Archive for March 2016

A New Metric, and 2016 NCAA Tournament Picks and Analysis

March 16, 2016

A New Metric: I added each team’s Sagarin Schedule Strength rating to its Scoring Margin to create a Power Ranking. In terms of scale, the two fit together quite well, such that the absolute value of each different statistic between teams 1 and 100 is roughly the same. The idea is that over the large sample size of an entire season, a team’s Scoring Margin should be completely indicative of its supremacy once adjusted for its Schedule Strength. Here are the results after doing so, with each team’s seed listed:

2 Michigan State 94.98

1 Kansas 94.66

3 West Virginia 93.51

1 North Carolina 93.07

1 Virginia 92.48

2 Villanova 92.47

5 Purdue 91.77

2 Oklahoma 91.67

5 Indiana 90.22

4 Kentucky 90.14

 

4 Duke 89.89

6 Arizona 89.84

3 Miami 89.46

3 Texas A&M 89.24

2 Xavier 88.93

7 Iowa 88.89

1 Oregon 88.70

4 Iowa State 88.59

5 Maryland 88.01

5 Baylor 87.97

 

3 Utah 87.78

11 Witchita State 87.59

9 Butler 87.21

11 Gonzaga 87.12

9 UConn 87.09

9 Cincinnati 86.48

4 California 86.85

10 Pittsburgh 86.01

11 Michigan 85.86

6 Texas 85.73

6 Seton Hall 85.33

6 Notre Dame 84.79

 

Best of the other lower seeds:

 

12 Yale 81.39

13 Hawaii 80.22

14 Stephen F Austin 81.50

 

Some strategic observations based on this:

 

