BracketVoodoo

by Cameron Matthews

BracketVoodoo is back for its eleventh season and our prediction engine has analyzed every play of every game for all 68 teams in this year’s bracket. Below are our favorites to win, based on a combination of team strength (see our Bracket Voodoo Power Rankings) and path to the championship. These odds to win it all incorporate the difficulty of each team’s potential matchups to get out of its region.

This year’s teams at the top have fairly average win probabilities with a bit of a top-heavy skew compared to a typical year. This is in stark contrast to last year, where the favorites had low win probabilities entering the tournament, indicating parity across the bracket. Our top contender this year is Houston with a 22% win probability, or a little better than a 1 in 5 chance. The top five teams have a cumulative win probability of 65%, vs a historical average of 62%.

We'll take a deeper look at a few of these contenders.

HOUSTON is the top ranked team in our Power Rankings and our pre-tournament pick for the second year in a row as the team with the best chance to win it all. Per our models, the Cougars have a 51% chance to make the Final Four and a 22% chance to win it all. Coach Kelvin Sampson has his team back for their 6th tournament appearance in a row and they will be looking to avenge a Sweet 16 exit from last year. 

This year’s team is another defensive powerhouse, holding the top spot in BracketVoodoo’s Defensive Power Rankings. And they’ve added a new offensive dynamic as well, with 3pt threats and experienced playmakers producing a top 20 offense to complement their stellar D. 

So should you pick the Cougars to win it all this year? That depends on your specific pool characteristics. By Tuesday we should have strong data from around the country of which teams are over and under picked relative to their win probabilities - in other words, which teams have an edge. If you’re in a small to midsize pool and think less than a quarter of participants will take Houston, perhaps believing past tournament performance has some bearing on how they’ll do this year, then they will probably still have good value. But we should have more insight on that in the coming days. In the meantime, let’s take a look at a few of the other contenders.

UCONN Last year’s champion is back as this season’s top overall seed and our second most likely team to win it all at 16%. While that might feel like a slight compared to Houston 6 pts  higher, the Huskies win probability is as high as last year’s top seed, highlighting how much the models believe this year’s tournament is a two horse race. And deservedly so. Dan Hurley’s team is a top 5 offense with NBA talent and the experience to make a deep run. 

Remember, though, that win probability is dependent on path/strength of schedule. You can only play who is in front of you, and UConn’s draw in the East is formidable, with three other top-10 teams in their region in Iowa St, Auburn, and Illinois they would have to get through. So while they have the talent and experience to go back-to-back, their road to get there is a tough one. 

AUBURN & ALABAMA are two 4-seeds both out of the SEC, and our #4 and #11 teams in the Power Rankings, respectively. We aren’t alone in our lofty projections for Auburn, with a 6% chance of winning it all. Their stats get a lot of attention from computer models, with efficiency on both sides of the ball. Those that say they are overrated point to a relatively weak schedule, but it’s hard to knock the results. The Tigers went 27-7 on the season and won the SEC championship. With a dominant power forward in Johni Broome and a deep bench, Auburn is a team to challenge UConn in the East. 

Over in the West, Alabama’s intense pace (#4 in our rankings) and high flying offense (#3), coupled with a relatively weaker region, give them a 14% chance to make the Final Four and 2% chance at the championship. Despite some late season struggles, our models like their chances. 

For those 6-seed and below dark horse candidates, BYU has the best chance of making the Final Four at 8%. The Cougars are our 15th most likely team to win it all at 1.4%. BYU boasts the top offense in our Power Rankings. Their cast of sharpshooters can shoot with the best teams in the country and can certainly make a deep run.

Don’t forget that figuring out the teams with the best win probabilities is only the first step in making your bracket selections. The next phase involves understanding the opponents in your pool and devising an optimized bracket strategy to outmaneuver them. BracketVoodoo's optimizer assesses the likely selections made by your opponents and compares them with win probabilities, evaluating the value of each matchup accordingly. In the next day, we'll continue refining our models and delving deeper into the data to scrutinize which teams present the best value propositions for various pool types, all of which will be integrated into our Bracket Optimizer.

Brad Null

We’ve gone through the favorites to win the tourney, and no surprise, Houston and Alabama the top overall seeds are on top, followed by a host of other top seeded teams. But knowing the favorites is only one part of filling out a smart bracket. You also need to look for undervalued teams—teams that have a real chance of advancing in the NCAA tournament but are being overlooked by the general public. In that sense, Alabama doesn’t look like it is necessarily that great of a pick with a 15% chance of winning and 20% of all people picking them.

Our BracketVoodoo optimization engine looks at this public pick data alongside all of the historical game and player data to evaluate millions of bracket combinations and determine optimal brackets for NCAA Tournament pools of various size and scoring parameters. Here are five teams that our analysis has uncovered as teams that could give you an edge in your bracket this year.

1.          Houston - #1 seed / Midwest Region – Houston was the #1 team for much of the year, and would have been the number one overall seed if Memphis had not knocked them off in AAC championship game. That loss, combined with injury concerns for star guard Marcus Sasser, seems to have spooked people a bit. We still have them as the overall favorite to win the tourney at just under a 16% chance. (It was closer to 20% a week ago.) And with only 14% of the public picking them, versus 20% for Alabama, they are not a bad Champion pick to build your bracket around for a smaller pool.

 2.          UCLA - #2 seed / West Region – We see the Bruins as third best team in the country, and also with the third highest odds of winning the tourney at 9.4%. However, due perhaps to their own injury scare, only about 6% of the public is going with UCLA, giving them good value as a team to build around in a small to medium sized pool.

 3.          Tennessee - #4 seed / East Region – Tennessee started the season strong, peaking at 18-3 after a convincing win over Texas in late January. They have floundered since then, to say the least, going 5-7 over their last 12 games (albeit with a win over #1 Alabama). They have been hit by the injury bug as well, but despite all of this, our algorithms still see them as a dangerous team with a 6% chance of winning it all. With under 2% of people picking them, they are a viable play in a larger pool.

4.          Creighton - #6 seed / South Region – This team may just be flying under the radar, but they are a solid pick if you are looking for a middle seed to take a run in a medium sized pool or win it all in a large pool. They’ve got about a 4% chance of making it to the final versus 1% of people picking them. 

5.          Utah State - #10 seed / South Region – Looking for a big underdog to make a splash with in a huge pool or one with underdog bonuses? Check out the Aggies. They have a 10% chance to make the Sweet Sixteen versus only 2% of the public picking them to get past Missouri and Arizona.

While everyone should consider these teams, how far you should project them in your bracket is a function of your pool specifics, most notably your pool’s size and scoring system. UCLA may not be a great pick in your mega-pool in L.A. (unless your buddies are all USC alums). On the other hand, you might seriously consider Utah State to win it all in that 3000 person mega-pool, but the Aggies would be a foolish play in your 10-person office pool.

Cameron Matthews

BracketVoodoo is back for our tenth year analyzing the March Madness tournament. Our prediction engine has analyzed every play of every game for all 68 teams in this year’s bracket and below are our favorites to win. Keep in mind, these may not be exactly the same as who are Power Rankings say are the “best” teams, as these odds to win it all also incorporate the difficulty of each team’s potential matchups in the bracket.

The teams at the top have low win probabilities compared to a typical year. In the past we’ve sometimes seen heavy favorites at 25% or even 30% to win. This year, our top contender is Houston with just a 16% win probability. And the top 5 teams have a cumulative win probability of 54%, vs a historical average of 62%, suggesting a high level of parity, more balance in the seeding, or both. Some of this is due to uncertainty due to key player injuries or off the court issues with a number of the top teams.

