44) 90 81.4 10 75.1 128
3 Creighton (8) 10 79.5 18 67.4 58
It's weird these days seeing Creighton as a 3-seed, but another game pitting 2 good offenses against each other, defense prevailed, 76-66.
RPI PPG Rk PAG Rk
15 Eastern Kentucky (48) 97 79.0 21 70.6 86
2 Kansas (1) 3 79.6 17 70.5 85
On paper, Kansas is marginally better than EKU. With my whole system hinging on Kansas and my questioning everything about the system, this matchup makes me nervous on paper. It looks like Mercer and Duke, except with worse teams. Kansas, and my entry, survived, 80-69. Now I'm seeing each win for my brackets as small victories for my system. I'm seeing things more clearly. The picture is getting less convoluted by guys in tv studios telling me who should be winning these games, but I also realize that Kansas won't win this tournament the way things are shaking out.
RPI PPG Rk PAG Rk
9 Oklahoma State (22) 45 80.3 13 69.1 72
8 Gonzaga (15) 20 76.9 31 65.0 38
Coin flip 8-9? Hardly. Better defense prevailed over better offense, 85-77.
RPI PPG Rk PAG Rk
9 George Washington (34) 29 73.4 60 66.9 53
8 Memphis (28) 37 77.6 27 70.4 84
Although I won it, GW looks like Dayton on paper and Memphis looks like Baylor. Could have gone either way.
RPI PPG Rk PAG Rk
11 Providence (51) 40 73.2 62 69.5 76
6 North Carolina (24) 25 76.3 35 69.1 72
Again take out the names and RPI and it's a close one, although UNC has edge on offense. They win 79-77.
RPI PPG Rk PAG Rk
12 Stephen F Austin (32) 52 76.5 33 62.6 21
5 Virginia Commonweath (13) 13 75.0 46 64.9 37
VCU's RPI helped their ranking, and lost 77-75. Another testament to PPG and PAG rankings.
RPI PPG Rk PAG Rk
9 Kansas State (54) 51 69.7 96 65.7 45
8 Kentucky (19) 17 75.9 39 66.6 51
2 equal defenses, 1 superior offense. UK wins 56-49.
RPI PPG Rk PAG Rk
14 Coastal Carolina (62) 196 72.7 67 68.4 66
3 Iowa State (4) 7 82.9 5 73.9 117
After I crunched the numbers, Iowa State is in the top 4. Fortunately for me their offense dominated with a 93-75 win.
RPI PPG Rk PAG Rk
13 Tulsa (56) 73 73.1 63 67.4 58
4 UCLA (9) 14 81.8 9 70.8 88
Again, better defense can't control better offense. UCLA 76-59.
At the end of day 2, I had 8 losses but, thankfully, minimal damage. Only those losses by Oklahoma and Duke have hurt me so far. With day 3 and the second round looming. My system looked vulnerable as the successes of the first round winners were easily explained. I wish I had done this differently, although I could already tell I was onto something.
But, would these same predictive indicators hold things true for the outcomes of Saturday's games?
Early evidence said "yes," demonstrated by 4 blowouts to start the day.
RPI PPG Rk PAG Rk
5 St. Louis (37) 27 70.0 93 61.2 15
4 Louisville (6) 19 82.1 7 61.0 13
UL 66, SLU 51
RPI PPG Rk PAG Rk
9 Pittsburgh (43) 39 72.5 69 62.4 1
1 Florida (2) 1 70.7 87 57.9 3
Gators win low-scoring affair 61-45.
RPI PPG Rk PAG Rk
7 Texas (41) 36 74.1 54 70.0 80
2 Michigan (12) 11 74.4 51 65.4 42
Michigan's defensive superiority gives them 79-65 win.
RPI PPG Rk PAG Rk
12 North Dakota St. (27) 35 76.4 34 65.4 42
4 San Diego St. (26) 15 70.6 88 56.6 2
Interesting how these 2 are ranked by my system. North Dakota was stymied 63-44. San Diego State's line is nearly identical to Florida's, but because of their RPI, ranked much lower by me.
So after going 4 for 4, something was bound to happen. I'll take 75% correct any day during these tournaments, so I was instinctively nervous. Perfect days just don't happen, especially in the first and second rounds.
RPI PPG Rk PAG Rk
11 Dayton (47) 43 73.4 60 67.6 59
3 Syracuse (30) 16 68.2 109 59.5 9
The good defenses have been shutting down mediocre offenses so far, and Dayton again faced a top-tier defense with a mediocre offense extremely similar to Ohio State, so naturally they knocked off Syracure 55-53. Another Sweet 16 team done.
RPI PPG Rk PAG Rk
7 Oregon (16) 28 81.8 9 74..0 118
2 Wisconsin (11) 6 73.5 59 64.6 34
Wisconsin kept the Ducks in check 85-77.