  • If these ratings are to be believed, then the most likely Final 4 is Michigan State, West Virginia, Kansas and Oklahoma. However, West Virginia is eliminated as true contender, albeit just barely, based on their Offensive Efficiency number being <115.0. Therefore, we will be selecting the next highest ranked qualifier in that region, North Carolina, while selecting West Virginia to advance to the Elite 8.
  • The lowest ranked #1 seed, by every metric and by a mile, is Oregon. The irony here is that the weakest #1 seed was actually rewarded with one of the weakest regions in the history of the tournament. Even so, we will look to beat the Ducks early with Duke, who although lacking in depth, have the superior power ranking and match up well in terms of efficiency, as both teams have high-powered offenses but relatively poor defenses. Neither team rebounds well either. Duke is one of three teams in the tournament that have two players projected as NBA First Round picks (California, Kentucky), and sometimes that means a lot in March. Additionally, Oregon stands the best chance to lose in the second round of all the #1 seeds per KenPom (20%), slightly decreasing the risk of this pick.
  • The lowest ranked #2 seed is Xavier, and both KenPom and FiveThirtyEight give them by far the lowest chance to make the Elite 8 of all the #2 seeds (below 30%), so we will play the numbers and keep them out of it, with West Virginia beating them in the Sweet 16. Our season-long opinion has been that the Big East is overrated.
  • Villanova is actually the second highest ranked #2 seed, but this is a bit misleading, as they have to potentially face two of four highly underseeded teams analytically speaking (Arizona, Miami, Witchita State, Iowa) in order to advance. Iowa, for starters, is by far the strongest #7 seed, which immediately decreases Villanova’s chance of advancing relative to its #2 seed peers. In addition, their Three Point Rate of 43.8% combined with a Three Point Shooting % of just 34.4% is a major red flag and a historical eliminator for a Final Four contender. Keeping with the bearish opinion on the Big East, we will try to stop them well short of that.
  • Value rests in locating the correct double digit seeds to advance to the Sweet 16, because let’s face it, some always do. Witchita State and Gonzaga appear fairly imposing in this department based on this metric. Gonzaga, especially, won’t have to contend with any far superior teams to get there. In fact, by drawing the lowest ranked #3 seed in the second round (Utah), this is a fairly riskless choice, as not too many will be taking the Utes past a dazzling Michigan State squad there. Witchita will have a tougher road, having to play the toughest 6 seed in the first round (Arizona), but there is still value here. The Shockers boast the #1 Defensive Efficiency in the nation, and are especially dominant in 2 Point FG% defense, which is where Villanova and Miami excel (#6 and #25 respectively in 2 Point FG Shooting %). Their achilles’ heel has been their offense, which ranks just #81 in terms of efficiency, but Arizona and Miami don’t play dominant defense (#40 and #44 respectively), and the Shockers help their chances in this department by taking extremely good care of the ball and making the most of their possessions (#8 in Turnovers Per Possession). Villanova could be a problem if Witchita State puts them on the line too frequently. The Wildcats rank #3 in the nation in Free Throw Shooting %, while the Shockers rank #327 in Opponent FTA/FGA. This is a result of playing aggressive defense inside the arc, however. Additionally, Witchita State is very strong on the glass, ranking #2 in Defensive Rebounding %. This could neutralize the opposition on the offensive boards and cut down on second chance points, as none of these opponents are incredibly strong on the offensive glass, and could be especially troublesome for Villanova, who ranks just #157 in Offensive Rebounding %.
  • Yale appears the most likely #12 seed to upset a #5, adding value to our Duke pick. Baylor’s main strength lies in its ability to crash the boards offensively (#3 in Offensive Rebound %). Yale can completely neutralize this by its strong rebounding on that end (#6 in Defensive Rebound %). Baylor also does not rebound defensively nearly as well as Yale does offensively (#107 vs. #5). Of course, a Baylor win conversely poses an extreme rebounding matchup problem for Duke in the next round (#3 Offensive Rebound % vs. #330 Defensive Rebound %). Still, regardless of that one rather glaring matchup dilemma, both KenPom and FiveThirtyEight agree with our metric and favor Duke slightly over Baylor, and FiveThirtyEight even predicts Duke as the most likely #4 seed to advance to the Elite 8, even above Kentucky.
  • Hawaii appears the most likely #13 seed to upset a #4, so we will select Maryland over California in the second round. The Terps’ higher ranking by our metric adds support to this conclusion. Both of these teams have the talent, on paper, to give Kansas a very tough game, but we will stop short of sending the Jayhawks home quite that soon.
  • Stephen F Austin appears the most likely #14 seed to upset a #3, and actually has a higher rating than any #12 or #13 seed, which makes things a bit tricky considering that West Virginia appears to be the strongest #3 seed. (Here is another example of really bad use of analytics by the committee). This is an area of the bracket that could potentially devolve into complete chaos. For starters, this is a crazy first round game between two underseeded teams who rank 1st and 2nd in Turnovers Per Possession. They are essentially the same team, and the scary thing is that West Virginia actually is a lot more prone to turning it over on offense than SFA is (#285 and #182 respectively, so this could be a complete trainwreck of a basketball game). Still, West Virginia ranks #1 in Offensive Rebounding % and played a far tougher schedule, while SFA doesn’t crash boards well (#157 Defensive Rebounding %). This Offensive Rebounding dominance could pose problems for Notre Dame in the second round as well (#247 Defensive Rebounding %), even if the Irish aren’t as susceptible to the West Virginia press (#10 Turnover Rate). Michigan rebounds much better (#47) and also takes care of the ball (#6 Turnover Rate) and would probably be a tougher matchup for the Mountaineers, but we view Notre Dame-Michigan as a virtual coin flip game and therefore a risky pick to advance either of those teams beyond the second round.
  • On Indiana-Kentucky: This is admittedly an incredibly regrettable and unfair draw for the Hoosiers. However, this is not the Kentucky team of last year or of 2012 that beat Indiana en route to a National Championship. There are matchup advantages that Indiana can exploit here, and the metrics above prove that Indiana played about equal to Kentucky this season. One that sticks out, and is always a wildcard in March, is Three Point Shooting %. Indiana is one of only five teams in the tournament that shoots over 40% from beyond the arc, ranking #6 in the nation. But Kentucky ranks just #86 in Three Point Shooting % Defense. In combination with that, this Kentucky team is not strong at all on the boards, ranking just #260 in Defensive Rebounding %. Indiana, conversely, has been highly successful creating second chance opportunities and taking advantage of them thanks to their #12 Offensive Rebounding %. The idea here is that Indiana may be able to either light up Kentucky from three point land, or rebound a lot of their misses if they do not. These are really two very similar teams, with roughly equally great offenses and equally mediocre defenses, depending on your preferred metric. Meanwhile, Purdue ranks the most likely #5 seed to make a Final Four run, and will be a matchup problem for anyone. They simply tower over a very unbalanced Iowa State team from an analytics standpoint. We will be playing both #5 seeds over the #4s on this side of the bracket.
  • North Carolina doesn’t shoot very well. While our elimination model still gives them the best shot to come out of the difficult East region, this statistic bears mention: Over the last 18 years, which is as far back as we could find data, not a single team has made the Final 4 with a Three Point Shooting % as low as the Tarheels (31.4%). In fact, this historically poor shooting is more than two standard deviations from the mean over that time span (36.4%, 2.4%). The lowest Three Point Shooting % for a Final 4 team over that time span was 31.8% (Louisville in 2012). The highest, somewhat ironically, was the 2005 Tarheels (40.3%). Having to overcome this trend will lead us to send North Carolina home after they play the Midwest winner in the Final Four and avoid them as a National Champion pick. Not surprisingly, the average Three Point Shooting % statistics for Champions are even more daunting: 37.6%, with 2013 Louisville the lowest at 32.9%. Additionally, 14 of those 18 National Champions have had a Three Point Shooting % ABOVE the 36.4% mean for Final 4 teams.