HOUSTON is the top ranked team in our Power Rankings, and our pick as the most likely team to cut down the nets. With the number two overall seed, a controversial pick to some who thought Kansas was more deserving for their Quad1 wins (our models actually think the Jayhawks are the weakest 1-seed), the Cougars have a line of sight through the Midwest. Even with the semifinal groin injury to key guard Marcus Sasser and loss in the conference championship, our models like Houston’s matchups in the early rounds even if he does need to rest. Coach Kelvin Sampson has his team back for their 5th tournament appearance in a row with a superb defense.

With a better than 40% chance to make the Final Four, and a 16% chance to win the championship, they are just half a point higher than top overall seed Alabama in our models. So should you take the Cougars this year?

The answer depends on the specifics of the pool or pools you are playing in. If you think many of the other people in your pool will discount Houston’s chances, perhaps fearing that Sasser’s injury will be too much to overcome, they will still have good value. More on that in the coming days as the data starts rolling in for how popular a pick they are. In the meantime, let’s take a look at a couple other contenders.

TENNESSEE is an interesting case of a seemingly under-seeded 4-seed, even after a season ending ACL injury to Zakai Zeigler. Our models still indicate Tennessee is a strong team based on their full body of work and contribution of other players, but it is true that they look to be trying to figure out life after Zakai over the last four games coming into the tournament.

If they do put it together, they have what our Power Rankings have determined are the second weakest 1-seed in Purdue, and weakest 2- and 3-seeds in Marquette and Kansas St. in their region. That relatively easy path gives them an 8% chance of winning, good for fourth. 

UCONN is another 4-seed in the tournament with a solid chance to make a deep run. The Huskies are fourth in our Power Rankings and match up well against other teams in their region. The undersized Kansas Jayhawks, the top seed in the west and weakest 1-seed according to  our analysis, is just one example of how the regional is lined up favorably to UCONN. The big-man pair of Donovan Clingan and Adama Sanogo is a serious matchup challenge for many, including an undersized Kansas.

If you’re looking for a dark horse, CREIGHTON has the best chance of any 6th seed or lower to make the Final Four at 8%. The Bluejays are our 13th most likely team to win it all at 1.5%. Coach McDermott lost to eventual champ Kansas in last year’s tournament and they are an experienced group eying a deeper run.

Dan Loman

With the First Four games underway and the Round of 64 tipping off shortly, here are a few last-minute bracket tips before yours is due:

Aim to survive the first weekend, win in the second and third

A common fallacy in Bracketology is that because there is typically at least one 12 over 5 or 13 over 4 upset every year, you need to pick at least one of those upsets on your bracket. The problem is that if you try to pick every upset, there’s a better chance you’ll pick the wrong one, so from a game theory standpoint you’re better off going with chalk. For the stats nerds reading, here’s how it breaks down mathematically:

Since 1985, 5 seeds have beaten 12 seeds 65% of the time. If that rate holds, and you pick all 5 seeds to win, then for each game you have a 65% chance of getting it right. That results in an expected value of 2.6 correctly picked teams going into the next round:

4*(1*.65 + 0*.35) = 2.6

If you try and pick one 12 seed to win, then for each game you have a 75% chance of picking a team with a 65% chance of winning, and a 25% chance of picking a team with a 35% chance of winning. That results in an expected value of 2.3 correctly picked teams:

4*(.75*.65 + .25*.35) = 2.3

If you go too crazy picking first round upsets, you can dig yourself in a hole after the first weekend that can be tough to recover from. The objective isn’t to pick a perfect bracket, because that’s pretty much impossible, but to beat your opponents. Of course, you have to nail a few contrarian picks to differentiate your bracket from the rest of the pool, but the value generally isn’t there in the first few rounds. It’s better to focus on one or two value picks in later rounds, so you only have to worry about a few vig things going right rather than a bunch of little ones. Our engine likes Houston and Purdue as solid under-the-radar plays that could make a run this year.

Understand your competition

How deep of a sleeper to roll with, and how far to take them is a function of your specific pool. In small pools it makes more sense to be risk-averse, but in large pools your bracket really needs to stand out. For example, our engine gives Houston about a 3% chance of winning it all, but only 1% of brackets are picking them - that gives them strong value in larger pools, but the math doesn’t work out the same in say, a 10 person pool. If your pool has an alternate scoring system, like 1-2-3-4-5-6, keep in mind that early picks will have greater relative value.

Lastly, be aware of any group biases. If you’re in a pool with a bunch of Arizona alumni, it might be a good idea to fade Arizona from a strategic standpoint.

Don’t overthink it

Humans are inherently bad at parsing information. We’re wired to look for patterns and explanations everywhere - this is why correlation vs causation is such a cliche. We HATE the idea that randomness exists, and that outcomes may be totally different if run in a parallel universe. When we fill out our bracket, we need to tell ourselves a story for why it makes sense in order to feel good: Iowa is playing their best basketball right now, Duke has the most NBA talent, Kansas has the experience - these are all things that have run through my mind since the selection show that our analyses have shown aren’t too important.

And if that’s not bad enough, our memory sucks too. We’re horribly selective in what we remember and we’re equally bad at interpreting it. It’s likely there are a few games from the regular season that stick out in your mind, but it’s important to remember that that’s a one or two game sample out of 30. Just as importantly, don’t get hung up on prior NCAA Tournament results, as there’s no correlation between prior year and current year team results in the tournament. If you discounted Virginia before their championship run in 2019 because they burned you after losing to UMBC in 2018, you probably lost out. On the flipside, don’t automatically pencil Loyola-Chicago into the Sweet Sixteen just because they’ve outperformed in March recently.

Despite our annual complaints the committee usually does a pretty good job of sorting out all of these variables to the point where there’s no real value to be found in any particular narrative. Our power rankings, Kenpom ratings, and Sportsbook odds are generally more reliable indicators of team quality than any single game or narrative. 

Brad Null and Cam Matthews

Sunday night we broke down the favorites to win the 2022 NCAA Tournament and no surprise, top ranked Gonzaga is the most likely team to win at a whopping 26%, the highest percentage since… 2021 Gonzaga was also a 26% chance. The next highest team, Arizona (10%), is also the largest marginal difference between the 1 and 2 we’ve seen. This all feels like deja vu to us, but the Zags are certainly hoping it turns out a little different than last year. 

Understanding the favorites and their paths to victory is only the first step in filling out a winning bracket. Picking the best bracket for your pool means consciously playing against the other players in your pool - you don’t need a perfect bracket. Playing to beat 15 people is very different than 150.

Finding teams that are undervalued - teams that have a real chance of advancing to later rounds but are being overlooked by the average player - is key as pool size increases. The more entries, the more you need to differentiate. At BracketVoodoo we call this a team’s round-by-round odds vs public pick percentage (RbRvPPP). Let’s say you pick a champion that has a 5% chance of winning. However, only 1 in 50 brackets is picking them to win. That gives you a RbRvPPP of 2.5 (anything >1 is considered a “value” pick). In a pool of 50 players, that’s a pretty good pick! But make that same pick in a pool with just 10 people and 19 out of 20 times you are going to be in trouble. 

That’s where our Optimization Engine comes in to generate gambits for different pool sizes. We look at data from picks all over the country as well as all of the historical play-by-play data to evaluate millions of bracket combinations and determine optimal brackets for a variety of pools. Here are four teams that our analysis highlighted as teams that could give you an edge in your bracket. These are for standard pool scoring of 1-2-4-8-16-32, unless otherwise specified.

Gonzaga - #1 seed / West Region - The Bulldogs are once again the heavy favorite at more than a 1-in-4 chance to cut down the nets. Not surprisingly, they are underpicked by the public compared to their win probability. Pool players are likely reluctant to hop on the bandwagon after getting burned by other recent “sure thing” picks like 2015 Kentucky, 2019 Duke, or last year’s version of the ‘Zags. But Gonzaga has regrouped and has a favorable path. We think Gonzaga is a very strong pick for pools with ~60 or fewer entries.