RPI PPG Rk PAG Rk
12 Harvard (42) 46 73.6 58 60.5 12
4 Michigan State (20) 18 75.9 39 65.3 41
Thankfully my formula worked for this one with Michigan State winning 80-73.
RPI PPG Rk PAG Rk
7 Connecticut (31) 22 71.9 75 63.1 24
2 Villanova (5) 5 78.5 23 73.9 117
I shouldn't have to say who won this one. Another Sweet 16 out as Villanova falls 77-65. The way things are going in this tournament this is not an upset. First day of Round 2 and I'm 6-2. 7 of my 8 are still in. 12 of my 16 <gulp>. Sunday has to be perfect, but with my remaining teams I'm having serious doubt.
The first thing I read about the upcoming games on Sunday was that Iowa State had lost one of its key players to injury. With my having Iowa State ranked number 4, and East Region champions, I'm skeptical about their ability to beat North Carolina while sad for the player and his teammates.
RPI PPG Rk PAG Rk
3 Iowa State (4) 7 82.8 5 73.9 117
6 North Carolina (24) 25 76.3 35 69.1 72
On paper, the Tar Heels should have picked apart Iowa State, especially without State's injured forward. The Cyclones pulled together and won 85-83. I guess since neither team was exceptional at defense all year, the better offensive team won and I considered myself pretty lucky with that pick. My confidence in Iowa State's ability to make it to the Final Four has now diminished as they will face Connecticut and their 24th ranked defense next round.
RPI PPG Rk PAG Rk
10 Stanford (46) 41 73.5 59 68.6 68
2 Kansas (1) 3 79.6 17 70.5 85
Done-zo. Stanford defeats Kansas and reinforces the fact that my formula is fatally flawed. Just 2 hours into the 4th day of the tournament, and my brackets are officially busted. Stanford won 60-57. There is always next year, and the games do go on.
RPI PPG Rk PAG Rk
8 Kentucky (19) 17 75.9 39 66.6 51
1 Wichita State (7) 4 75.5 41 59.6 10
I'll call this one Anomaly Game #2. Evenly matched offenses, one superior defense. This is why they don't play games on paper. Would have missed this one every time.
RPI PPG Rk PAG Rk
14 Mercer (33) 81 79.3 19 66.4 50
11 Tennessee (52) 42 71.3 81 61.1 14
Since I didn't pick either of these teams to win their first game, I was really curious to see who I should have picked, and the easy answer is Tennessee by virtue of defense. And they did 83-63.
RPI PPG Rk PAG Rk
12 Stephen F Austin (32) 52 76.5 33 62.6 21
4 UCLA (14) 14 81.8 9 70.8 88
The way this tournament is going, you would think SFA would have won, but RPI and I won again, as did UCLA 77-60.
RPI PPG Rk PAG Rk
6 Baylor (29) 30 75.2 44 68.4 66
3 Creighton (8) 10 79.5 18 67.4 58
My ranking says Creighton all the way. Apparently Baylor had other ideas. They destroyed Creighton by a score of 85-55. Anomaly game #3. Over 4 days not bad, but I just lost Elite 8 team number 3. And only 10 of my Sweet 16 now remained. Luckily the remaining 1-seeds, Virginia and Arizona won their games. Still, my brackets are in shambles, and I have until Thursday to see what happens next to the remainder of my picks. 9 of 16 are still in, 5 of 8, 3 of Final Four, but National Champ is out. No way in heck I can win. What a disastrous day. At least Iowa State won.
The round of 16 began innocently enough with this tournament's darlings, Dayton taking on Stanford. On paper Wednesday night one of the unlikeliest of 3 round matchups.
RPI PPG Rk PAG Rk
11 Dayton (47) 43 73.4 60 67.6 59
10 Stanford (46) 41 73.5 59 68.6 68
Evenly matched, Dayton advances to the Elite 8 with an 82-72 win. After winning the very first game of the tournament, Dayton has become the best story.
RPI PPG Rk PAG Rk
6 Baylor (29) 30 75.2 44 68.4 66
2 Wisconsin (11) 6 73.5 59 64.6 34
Once again my higher-ranked team won, handily 69-52. Not to mention the better defense beat the better offense.
RPI PPG Rk PAG Rk
4 UCLA (14) 14 81.8 9 70.8 88
1 Florida (2) 1 70.7 87 57.9 3
Formula success again! Gators win 79-68.
RPI PPG Rk PAG Rk
4 San Diego St. (26) 15 70.6 88 56.6 2
1 Arizona (3) 2 73.1 63 58.1 4
Arizona is higher-ranked, and had better offense although teams were even on defense. Wildcats win it 70-64.
Once again I was thankful for my wins,and hoped the success would continue on Friday.