THE DIABOLICAL, UNAVOIDABLE REMATCH BETWEEN MICHIGAN STATE AND VIRGINIA:

It’s easy to forget that if Justin Anderson hadn’t opted to go pro after last season- a decision that looks just as perplexing now as it did then- Virginia would have been the runaway choice for the preseason champion pick. However, an interesting development this season has been the improvement of the Virginia offense after Anderson’s departure. (They now rank #9 nationally in offensive efficiency after finishing last season at #24). This senior-laden squad is highly efficient on both sides of the ball, and is one of only two teams in the nation to rank in the top 10 in both offense and defense, the other being Kansas, per Ken Pomeroy. For what it’s worth, over the last 14 seasons since he has been keeping this data, 50% of National Champions own that profile. Malcolm Brogdon has become a spectacular playmaker on both sides of the ball, winning both the ACC Offensive and Defensive Player of the Year awards, and should finish third in the Wooden Award voting behind Buddy Hield and Denzel Valentine. Brogdon doesn’t rely completely on the three point shot the way that Anderson did, and his versatility on the offensive end is less susceptible to cold streaks, thereby increasing this team’s efficiency.

Did you know that senior point guard London Perrantes shoots 48.0% from three-point range, the second highest among Power Conference players with at least 100 attempts? Seniors Anthony Gill and Mike Tobey anchor a formidable, experienced front court. This isn’t a flashy squad or one that likes to score in transition. The Cavaliers flourish almost entirely in a half court offense and operate strictly with off-the-ball screens. As a result, they are a patient bunch that wear down and frustrate defenses until they find the right shot. Out of 351 Division 1 teams, they rank 351st in pace of play. That statistic, combined with their highly efficient offense, means that opponents that fall behind early, which is usually the case, can feel like they are climbing up Mount Everest as they try to come back. This is an extremely well-coached team that will simply lull you to sleep, wear you down and suffocate you like an anaconda. They also refuse to make things easy for opponents, as their turnovers per possession numbers are among the top 95% in the land. I picked Virginia all the way to the final game last season, but they fell to Michigan State for the second straight year. This team is better than that one was.