Houston - #5 seed / South - Houston looks to be underseeded from our analysis. At #6 in our Power Rankings, we’re having a hard time not picking the Cougers to make a deep run in almost every pool size and type. Injuries to key players, especially Sasser back in December, may be making them undervalued by the public, but our projections take this into account and still aren’t deterred. We see them with about a 1 in 20 chance of winning the championship, while only about 1 in 100 are picking them. In a pool of just 8-10 people it may still not be worth it, but aside from those very small pools they seem like a great value.

Purdue - #3 seed / East - For mid to large size pools, the Boilermakers could make an interesting gambit coming out of their region. More than twice as many people are picking Baylor or Kentucky, but our models indicate their chances of advancing are only about 50% better than Purdue’s, who sit at 14 in our most recent Power Rankings. 

USF - #10 seed / East - For pools with upset bonuses or seed multipliers, the 10-seed Don’s could be a valuable pick for 2 or 3 wins. Even in a large pool with bonus scoring, you don’t typically want to try to pick the “right” 10+ seed to make a deep run, but a strategic Elite 8 finish could really help to differentiate your bracket in an upset pool, especially with a lot of entries. We aren’t as high on them as some, but worth taking a flyer on if you get extra points for the “10” next to their name.

Keep in mind, how far you should project these teams in your bracket is a function of your pool specifics, most notably your pool’s size and scoring system. 

Dan Loman

The great thing about the NCAA Tournament is that it has something for everyone. The non-sports fan can get behind the spirit, pride and the competition. The gamblers can get their fix. NBA fans get to watch the next generation of elite hoopers. For college basketball fans, it is of course the pinnacle of the sports year.

If you’re an NBA and NCAA basketball fan, well this year’s March Madness tournament has everything for you. With a light international crop of players entering the 2022 NBA draft, the top talent is playing in college basketball - and just about all of them will be heavily featured in the NCAA Tournament. According to The Athletic’s latest Big Board, the top 5 players* projected for the draft all play for top 3 seeds: Chet Holmgren of Gonzaga, Jabari Smith of Auburn, Jaden Ivey of Purdue, and Paolo Banchero and AJ Griffin of Duke. Over the past 10 tournaments (not counting 2020), this would be only the third year in which every top 5 pick is featured in the NCAA Tournament - and the average seed of 2.0 is tied for the lowest of all other years.

*Not counting Shaedon Sharpe, who has stated he plans on returning to Kentucky

If we expand out to the lottery (top 14 picks), all but one projected player (Dyson Daniels of the G-League Ignite) will be playing in this year’s tournament - no other tournament in the past decade featured as many lottery picks as 2022 is expected to. Not only that, but with an average seed of 2.69, these future NBA players are all playing on teams poised to make serious runs in March. On the eve of Selection Sunday, Jalen Duren of Memphis and Keegan Murray of Iowa are the only ones of the group playing for teams lower than a top 3 seed. That puts 2022 just slightly behind 2017 and 2019 in terms of team quality of lotto picks:

With an unusual amount of star power featured in this year’s NCAA Tournament, it’s sure to be a blast. But what does this mean for your bracket? Should you give extra credit to Duke, for example, for having more NBA caliber talent on the roster? As you might expect, the results are mixed. It tends to be easier remembering the successes, so powerhouse teams with multiple lottery picks like 2012 Kentucky and 2015 Duke stick out. Speaking of the Blue Devils and Wildcats, two of the most dominant teams post-Anthony Davis - 2015 Kentucky and 2019 Duke - weren’t completely disappointing but likely killed your bracket if you picked them to win it all. And then there’s the 2014 Kansas Jayhawks, who despite featuring top 3 picks Andrew Wiggins and Joel Embiid failed to even make the 2nd weekend. The oft-forgotten 2013 Indiana Hoosiers are another team with two top 5 picks that lost in the Sweet Sixteen.

Overall, teams featuring future NBA talent outperform expectations slightly based on their initial seed, though not to a significant degree. Here’s how it breaks down by year:

There’s also a correlation/causation issue in this analysis, where players might assert themselves as lottery picks because of a strong NCAA Tournament performance (Kemba Walker comes to mind). With this year’s NBA draft as loose as it is, it seems very likely the tournament could do the same for one of this year’s prospects.

Cam Matthews

Another year of March Madness is upon us! Our prediction engine has analyzed each team in the bracket and every play of this 2021-22 season. Below are BracketVoodoo favorites to win this year’s tournament. Outside of the top 3, the rest of our top 10 doesn’t closely follow our Power Rankings. That’s because our “path to championship” odds also incorporate the difficulty of each team’s potential matchups.

Gonzaga is our heavy pre-tourney favorite for the second year in a row. And while their hopes of a perfect season in 2021 fell just short, they have regrouped and look like the clear team to beat. The other storyline we’re following is a wide open South region, where we project four teams in our top 10 mostly likely champs.

GONZAGA is the top overall seed, the top ranked team in our Power Rankings, and our pick as the most likely team to cut down the nets. Upperclassmen Drew Timme, Andrew Nembhard, and freshman sensation Chet Holmgren lead the 26-3 Bulldogs, who are looking to avenge their 2021 title game loss to Baylor. With a better than 1-in-2 chance to make the Final Four, a 26% chance to win the championship, and nearly 3x as likely as the next team down the list, they are the most heavily favored team in our models since… 2021 Gonzaga and 2015 Kentucky before that. You might recall that neither of those heavy favorites wound up showered in confetti when it was all said and done. So should you take the Zags this year?

The answer depends on the dynamics of your particular pool. If you think many of the other people in your pool will discount Gonzaga’s chances, believing they can’t win the big one, they may still have good value. More on that in the coming days. In the meantime, let’s take a look at some of the other contenders.

ARIZONA is the 1-seed in the South and the BracketVoodoo #2 ranked team. The Wildcats, a team with a blazing pace, have a 29% chance to make the Final Four and a 9% chance to win it all after winning the Pac 12 regular season and conference tournament. Still, our models place them significantly below Gonzaga. They face a tough potential Sweet 16 matchup with 5-seed Houston, and tough regional competition in Villanova or Tennessee. Arizona is a strong team with a tough draw.

With a win in the Big 12 Championship over Texas Tech, KANSAS gets the Midwest region 1-seed, our #3 ranking, and is our 3rd most likely to win it all. The Jayhawks had the 3rd toughest strength of schedule on the season and are strong on both offense and defense. Five of their six losses were to Quad 1 opponents, and they are playing excellent basketball right now. Their path through the Midwest looks more favorable than some of the other regions.

Other notables…

HOUSTON, who crushed #9 seed Memphis in the AAC Tournament, appears to be vastly underseeded compared to our power rankings. The committee seems to have knocked them for their lack of Quad 1 wins, but our engine loves how they generally dominated weaker competition while losing very close to stronger opponents. And with a Sweet 16 site less than a 3 hour drive to San Antonio, the Cougars may have the wind in their sails to come out of the South.

LSU has the best chance of any 6th seed or lower to make the Final Four at about 1 in 12. Despite entering the tournament just 7-10 over their past 17 games, and with their coach Will Wade fired amidst allegations of major violations, our engine rates them highly. Driven by a strong non-conference performance the Tigers are 19th in our rankings, which is actually 13 spots ahead of potential 2nd round opponent Wisconsin and just 7 spots below potential Sweet 16 matchup Auburn.