RPI PPG Rk PAG Rk
11 Tennessee (52) 42 71.3 81 61.1 14
2 Michigan (12) 11 74.4 51 65.4 42
My rankings prevail although the better defense didn't slow the better offense. Michigan 73-71.
RPI PPG Rk PAG Rk
4 Michigan State (20) 18 75.9 39 65.3 41
1 Virginia (23) 8 65.9 128 55.3 1
My rankings prevail, although better defense didn't slow the better offense. Maybe because Virginia was so poor at scoring, and Michigan State was better than average defensively, that Michigan State won, 61-59.
RPI PPG Rk PAG Rk
7 Connecticut (31) 22 71.9 75 63.1 24
3 Iowa State (4) 7 82.8 5 73.9 117
Despite my rankings, no surprise Iowa State lost here 81-76. Better defense by UCONN and the lack of player depth hurt Iowa State, and my Final Four, which is down to 2.
RPI PPG Rk PAG Rk
8 Kentucky (19) 17 75.9 39 66.6 51
4 Louisville (6) 19 82.1 7 61.0 13
RPI got it right. I didn't. 3 of my Elite 8 are still in, 1 Final Four team. Maybe the RPI people and the tournament selection committee should start hanging out together. Some of the seeding didn't seem to make much sense. How can a number 17 in the RPI, a runner-up to Florida in its conference "earn" an 8-seed?
Saturday, March 29 and Sunday March 30--Regional Finals
RPI PPG Rk PAG Rk
11 Dayton (47) 43 73.4 60 67.6 59
1 Florida (2) 1 70.7 87 57.9 3
Florida cruises to a 10-point win, and now looks tough to beat, winning the South Region.
RPI PPG Rk PAG Rk
2 Wisconsin (11) 6 73.5 59 64.6 34
1 Arizona (3) 2 73.1 63 58.1 4
This one looks easy to pick, right given the trends I've seen in this tournament? Arizona, ranked higher, better defensively, teams equal in offensive production. Wisconsin wins in OT 63-64, and the West Region;and out goes the last of my final four teams.
RPI PPG Rk PAG Rk
7 Connecticut (31) 22 71.9 75 63.1 24
4 Michigan State (20) 18 75.9 39 65.3 41
During the selection Sunday show on ESPN I remember hearing Dick Vitale predict Michigan State and Louisville would meet in the Championship game. Who was I to judge? Especially now. I didn't have either team reaching this point and it's obvious who would win on paper. After all, the better defenses were consistently beating better offenses, which is exactly what happened here as the Huskies advanced with a 60-54 win. Oddly enough, the teams winning these games scored about 10 fewer points than their season average.
RPI PPG Rk PAG Rk
8 Kentucky (19) 17 75.9 39 66.6 51
2 Michigan (12) 11 74.4 51 65.4 42
These 2 appeared evenly matched but Michigan had a marginal edge defensively, but failed to beat Kentucky as the Wildcats prevailed 75-72 thanks to a 3-pointer with 2.7 seconds left to play. Kentucky's run continues and their seeding comes into question again after knocking out 3 of last year's final four reach this year's final four.
Saturday April 5--National Semi-finals
RPI PPG Rk PAG Rk
7 Connecticut (31) 22 71.9 75 63.1 24
1 Florida (2) 1 70.7 87 57.9 3
Here we go. Clearly Florida is the better team. No way UCONN can beat them and advance to the Final right?
Wrong again. This was getting old about a week ago, but this one seals it. Connecticut won 63-53.
RPI PPG Rk PAG Rk
8 Kentucky (19) 17 75.9 39 66.6 51
2 Wisconsin (11) 6 73.5 59 64.6 34
This one's a no-brainer, too. Better defense should beat the better offense. Nope. Not tonight. Again on a 3-point shot in the waning seconds, Kentucky goes on to face Connecticut for the National Championship.
Monday April 7--National Championship
RPI PPG Rk PAG Rk
8 Kentucky (19) 17 75.9 39 66.6 51
7 Connecticut (31) 22 71.9 75 63.1 24
The way this tournament has gone, UConn would appear to have the edge, but Kentucky has been able to rise above their predictive indicators and surprise a lot of people during their run to this game. Unfortunately for them, things held true to form as Connecticut wins the game and the National Championship. Warren Buffet's billion dollars were safe over 2 weeks before this game and it should come as a surprise to no one.
If it were as simple as plugging a bunch of numbers and accurately predicting the winners of all 63 games, I'd sell the formula for $100 to each person that wanted it. The money involved in the tournament bracket pools makes it that some people are willing to do just about anything to gain an edge. Also, with all the intangibles at play in these games, numbers will never be able to project the outcome of the tournament. Numbers can help us tell some things about the teams and decide on which ones we should pick, but that's all.
Thanks for reading.
Kurt