Unfortunately for Virginia, the Spartans inexplicably fell into their bracket yet again this year. That’s the only team that I think can beat them. I like the winner of the Midwest region to win the tournament, but as much as I like Virginia, I’m picking against them this year for two reasons. For one, for all the controversy, the committee really synthetically awarded Michigan State the #1 seed here, since the Regional Final will be played in Chicago. Also, Virginia’s matchup against Purdue figures to be a lot tougher than whoever Michigan State meets in the Sweet 16 (we are going with Gonzaga, remember), and for that reason, from a probability standpoint, Michigan State seems the better play. The metric above adds yet more evidence to the idea that the Spartans are simply the best team in the country.

PICKS:

South: #1 Kansas over #11 Witchita State

West: #2 Oklahoma over #4 Duke

East: #1 North Carolina over #3 West Virginia

Midwest: #2 Michigan State over #1 Virginia

Championship: #2 Michigan State over #1 Kansas

Final Four Elimination Criteria

March 14, 2016

 

When breaking down your bracket, I always find it helpful to work backwards rather than forwards. That is, by eliminating teams that you believe cannot make a deep tournament run, it becomes easier to narrow down your Final Four contenders first and fill out your bracket accordingly. This gives you the ability to weigh risk and return. What teams can you afford to take shot for an upset against, and which ones should you go ahead and pencil in to at least the Elite Eight? I’ve written at length about my Final Four elimination criteria in the past, but I wanted to create a more concise summary.

It’s always important to remember of course, that this is March, and anything can happen. These elimination criteria attempt to maximize the probability that a given team WILL NOT make a run to the Final Four. That is different than saying that such a run is impossible, as there are always exceptions to the rule. Last year, for example, Duke’s defense was ranked just outside the Top 50 in Defensive Efficiency, and their offense, being just outside the Top 2, was not theoretically good enough to compensate for it. (It was 3rd, so it was right on the cutoff line). Still, who could have predicted or expected that Duke would be able to increase their Defensive Efficiency ranking all the way up to 12th over a six game Championship run? This was a function of incredible coaching adjustments and the understanding of where the team’s problem was, and that Duke team had the personnel to make those adjustments and corrections work perfectly. Such an improvement is the exception, and not the rule.

Here are the three critieria that we will use to eliminate contenders from our Final Four selections:

TOO MUCH DEPENDENCE ON THE THREE= DEATH IN MARCH

Since 1998, only three teams with a Three Point Rate (3 Pt FGA/ Total FGA) above 40% have made the Final Four. That is only 4% of all Final Four teams over that timespan, which is beyond a two standard deviation event, and all three of them ranked in the top 10% in the nation in Three Point Field Goal Percentage, adding credence to their outlier status.

The exception to this rule, because its sample size is so rare, involves teams that do show a Three Point Rate above 40%, but that actually show a Three Point Field Goal Percentage higher than that. It is always hard for a team that is highly dependent on perimeter shooting to string together consecutive wins, but a team that is highly skilled in this department is correct to allocate their shot selection in this manner, and should not be eliminated from consideration for this reason.

So, eliminate all teams from Final Four contention with a Three Point Rate above 40%, unless their Three Point Field Goal Percentage is higher than their Three Point Rate.

OFFENSE WINS CHAMPIONSHIPS

 Since 2008, 95% of Final Four teams have had either an Offensive Efficiency rating greater than 115.0 OR a Defensive Efficiency rating below 90.0, per Ken Pomeroy. The former is much more frequent than the latter, demonstrating that a high-fueled offense with limited defensive capabilities is better suited for a deep tournament run than a top notch defense that lacks offensive punch (see Duke winning it all last year).

So, eliminate all teams from Final Four contention that show an Offensive Efficiency Rating below 115.0 AND a Defensive Efficiency rating above 90.0.