Brad Null & Cam Matthews

Sunday night we broke down the favorites to win the 2021 NCAA Tournament and no surprise, undefeated and top ranked Gonzaga is the most likely team to win at a whopping 26%, the highest percentage since 2015 Kentucky was a nearly 1-in-3 chance. The next highest team, Illinois (12%), is also the largest marginal difference between the 1 and 2 we’ve seen since then. But they call it March Madness for a reason, and the ‘Zags are far from a sure thing.

Understanding the favorites and their paths to victory is only the first step in filling out a winning bracket. Picking the best bracket for your pool is a little like the old saying, “You don’t need to be faster than the bear, just faster than the guy next to you.” In the case of March Madness you don’t need a perfect bracket, you just need to run faster than the other people in your pool. Running a race against 9 people is very different than 900.

Finding teams that are undervalued - teams that have a real chance of advancing to later rounds but are being overlooked by the average player - is key as pool size increases. The more entries, the more you need to differentiate and look for under-picked teams. At BracketVoodoo we call this a team’s round-by-round odds vs public pick percentage (RbRvPPP). Let’s say you pick a champion that has a 2% chance of winning. However, only 1 in 200 (.5%) brackets has them picked to win. That gives you a RbRvPPP of 4.0 (anything >1 is considered a “value” pick). In a pool of 200, that’s a great pick! 2% of the time you’d expect about a coin flip of winning (vs the other expected entry with the same champion). That’s an ROI of 2x over an “average” bracket with a 0.5% chance. But make that same pick in a pool with 10 people and in 98% of the outcomes you’re going to be in big trouble. So even though your team has a high RbRvPPP, those are bad odds for the pool size.

That’s where our Optimization Engine comes in to generate gambits for different pool sizes. We look at public pick data alongside all of the historical game and player data to evaluate millions of bracket combinations and determine optimal brackets for NCAA Tournament pools of various size and scoring parameters. Here are five teams that our analysis has uncovered as teams that could give you an edge in your bracket this year. These are for standard pool scoring of 1-2-4-8-16-32, unless otherwise specified.

Gonzaga - #1 seed / West Region - What can we say, the Bulldogs are the heavy favorite at more than a 1-in-4 chance to cut down the nets, and surprisingly they are not as heavily picked by the public as other recent favorites like 2015 Kentucky or 2019 Duke (both of whom we advised against and both of whom did lose). The undefeated ‘Zags have a deep roster loaded with NBA talent and they are the favorite for a reason. Maybe feeling burned by previous “sure thing” busts, the market has overcorrected in 2021. Whatever the case, we think Gonzaga is a very strong pick for pools of ~20 entries or fewer.

Baylor - #1 seed / South Region - sandwiched between the small and medium sized pools is a range of ~20-30 bracket entries, where some differentiation will start to help. The Bears are a good counter to the favorite ‘Zags for pools of this size. With about a 10% chance of winning and also about a 10% public pick percentage, they provide a little more value, but still with good likelihood, than the other 1 seeds.

Iowa - #2 seed / West Region - Well that’s interesting. The team that would (likely) have to beat the ‘Zags just to get to the Final Four looks like a great gambit to win the whole tournament for medium sized pools of 35 to 100+ entries. The Hawkeyes, ranked #5 in our Power Rankings and the strongest 2-seed, have about a 1-in-3 chance of beating the ‘Zags in that hypothetical head-to-head matchup in the West and a decent path overall. They become a good value at that inflection zone between 30-40 pool entries all the way up to 120 or so, where it’s definitely advantageous to get a little contrarian.

Houston - #2 seed / Midwest Region - For larger pools of 150 to 400 entries,  where a differentiation becomes really important, the Cougars look like a solid play. At #6 in our Power Rankings, we give them about 5% chance at the championship, yet their public pick percentage is only 2.4%, giving them a RbRvPPP of >2. Part of what’s at play is that the #1 seed in their region, Illinois, seems to be a popular contrarian foil to Gonzaga. Too popular in fact. With a 12% chance to win, they are being picked in more that 15% of brackets. All that adds up to Houston being a great value for those larger pools

Tennesee - #5 seed / Midwest Region or Wisconsin - #9 seed / South Region - For very large pools of 500+, you should be looking for significant differentiation and teams with high RbRvPPP values. The Volunteers have about a 1.5% chance of winning in our models and are only being picked in 0.4% of brackets nationally, for about a 3.5 value rating. Similarly, the Badgers have about a 1% chance of winning in our models and are only being picked in 0.3% of brackets nationally. These are not teams you’re likely going to want to put in the winners circle in your friendly pool with your 10 buddies unless you like to live on the edge, but having them advance deep in your giant pool or even a seed-weighted scoring pool looks like a pretty good play.

While everyone should consider these teams, how far you should project them in your bracket is a function of your pool specifics, most notably your pool’s size and scoring system. 

Dan Loman

It’s been a weird year to say the least. From empty stadiums to covid pauses, uncertainty has been the theme of the 2020-21 season. Despite all that, I’m thrilled to be filling out a bracket again and here to offer some tips for a tournament that could be madder than ever.

Embrace the randomness...by not trying to fight it

That sounds counterintuitive. If uncertainty is higher than ever this year, shouldn’t we prepare for more upsets? Well there might be more madness than usual in 2021...but good luck picking the teams that create it. If you get too crazy picking early upsets there’s a better chance you’ll end up in a deeper hole by picking the wrong ones.

Take it from someone who had a perfect bracket in 2020 - a better strategy is to focus on a few long-term plays that carry more value. That way as long as those teams make it through the first few rounds unscathed, you can sit back and enjoy the craziness knowing your bracket isn’t busted.

Don’t fall for false narratives

Part of not fighting randomness is not falling for biases and false narratives, and there’s an extra amount of noise this year. Will the limited crowd affect some teams more than others? What’s the impact on limited travel after the first weekend? Which teams will lack the mental fortitude to survive eating cold food? While it’s smart to keep up on Covid-related news like Kansas having to navigate a smaller roster, you’re more likely to make a mistake by overthinking all of the additional variables than hitting on any real signal.

At BracketVoodoo we’ve spent a lot of time studying the impact of various talking points on tourney performance, and the results are mostly boring: Momentum is a myth, much of tourney performance comes down to shooting luck, and recent tournament performance doesn’t matter (actually, Tom Izzo’s Spartans would go on to lose to 15 seed MTSU less than a week after that last post was published to further hammer home the point). Though we can’t test covid-related theories, I’d suspect it’s more of the same.

Stand out from the pack

Cam Matthews

March Madness is back! It feels good to say that, even if things are a little more in flux than a normal Selection Sunday. The name of the game this year is uncertainty. 68 teams have been selected, but some may need to be replaced before Tuesday. Four play-in games are scheduled for Thursday and thirty two opening round games for Friday and Saturday, but it is unknown whether some will need to be played shorthanded or even forfeit. COVID-19 protocols could very well reshape the bracket that was announced this evening. Also unclear is how the Indianapolis bubble, with no travel and limited fans, will change the tournament dynamics. Patience is key.

That said, we have to start somewhere. Our prediction engine has analyzed each team in the initial bracket and every play of this one-of-a-kind season. Here are the BracketVoodoo favorites to win this year’s tournament. The top 10 list stays pretty true to our Power Rankings at the top, but starts to diverge in the latter half as it also incorporates the difficulty of each team’s championship path.

Gonzaga is the top overall seed, the top ranked team in our Power Rankings, and our pick as the most likely team to cut down the nets in 2021. Jalen Suggs, Drew Timme, and Korey Kispert lead the 26-0 Bulldogs, who would become the first team since Indiana in 1976 to record a perfect season if they run through the tourney. With a 1-in-2 chance to make the Final Four, a better than 1-in-4 chance to win the championship, and more than twice as likely as the next team down the list, they are the most heavily favored team by our models since 2015 Kentucky had a 1-in-3 chance.

No doubt Gonzaga will be popular in many brackets. A word of caution, though, before you pencil them in the winners box. Heavy favorites, like that 2015 Kentucky team, sometimes don’t meet expectations. Long time BracketVoodoo fans patrons will remember us advising against selecting the Wildcats in 2015 due to being too popular. This year with the added variance introduced by the bubble health and safety protocols, we may find yet another heavy favorite that is too overpicked to be valuable in your pool. More on that in the coming days, though, stay tuned.

After the favorites, the championship probabilities for the next teams down the list are relatively spread. The 1-seeds from the other regions, Illinois, Baylor, and Michigan are together in a tier at 9-12%, and the two seeds all come in between 4-7%.

Illinois is the 1-seed in the Midwest and the Bracket Voodoo #2 ranked team. The Illini have a 35% chance to make the Final Four and a 12% change to win on the strength of winning college basketball’s toughest conference. Still, our models place them significantly below Gonzaga. They face a tough potential matchup with 2-seed Houston, but have an easier path than other 1-seeds as their potential 3 and 4 seeds matchups (West Virginia and Oklahoma St.) are not as strong.

Baylor, the South region 1-seed and our #3 ranked team is our 3rd most likely to win. The Bears are a lights out 3-pt shooting team with some questions on defense. With their only losses coming to Kansas and Oklahoma St in the Big12 tournament, they know how to win games. But they face very tough potential matchups in Ohio St and 5-seed Villanova, who our power rankings view as underseeded as the 9th best team.

4th and 5th on our list are two more Big 10 teams, Michigan and Iowa at 9% and 7% chance of winning. Both have the talent to make a run, but have questions around depth due to injury concerns. Iowa is also on course for a  tough matchup with aforementioned Gonzaga should both hold serve in their half of the West.

All in all, this tournament has a clear favorite in Gonzaga followed by parity and uncertainty. Excluding the heavily favored top-seeded Bulldogs, this year’s champion is the least likely to come from the 2-10 grouping since we began tracking 8 years ago. In the chart below you can see Gonzaga’s 26% win probability, rising to only ~80% when adding the next 10 teams.

Cumulative Championship Probability of Tourney Favorites

Cumulative Championship Probability of Teams 2-11

Dan Loman

In our previous blog post, we outlined some indicators for early round upsets by double-digit seeds. This time we’ve switched gears to identify some at-risk higher seeded teams you might want to avoid.

One of our top pieces of advice to win your pool is to focus on one or two big picks, and hitch your wagon to them. Of course, if you get those picks wrong, you’re dead in the water. How can you avoid picking the wrong favorite? This blog post is here to help.

For this analysis, I bluntly defined two populations: Disappointments are teams who won at least 1 game fewer than expected (like Tennessee and Arizona last year). They make it out of the first round, but hurt your bracket if you had them going too far. Busts are teams who won at least 1.5 games fewer than expected, and comprise of the most notable high seeds that killed your bracket (like Virginia). I used the same features as the previous model to assess the likelihood of a team falling into one of these groups: advanced team stats, miscellaneous data from Kenpom.com, and proxies for momentum.

Here’s what stuck out among the Disappointments:

They are 3 happy. Yesterday we saw that Cinderellas tend to be better 3-point shooting teams, so why is the opposite the case here? It’s all about variance; if a team is the underdog, they’ll need a few good 3 point shooting games to advance, while a team that landed a high seed largely off the strength of their shooting only needs one bad shooting night to be bounced early. Michigan St. is a great example of this last year, when they shot just 8/27 from 3 against Syracuse despite shooting 40% there for the year.

They rely on (undersized) bigs. High seeds in our database that failed to live up to expectations tend to get more scoring out of their big men than guards. This falls in line with some prior research we’ve done here, that suggested that teams are helped by strong lead guard play. They fared worse when the bigs were undersized, like with Creighton and Villanova in 2014 (Dougie McBuckets!).

In this year’s field, Virginia Tech, Auburn and Houston are teams you might want to be wary of penciling in to the second weekend or Elite 8. The Hokies check both of those boxes by taking 45% of their shots from deep and hitting at a 39.4% clip, and by only playing one guy in their rotation over 6’6’’. Bruce Pearl’s Tigers have proper size but take nearly 50% of their shots from 3, while Houston’s tallest rotation player is 6’8”.

And here are the top qualities of Busts:

They play a high variance style. Similar to Disappointments but with a major distinction: the top predictor of teams that lose much earlier than expected is that they play with a slow pace. Slower pace means fewer possessions, which increases the variance of outcomes. That’s great for the underdog, but potentially bad news for the favorite. Busts also shoot more 3s and have a higher offensive rebounding percentage, suggesting they crash the boards more.

They lack size. A high variance style can be countered by overwhelming a lesser opponent with size, like Bo Ryan’s Wisconsin teams who always played slow but had NBA size up front. Be wary of “small giants”.

Though we still think Virginia is a strong title pick based on their overall team quality and undervaluation by the market, they do exhibit the biggest red flags among favorites in the field. They play at by far the slowest pace in the country, are great at shooting the 3 and crash the boards more often than last season. The good news for the Cavs is that recency bias is overstated and historical team tourney performance doesn’t matter, so they are no more or less likely to repeat their historic loss from 2018. Michigan St. is another team that shows similar qualities, but they aren’t nearly as slow as UVA.

One of the more interesting teams in the field is North Carolina. The Tar Heels have little risk of falling early due to their blistering pace, but could be a vulnerable second weekend team due to the amount their offense relies an undersized big man - 6’8” Luke Maye. Duke and Tennessee seem like the safest options to at least not go home earlier than expected. Both teams are big, don’t rely too much on outside shooting, and Duke plays at one of the fastest paces in the country.

Dan Loman

The bracket is out for 2019! I hope you’re as excited as we are as we gear up for another year of March Madness.

One of the big pieces of advice we give here at Bracket Voodoo is not to chase upset picks. If you go reaching for too many early round upsets, there’s a good chance you’ll pick the wrong ones and find yourself dug in a hole at the end of the first weekend. From a game theory perspective, that makes perfect sense and will help you win your pool. But if you’re in a contest with your friends and coworkers, you know that one of the most fun parts of the tournament is nailing that upset pick. So if you decide to  get a little risky in the first few rounds, this blog post will help you pick the right ones.

In the first part of a two-part analysis, I focused on higher seeds that outperformed expectations since 2010. This study is in the same spirit as ESPN’s Giant Killers series, but with different parameters: I looked at teams seeded 10-15 (sorry UMBC) - the ones ripe for that first round upset or Elite 8 cinderella run. From there I homed in on two different groups: Upsetters - teams who won at least one game in the tournament (like 13 seed Buffalo in 2018 or 11 seed Rhode Island in 2017) - and Cinderellas, the subset of Upsetters who pulled off a hugely unlikely win (like 15 seed Middle Tennessee State in 2016) or went on a hugely unlikely run (like Loyola-Chicago last year).

So what kind of double digit seed is ripe pull off an upset? Using a model trained on NCAA Tournament teams from 2010-2018 with features like advanced team stats, miscellaneous data from Kenpom.com, and proxies for momentum, these features stuck out as the most among Upsetters:

They win the turnover battle. A common finding with the Giant Killer series and one that makes intuitive sense, double-digit seeds that pull off first round victories tend to take the ball away more than they turn it over themselves. While turnover defense is the most important thing to look for, it doesn’t hurt for teams to be careful taking care of the ball themselves. Winning the turnover battle is a great way for a less talented team to steal a victory over a more talented opponent. Stephen F. Austin in both 2014 and 2016 showed how a low-seeded team who forced a ton of turnovers can fare in the tournament; both years they ranked in the top 10 in forced turnover rate, and both years they knocked off a higher seeded opponent (5 seed VCU and 3 seed West Virginia, respectively).  In the 2019 field, St. John’s, Florida and Temple have the best-rated turnover differentials among double-digit teams. And while we wouldn’t fully endorse picking a 15 over 2, Abilene Christian ranks eighth nationally in forced turnover rate.