 BALANCE IS EVERYTHING

 The most significant thing that we find when researching Ken Pomeroy’s data, which is available back to 2002, is that it is incredibly rare for a team to advance to the Final Four if it is ranked outside of the Top 25 on both sides of the ball. This has happened only twice over that timespan, which means that over 96% of Final Four teams are ranked in the Top 25 on at least one side of the ball:

2010 Michigan State: #36 AdjOff, #27 AdjDef

2011 Butler: #48 AdjOff, #44 AdjDef

And, even if they are Top 25 on one side of the ball, if they happen to fall outside of the Top 50 on the other side of the ball, over 96% of Final Four teams are ranked in the top TWO on the opposite side of the ball in order to compensate for such weakness. An exception to this has also only happened twice, and these remain the two most shocking Final Four teams in history:

2006 George Mason: #58 AdjOff, #13, AdjDef

2011 VCU: #25 AdjOff, #84 AdjDef

You read that correctly. For the opposite to occur, in both instances, is beyond a two standard deviation event.

There are ways to prioritize and create a hierarchy of teams that show a specific profile as it relates to this criteria among those that are not eliminated. For example, 86% of Final Four teams rank in the Top 50 on both sides of the ball, 80% of Final Four Teams rank in the Top 10 on one side or the other, and 59% of Final Four Teams rank in the Top 25 in both. But the overriding moral of the story here is that it is highly preferable to be among the nation’s best in both offensive and defensive efficiency.

So, eliminate all teams that are ranked outside the Top 25 on both sides of the ball. Also, eliminate all teams that are ranked outside the top 50 on either side of the ball UNLESS that team ranks in the Top 2 nationally on the stronger side of the ball.

CONCLUSION

Amazingly, after applying these three simple criteria to the entire field, we are left with only 12 teams in the entire country that statistically speaking, stand a chance to make the Final Four, and roughly in this order. I will save you the work and just tell you:

Kansas

Virginia

Michigan State

North Carolina

Oklahoma

Kentucky

Oregon

Purdue

Witchita State

Miami

Xavier

Arizona

 

Happy Picking.

FINAL BRACKETOLOGY

March 13, 2016

 

1KANSAS Virginia MICHIGAN STATE Oklahoma

2Villanova West Virginia NORTH CAROLINA OREGON

3Utah Miami Seton Hall Xavier

4Texas A&M Purdue Texas Indiana

5Iowa Iowa State Duke KENTUCKY

6Maryland California Notre Dame Providence

7UCONN Cincinnati Baylor Arizona

8Dayton Wisconsin Temple Michigan

9ARKANSAS LR St. Mary’s Colorado Butler

10GONZAGA Oregon St Pittsburgh Texas Tech

11South Carolina ST JOSEPH’S Tulsa/ Witchita St USC/ Syracuse

12HAWAII CHATTANOOGA STEPHEN F AUSTIN YALE

13FRESNO ST NORTHERN IOWA AKRON STONY BROOK

14MIDDLE TENNESSEE SOUTH DAKOTA ST BUFFALO IONA

15CSU BAKERSFIELD WEBER STATE UNC ASHEVILLE GREEN BAY

16FARLEIGH D/ SOUTHERN TEXAS SOUTHERN DUNK CITY HOLY CROSS/ AUSTIN PEAY

Thoughts on 3 Point Rate and A Possible Exception To The 40% Eliminator Rule

March 2, 2016

3 Point Rate is defined as: (3 Point FG attempts/ Total FG attempts). Since 1998, only three teams have made the Final 4 with a 3 Point Rate above 40%. Here are their stats, with 3 Point FG % first, followed by 3 Point Rate, and national ranking in 3 Point FG%:

2011 VCU: 37.0/41.2 (55th)

2005 Louisville: 40.0/ 41.3 (10th)

2001 Duke:  38.5/41.8 (29th)

Since just 3 of the last 72 Final Four teams (4.1%) have shown a Three Point Rate above 40%, this trend is significant enough to eliminate from Final Four consideration any team that shows this very infrequently successful profile. This makes sense when you think about it. A team that relies so heavily upon making more difficult shots in order to win is going to be less likely to be able to repeat making these shots over consecutive games in a one-and-done sort of scenario, and the historical stats support this rather obvious trend.

Not surprisingly, all three of these teams shot fairly well from behind the arc in terms of national ranking, which allowed for them to overcome this extreme historical red flag in terms of inefficient shot allocation and advance to the Final Four in spite of it. This got me to thinking though: Have there ever been teams that showed a 3 Point Rate above 40%, but actually shot a higher 3 Point FG% than that?