They have size/pedigree. Among double-digit teams in the dataset, major conference teams were far more likely than mid-majors to pull off an upset. Of course, major conference teams also had better seeds on average, but even adjusting for seed, Upsetters tend to be bigger, especially at the guard spots. While not as sexy of an upset pick, an 11 seed major conference team (like Ohio St or Arizona State/St. Johns) stands a much better chance of advancing to the Round of 32 than the trendy 13 seed mid-major.

They don’t have a deep bench. This was a topic we covered at BracketVoodoo before, and ironically Duke is again a subject. Benches run shorter in the NCAA Tournament - roughly 6% more minutes in the Tournament come from starters - so teams that rely on depth are at a disadvantage. That benefits St. John’s and Saint Louis, who both fall in the bottom 13 of bench utilization, and you might want to be wary of New Mexico St. and UC Irvine, who have the first and seventh highest rates of minutes coming from their bench, respectively.

What interestingly doesn’t strongly correlate with upsets, is anything to do with 3-point shooting. While it is true that upsets are often driven by hot shooting, there is little correlation with 3 point shooting during the regular season. It seems the best bet for an upset is a team that got into the tournament off of its other strengths, and then shot abnormally well.

.tg { margin: 20px auto; border-collapse:collapse; border-spacing:0; } .tg th, td { font-family: Arial, sans-serif; font-size: 12px; border-style: solid; border-width: 1px; overflow: hidden; word-break: normal; } .tg th { padding: 5px 20px; font-weight: bold; } .tg td { padding:3px 20px; } .tg .tg-yw4l { vertical-align:top; text-align: center; } .accuracies { margin: auto; text-align: center; }

Ford Higgins

After the first weekend of Madness, there’s no need to remind our readers that game projections are notoriously difficult - just ask the Cavaliers, Musketeers, Bearcats, or Tar Heels what they think of the latest power rankings. The unpredictability is one of the reasons why we love the tournament, but it presents a conundrum when we sit down to fill out our brackets. For our readers who rely on public statistical systems (including ours) for bracket assistance, I wanted to dive deeper into how to evaluate projections, and how to avoid pitfalls hidden in the data.

Methodology

I judged the accuracy of different projection systems by analyzing projected game outcomes, incorporating home court advantages when possible, and comparing those projections against actual game results. The dataset includes most Division 1 regular season and conference tournament games from the 2017-18 season. Because of data issues with the underlying data source, a small percentage of games are discarded.

I evaluated three different systems:

Ken Pomeroy

Sonny Moore

Jeff Sagarin

I also included a home court advantage factor. Moore and Sagarin provide specific home-court advantages that can be applied across all games, which assumes every team has an identical home court advantage. KenPom, on the other hand, applies team-specific home court advantages in his own predictions and I simply used the average of the advantages in place of the individual values to make predictions with his system. I also assumed home court advantage applied to the listed home team in neutral site games.

System Accuracy In Conf Out of Conf Home Away Games Home Court Pomeroy 73.50% 70.40% 78.40% 78.20% 62.80% 5286 3.12 Moore 72.40% 69.50% 76.90% 76.40% 62.00% 5315 4.00 Sagarin 73.70% 70.90% 78.30% 76.90% 65.30% 5283 3.17

I hypothesized that all three projection systems would begin the season with limited accuracy, with each system becoming more accurate over the course of the season (as more information about each team became available). Surprisingly, this proved false, each system enjoyed an early season peak, declined until final exams and the holiday season, after which performance leveled off or improved slightly until conference play started. Some of the early-season projection accuracy can be attributed to the non-competitiveness of “buy games”, with high major teams padding their win total against low major teams in need of money (including 2018 tourney darling, Texas Southern). I was also surprised to see projection accuracy continue to decline throughout the conference season, especially during February, when teams are usually facing opponents for a second time in a season. In the end, all three systems ended the season around 73% accuracy.

Next, I sliced the projections into common splits and analyzed how each system performed for these subsets:

In-conference versus out-of-conference and games

Picking home team winners versus picking away teams

In general, we would expect the accuracies for these splits to be randomly distributed around the overall average of ~73%. However, since we know that most out-of-conference games are played during the first half of the season, I expected the in/out conference splits to follow the overall time series patterns discussed above - and that’s exactly what I saw. Since I expected richness of head-to-head matchup history, to improve projection accuracy, I was especially surprised by these results.

Looking at home/away splits provided us with the most surprisingly results of all. Our #1 ranked system, Jeff Sagarin, was over 10% more accurate when picking home teams to win, versus picking away teams to win. This pattern not only holds up for the other projection systems, but is even more pronounced for Moore (76.40% vs 62.00%, -13%) and KenPom (78.20% vs 62.80%, -16%). Seeing all three systems share the same bias toward over-picking away teams suggests it may be time to revisit how home court advantage is factored into these models.

Note: Since KenPom generates team-specific home court factors, it is likely possible that incorporating those values would slightly improve the home/away splits.

The Sweet Sixteen

Kansas State vs Kentucky - South Region (Atlanta)

(5) Kentucky (9) Kansas St. KenPom 20.44 15.36 KenPom OOC 79.10% 83.50% Moore 82.35 78.43 Moore OOC 76.30% 81.30% Sagarin 88.21 84.31 Sagarin OOC 77.80% 82.30% BracketVoodoo 43.16 38.62 Vegas Spread -5.5 Vegas ML -255 +215

All three systems have Kentucky as a four or five point favorite, though the Atlanta location could make it a semi-home game for the Wildcats in blue. Interestingly, all three systems excel at correctly predicting outcomes for Big 12 out of conference games while having the lowest accuracy of any power conference with games involving an SEC team playing out of conference.

Loyola-Chicago vs Nevada - South Region (Atlanta)

(11) Loyola (IL) (7) Nevada KenPom 15.26 18.13 OOC 73% 77.20% Moore 74.98 78.4 OOC 75.40% 77.00% Sagarin 81.01 84.37 OOC 73.90% 80.20% BracketVoodoo 36.63 40.01 Vegas Spread -1.5 Vegas ML 105 -125 OOC Games 115 101 OOC Games 190 139

This game projects as a close game, with each system pegging the difference in teams to be about three points. It is also the only game where neither team comes from a power conference, which could be influencing why the systems rate them closely. The relative lack of accuracy for both teams’ conferences in non-conference games suggests that the models might be less trustworthy in this matchup than others.

Florida State vs Gonzaga - West Region (Los Angeles)

(9) Florida St. (4) Gonzaga KenPom 16.21 24.35 OOC 82.00% 78.70% Moore 81.14 81.92 OOC 77.20% 76.20% Sagarin 85.45 89.67 OOC 80.20% 77.00% BracketVoodoo 40.92 44.55 Vegas Spread -5.5 Vegas ML 210 -250 OOC Games 206 122

This game features the most variety among the systems - KenPom and Sagarin have Gonzaga as varying degrees of favorites, whereas Moore has the game as basically a coin flip. But Moore has also had the least success with both conferences, so I would shade towards the other systems. On top of that, the LA crowd could also be slightly in the Bulldogs’ favor as a West Coast team.

Texas A&M vs Michigan - West Region (Los Angeles)

(7) Texas A&M (3) Michigan KenPom 16.91 23.41 OOC 85.30% 79.10% Moore 79.95 82.74 OOC 82.70% 76.30% Sagarin 85.43 89.01 OOC 83.80% 77.80% BracketVoodoo 42.22 44.41 Vegas Spread -2.5 Vegas ML 130 -150 OOC Games 185 190

Michigan is favored by all three systems, with KenPom favoring the Wolverines the most. The Big 10 enjoys the second most accurate out of conference predictions across the systems whereas the SEC has the least accurate, providing an interesting contrast in conference features. However, KenPom has the largest split between the two teams and performs the best for both conferences, increasing my confidence in his prediction of Michigan.

Villanova vs West Virginia - East Region (Boston)

(1) Villanova (5) West Virginia KenPom 31.7 22.29 OOC 86.60% 83.50% Moore 88.62 83.63 OOC 86.40% 81.30% Sagarin 94.79 90.2 OOC 85.90% 82.30% BracketVoodoo 51.07 45.71 Vegas Spread -5.5 Vegas ML -230 195 OOC Games 134 139

Villanova, the top remaining seed by both the committee and all three systems, is a five point favorite according to Moore and Sagarin. KenPom has the difference between the two teams as nearly double that at about nine and a half points. The crowd in Boston will likely be pro-Nova, which would further boost their chances. Furthermore, all three models do their best work with Big East out of conference games, making Villanova even more favored by them.

Texas Tech vs Purdue - East Region (Boston)

(3) Texas Tech (2) Purdue KenPom 21.99 26.99 OOC 83.50% 85.30% Moore 82.83 88.1 OOC 81.30% 82.70% Sagarin 88.45 92.35 OOC 82.30% 83.80% BracketVoodoo 44.02 47.76 Vegas Spread -1.5 Vegas ML 105 -125 OOC Games 129 185

Despite this being a 2/3 matchup in the East region, every system has Purdue as about a five point favorite over Texas Tech. However, the systems are likely unaware of Isaac Haas’ absence for Purdue due to a fractured elbow, which should help Texas Tech close the gap.

Kansas v Clemson - Midwest Region (Omaha)

(1) Kansas (5) Clemson KenPom 23.43 20.25 OOC 83.50% 82.00% Moore 85.6 80.84 OOC 81.30% 77.20% Sagarin 91.72 85.97 OOC 82.30% 80.20% BracketVoodoo 45.98 42.81 Vegas Spread -4.5 Vegas ML -210 180 OOC Games 139 206

This matchup between 1 seed Kansas and 5 seed Clemson is the closest any game comes to a true home/away situation in the Sweet 16, and could make us feel even more confident in the consensus pick of Kansas. KenPom rates the teams as closer than either Moore or Sagarin, which is a departure from some of our other matchups where KenPom had the largest differential.

Duke v Syracuse - Midwest Region (Omaha)

(2) Duke (11) Syracuse KenPom 28.95 14.04 OOC 71.10% 71.10% Moore 88.5 79.06 OOC 72.40% 72.40% Sagarin 93.64 83.15 OOC 73.80% 73.80% BracketVoodoo 50.11 39.02 Vegas Spread -11.5 Vegas ML 575 -795 OOC Games 149 149

Note: Since these are both ACC teams, for this game we looked at each system’s accuracy for ACC conference games (IC)

Our final game holds the distinction of being the only game between two conference opponents for this round. The two ACC foes only met once during the season, however, so we don’t have much history to go on. Duke was able to get past Syracuse by a score of 60-44 at Cameron Indoor, though Syracuse has succeeded thus far in the tournament by holding teams to 60 points or less. It will be interesting to see if the Orange’s magic continues or if it will end in a double digit victory like the systems are predicting, despite the drop in confidence for in conference predictions by the models.

Final Four

The biggest wild card region remains the South, based upon the lower accuracies for the remaining teams’ out of conference games by the three models. Villanova is heavily favored by KenPom to come out of the East, but Moore and Sagarin think a potential matchup between the Wildcats and Boilermakers of Purdue is more of a toss-up. The West regional has similar matchups statistically, but two less predictable conferences involved with the SEC and WCC, making it tough to predict. In the Midwest, Duke is a clear favorite over Syracuse, but the Orange’s 2-3 zone could be a confounding variable the models struggle to incorporate into their ratings and the in conference factor adds more uncertainty to the matchup of Hall of Fame coaches. If the higher seeds in both matchups advance, each system would favor Duke over one seed Kansas, though any home court advantage for the Jayhawks turns it into a toss-up.

Which Projections Are Best?

The answer, it turns out, is that the “best” projection varies from game-to-game. That’s not entirely unexpected, but now we have a toolkit for understanding how we can take advantage of those variations and pick the best brackets.

Dan Loman

(AP) Shelvin Mack hit 7 of 12 3-pointers to help Butler upset Pittsburgh in 2011

The Basketball Gods giveth, and they taketh away.

Here at Bracket Voodoo, we do our best to give you the crucial information you need to win your March Madness pool. From our state-of-the-art Bracket Optimizer to our Power Rankings and Game Projections, we’ve crunched the important numbers from a wild year of college basketball to help you maximize your chance or earning that sweet office pool cash and family bragging rights.

But, as we find every year, even the best statistical models fall victim to the sheer randomness of the tourney. The crazy, unpredictable upsets are what make March Madness so special and, well, maddening. Yes, in the one-and-done format of the NCAA Tournament, where one off night sends you packing, luck undeniably plays a major role.

When we think of the concept of luck in basketball, we think of old adages like “the ball didn’t bounce our way” or “shots just weren’t falling today”. In basketball analytics, luck is more clearly defined and heavily researched. Nylon Calculus has done a lot of work analyzing luck (in the form of 3 point defense) and even using it to adjust team and player ratings. When I say luck here, I’m referring specifically to shooting performance - teams and opponents shooting 3 pointers and free throws at a different rate than they did during the season. We call it luck because it’s largely dependent on small sample size variation and factors outside of team control.

So to analyze the impact of luck in the NCAA Tournament, I created a “Luck Factor”, which is the difference between the actual points a team and their opponent earned through 3 pointers and free throws versus what we’d expect them to earn based on players’ percentages during the regular season (minimum of 40 attempts). I compared this Luck Factor to a “Degree of Upset”, which ranges from -1 (huge favorite that lost) to 0 (expected results) to 1 (huge underdog that won). As you can imagine, shooting luck correlates very strongly with unexpected tournament performance and explains over 17% of the variation in tournament upsets over the past decade:

It’s no surprise that teams rely from a little bit (or a lot) of luck to outperform expectations in March. Final Four teams benefited from an average of nearly 3 points per game onto their point differential from shooting luck, and that number increases to 4.6 points per game for NCAA Champions. And of the biggest tournament upsets over the past 10 years, all but one had some shooting luck involved and three got a ton of help:

Was Georgetown prone to choking in the tournament? Or were they just victims of bad luck? While the former makes for a more interesting narrative, we’ve found that because tournament performance doesn’t correlate year over year, the latter is more likely.

Can we predict which teams will get lucky? Well, not very well, hence why we call it luck. However, we found that below average 3 point shooting teams actually tend to experience greater shooting luck in the tournament than their sharpshooting counterparts. In aggregate, 3 point percentage in the NCAA Tournament is lower than in the regular season, with a difference of around 1.5%. That bodes well for teams that didn’t rely on elite shooting to get them to the tourney because they can win in other ways, so if they get hot from beyond the arc, they’re even more dangerous.