Since 1998, there have only been four such teams in the tournament:

2014 Michigan: 40.2/ 40.0 (#2 seed, lost in Elite 8)

2008 American: 40.8/ 40.2 (#15 seed, lost 1st Rd)

2002 Utah: 40.5/ 40.3 (#12 seed, lost in 1st Rd)

2000: Creighton: 41.6/40.4 (#10 seed, lost in 1st Rd)

Also, 2010 Cornell shot 43.3% from distance with a 39.8% Three Point Rate, dangerously close to the threshold but just below it. That team, as a #12 seed, made a Cinderella run to the Sweet 16.

We can draw two main conclusions from this data.

– Since 1998, 1182 teams have earned a trip to the NCAA Tournament, and a mere 4 of them fit this profile described as: Three Point FG % > Three Point Rate when Three Point Rate > 40%. That’s 0.034%. You don’t have to be a mathematician to ascertain that such teams are extremely rare.

– Even with such a small sample size, it is interesting to observe that none of the four teams that did fit this profile technically underperformed their seed. Three lower-seeded teams lost in the first round as they should have, while second seeded Michigan advanced to the Elite 8 as they should have.  The fact that the sharp-shooting Cornell team made their underdog Sweet 16 run with a team heavily dependent upon the three point shot is worth remembering.

Amazingly, this year’s tournament will see two of these very rare profiles in the same year, which has never happened before:

Indiana 42.0/ 41.1 (5th)

Oklahoma 42.8/ 40.5 (2nd)

While historical red flags do admittedly merit tossing the other highly regarded team in this year’s field with a 3 Point Rate above 40%- that’s Villanova (33.5/ 44.5) – from Final Four consideration, one might tread more carefully with Indiana and Oklahoma, as the obscurity of their success behind the arc in combination with their affinity for shooting behind it is a massive wildcard and not something that is commonly observed among tournament teams. Additionally, the spread between their respective 3 Point FG% and 3 Point Rate (0.9% for Indiana and a whopping 2.3% for Oklahoma) is larger than the average of even those four above (0.55%). These are two very rare teams indeed.

In layman’s terms, we could make the fairly obvious statement that a team that lives and dies by the three is a lot more likely to make a deep tournament run when they are the very best in the nation at shooting the three. It is interesting indeed that over the years, teams that shoot the largest percentage of their shots from distance are not usually among the best at shooting from distance, and the teams best at shooting from distance are not usually among the teams that allocate a higher portion of their shots from distance when compared with other teams.

In fact, digging deeper, I was only able to find six more teams over those last 18 seasons in all of Division 1 that even showed both a Three Point Pct and a Three Point Rate both above 40%, even with Three Point Rate > Three Point Pct:

2006 Notre Dame: 40.3/ 41.6 (no tourney)

2005 Louisville: 40.0/ 41.3 (#4 seed, lost in Final 4)

2005 Samford: 42.1/ 50.1 (no tourney)

2004 St. Joseph’s 40.4/ 42.5 (#1 seed, lost in Elite 8)

2003 Penn 41.1/ 42.6 (#11 seed, lost in 1st Round)

1998 New Mexico 40.5/ 41.5 (#4 seed, lost in 2nd Round)

CONCLUSION: I’m not sure I can explain why, generally speaking, that the teams best at shooting the three don’t take the largest percentage of their shots from three, or why the teams that take the largest percentage of their shots from three are usually not very good at hitting them. It seems very inefficient from a statistical standpoint, and that falls on coaching and a lack of analytical focus to some extent. But, given the rarity of teams such as Oklahoma and Indiana that actually do utilize their shooting in a manner that makes sense, I would be very hesitant to toss them out of consideration simply due to their 3 Point Rate being above the historical threshold when considering that their 3 Point FG% is actually higher. It’s worth wondering if this could be the exception to the “40% 3 Point Rate Eliminator Rule”, and makes intuitive sense at the very least.