So what can you do to counter the randomness of March? For starters, acknowledge that it exists. If you’re in multiple pools, or can afford multiple entries, diversify your brackets to expose yourself to a variety of different outcomes. Don’t go trying to predict luck in pursuit of the perfect bracket - you will likely fail. Making safe picks with the acknowledgement that some of them will lose instead of going crazy with first round upsets is the best way to keep yourself alive after the first weekend (but you still need to be contrarian later on to set yourself apart from the field). Finally, you can try and please the Basketball Gods. Burning a jersey in sacrifice or getting a premature tattoo celebrating your favorite team’s title run is always good karma.

Dan Loman

(USA Today) Moritz Wagner and Michigan shouldn’t have to worry about the long layoff affecting their title aspirations

Entering my mid to late 20s, I’m finally beginning to understand some of the old man rants of my father growing up. As the college basketball landscape continues to change, my friends and I have been catching each other complaining about how things used to be back in OUR day. Back when the Pac 12 was the Pac 10. When the A10 and Mountain West were legitimately good mid major conferences that garnered multiple at-large bids. When conference alignment actually made geographic sense! Just as my dad would confuse Texas and Oklahoma of belonging to the old Southwest Conference, I’ll never keep it straight that Maryland isn’t in the ACC anymore.

Back in our day, the Big East absolutely ruled college basketball. It wouldn’t be uncommon to have Syracuse, UConn, Georgetown, Villanova, Pittsburgh, Louisville, Marquette, West Virginia and Notre Dame all ranked or close to it, and when they all squared off in their conference tournament it was a major event. The Big East Tournament was always held in MSG, while the ACC tournament was played in North Carolina and the Big 10 tournament was somewhere in the midwest - Indianapolis or Chicago. Nowadays all the big conferences seem to want to hold their tournaments in New York, which makes the logistics difficult.

As a result, the Big 10 was forced to hold their conference tournament a week early this year, leading to a longer than usual layoff for their teams before the NCAA tournament. Much has been speculated on how the layoff could affect Michigan, Michigan State, Purdue and Ohio State next week - will they be more refreshed? Or rusty?

The bad news is that layoffs this long leading up to a team’s first NCAA Tournament game (up to 14 days if Ohio State lands a Friday game) are virtually unprecedented for major conferences. The closest comparison is Villanova in 2011, who lost in the opening round of the Big East tournament on Tuesday, 03/08 (the Big East Tournament lasted FIVE days!), and then played their opening NCAA Tournament game on Friday, 03/18 ten days later. The Wildcats lost in the first round to George Mason, but as a 9 seed didn’t have very high expectations in the first place.

The good news is that a 10+ day layoff is common among non-power conferences, and gives us a large enough sample for comparison. In aggregate, days off before the NCAA Tournament has no correlation with tourney success over the past 10 years:

*Expected Wins calculated using team and opponent quality

Seven teams have gone 13 days before playing in their first NCAA Tournament, of which three (Cornell in 2010, Florida Gulf Coast in 2013 and Wichita State in 2015) made Sweet 16 runs by upsetting higher seeded opponents. Some of the most surprising Final Four teams - Bulter in 2010 and 2011, Wichita State in 2013 and VCU in 2011 - all started their NCAA with at least 9 days of rest.

On the flip side, Gonzaga and Wichita State fell in the second round as top seeds in 2013 and 2014, respectively, and plenty of big conference teams - Purdue in 2011, Virginia in 2015 and Pittsburgh in 2011 - disappointed following a week of rest. Yearly popular upset pick Belmont struggled to advance past the first round after entering the tournament following significant time off.

By Dan Loman

If Grayson Allen and Duke does well in the NCAA Tournament, it’ll be because of their talent and not because they peaked at the right time.

Everyone has their own criteria and things that they look for when filling out a bracket. Whether it’s picking the home team, the team with multiple future NBA players, or the team with the coolest mascot (go Norse!), most college basketball fans have something that they lean on when penciling a team to the next round. Here at Bracket Voodoo, our goal is to dispel the myths of bracketology to help you optimize your bracket and increase your odds of cutting down the nets in your office pool. Last year, we focused on recency bias and found that historical team tourney performance doesn’t matter (if you don’t believe me, just ask Jay Wright and Villanova). This year, we looked at another common talking point - are teams with momentum heading into the tournament more likely to succeed?

To answer this question I looked at three different performance metrics using data from the past 5 seasons (2012-2016): Winning the conference tournament, wins over the past 10 games and win percentage in February and March vs win percentage from the beginning of the season. To normalize things a bit, I only considered teams from conferences which got more than one team into the tournament, and used tournament wins over expected as a measure for success (expected tournament wins were calculated as function of team quality and opponent quality). Ultimately, I found that momentum heading into the tournament has virtually no correlation with tournament success.

Winning the Conference Tournament:

One of the major things people look at when filling out their bracket is a team’s success in their conference tournament. Surely a team that won their conference tournament is riding high and has proven capable of winning in March, so they’d be a good bet to advance far in the NCAA Tournament. Recent history would suggest otherwise. As the chart below shows, getting into the tournament as an automatic bid vs an at large bid makes virtually no difference when it comes to tournament success:

The seven biggest overachievers of the past 5 years, a group that includes UConn and Kentucky in 2014, Wichita St. in 2013 and Syracuse last year, did not win their conference tournaments. If we went back to 2011, the First Four to Final Four VCU team didn’t, either. In fact, on average, teams that lost in their conference tournaments have fared slightly better than teams that won it: conference tourney losers have won .02 games over expected compared to .07 games under expected for conference tourney champs. In aggregate though, the difference is slight and statistically insignificant, so it’s best to disregard conference champion status when filling out your bracket.

What this really says is that the selection committee does a good job of weighing the value of a conference tournament victory when seeding teams. Winning their conference tournament will give teams a boost in their seeding and overall rating, since they likely picked up good wins along the way. That increases expectations - by my calculations, conference tourney winners were expected to win roughly .4 games more in the NCAA Tournament than conference tourney losers (since I excluded teams from one bid leagues) over the past 5 years. So teams that got in as automatic bids tend to meet those higher expectations at about the same rate as at-large teams. 

Last 10 games:

There’s also virtually no correlation between a team’s record in its previous 10 games and its success in the NCAA Tournament. As the chart below shows, teams have performed similarly relative to expectations no matter how hot they were over the 10 games leading up to the tournament:

Interestingly, of the six teams that went 10-0 over their previous 10 games before the NCAA Tournament, only one (Florida in 2014) outperformed expectations. The group with a losing record in their previous 10 games also underperformed on average, though Syracuse went just 4-6 before the 2016 NCAA Tournament and managed to shock the world by reaching the Final Four. Overall, however, the difference among groups is small and statistically insignificant, suggesting that recent performance over a 10 game period is not predictive of NCAA Tournament success.

End of Season vs Beginning of Season:

They say that all games are created equal, and that games in November count just as much as the games in March when the committee decides who goes dancing. Turns out, the same more or less holds true for winning in the Big Dance - as the chart below shows, teams that do better late in the season tend to perform about the same as teams that appear to have peaked early.

You may remember in 2013, when a Trey Burke-led Michigan team won 95% of their games before February and was ranked #1 overall at one point in the season. They stumbled into the tournament winning just 50% of their games since, but ended up reaching the finals before losing to Louisville. On the flip side, Virginia won 92% of their games after February 1st in 2014 (compared to 76% before), and failed to meet expectations as a 1 seed. Overall, recent NCAA Tournament history suggests that a team’s overall body of work matters more than their recent performance.

To summarize, we recommend not getting caught up in recent performance when filling out your bracket this year. Don’t overvalue teams because they won their conference tournaments, or sleep on them because they struggled a bit down the gate. You’re better off looking at a team’s overall rating (like our power rankings here) and overall body of work throughout the year. And if you get stuck, check out our bracket optimizer for help with those tricky picks.