Everyone knows that you can’t fill out a bracket without picking at least one 12 seed to upset a 5 in the first round. While this well-known phenomenon has been backed up with data showing that it truly does happen more than you would expect, there has been almost no study into why exactly it happens. That is, until now.

The truth of the matter is the 12-5 upset isn’t a thing. Well, it is, but it only really applies to certain teams. Looking back over many years of NCAA tournament history, when you separate out power conference and small conference teams, it is only the 12 seeds from the big conferences that outperform what you’d expect from a 12 seed, while the small schools perform at a rate almost that is, well, exactly what’d you expect.

Looking all the way back to 1985, when the field expanded to the 64-team format (play-in games are not included in this analysis), power conference (ACC, Big 10, Big 12, Big East, SEC, Pac 12) 12 seeds have taken down the 5 seed 55% of the time, compared to only 32% for small conference (all other than the 6 conferences mentioned above) schools. This is not just an old phenomenon either, in fact quite the opposite, as the big schools have won 77% of the time over the past 20 years compared to 33% for the small schools. When looking at the past 10 years, it climbs even higher to 88% compared to 34%, and over the past 5 years all 5 power conference schools to be a 12 seed have won, while just 35% of the small conference schools have. Considering 11 seeds win at a 36% clip historically, it is clear the small schools are performing about how you’d expect from a 12 seed, while the big schools are the ones creating the mirage of all 12 seeds over performing.

Now, the natural reaction to this information is that power conference teams are always going to outperform the small schools, and it is not just the 12 seeds. In reality, however, that is not the case. Over the past 10 years, 11 seeds from the small conferences have a better first round winning percentage than their power conference counterparts. Same with 10 seeds, and 9 seeds, and 8 seeds. While there is nowhere near as big of a separation as there is with the 12 seeds, it shows that it is not just a case where big schools outperform their seeds. In fact, comparing performance across seeds 1 to 11 over any recent timespan, there is no notable difference in performance in the tournament between the big and small schools. While it is certainly clear that big schools overall have more success, they also tend to be seeded higher, and once you hold seed constant, the performance is pretty consistent across both groups, suggesting the selection committee has overall fairly seeded both big and small schools.

Also, it is worth wondering whether this is just a product of chance, given the fact that there haven’t been that many 12 seeds from the big schools. However, assuming 12 seeds truly have a 32% chance of winning, the odds that the 55% figure would occur is just 2% when factoring the number of games played. And if you’re more inclined to look at just the past 5 years rather than the full 33-year history, the odds we’d get these results from pure chance are less than half a percent.

Considering this information, it begs the question, what about these teams make them so likely to provide these upsets? It is a bit hard to say, but looking at the teams from recent history provides a bit of a glimpse. The general theme is these teams fall into one of two buckets: teams that got hot in the conference tournament and pulled out an automatic bid to get into the field when they otherwise would have been on the outside looking in, and rode that momentum into the tournament, while the other group is teams that crashed out early in their conference tournament, likely causing them to be under-seeded from one bad performance when their resume from the whole season merited a higher seeding. There was not a single power conference 12 seed to pull an upset who went out in either the semis or championship of their conference tournament in the past 10 years – every single one of them either went out before that point or won it all.

Looking at this year’s bracket, there is not a single 12 seed from the power conferences, but this does not necessarily spell doom for Davidson and their fellow 12 seeds. As mentioned above, small conference 12 seeds have still taken about one third of their matchups, so look for 1 or 2 of them to pull off the upset. After all, anything can happen in March.

The data for this article was supplied by The Washington Post at: http://apps.washingtonpost.com/sports/apps/live-updating-mens-ncaa-basketball-bracket/search/

The post The Truth Behind the 12-5 Upset first appeared on 'CATS STATS.

In the NBA, with the wealth of data available, many factors can be considered, like distance of nearest defender. However, that luxury is not as widely available in the NCAA. Thus, the best way we can look at shot selection is by looking at where shots come from on the court. With the help of data from CBS Sports’ website, we compiled a database of over 1100 games and 120,000 shots ranging from the beginning of the ‘14-15 NCAA season to the end of the 2015 calendar year, which amounts to the approximate end of non-conference play from this season.

By looking at every shot and the result of it, we can assign an expected points value for each coordinate on the court, and use that to evaluate shot selection across college basketball. Above, you can see a diagram of shot efficiency from different areas on the court, where the red areas are the most efficient and blue the least efficient, while the size of the square represents the frequency of shots coming from that location. One thing to note is that CBS groups all layups to one coordinate (the big red square by the hoop), which is why it may seem as though there is too much blue around the hoop. However, the efficiency of layups is indeed captured.

Once an expected points value is available for every coordinate, we can then go through player by player and find the expected points of every shot they have taken, and then average these values to find what kinds of shots they are taking as a whole. In order to avoid outliers, it is limited to players with over 200 shots.

Looking at the top of this category, there tends to be a wide range of type of player. With the 3 point revolution currently going on in the NBA (and all across the game of basketball), spurred on by the growth of analytics, you would expect that this list would be filled with 3 point specialists. Yet, judging from the numbers that is not the case. While there are guys on the list who have range out to beyond the arc, there seems to be more of a presence from bigs who get most of their shots right around the restricted area. Overall, it seems to be generally a combination of the aforementioned bigs along with small forwards who like to slash to the hoop a fair amount making up most of the leaderboard.

What does this tell us? Well, to a degree Marc Jackson had a point in his criticism of Stephen Curry and his effect on the game. The one in a million talents like Curry may have the ability to take 3’s and make them consistently at an extremely efficient clip. However, in the college game at least, the players who are more aggressive at trying to get efficient looks at the hoop are actually giving themselves the best chance to score. The 3 is certainly a more efficient shot than the midrange jumper, but people are beginning to forget the fact that even the 3 will never match the efficiency of a layup.

As for applying this metric to full teams, some of the notable programs near the top do tend to be teams with a heavy emphasis on 3-point shooting, like Duke and Villanova. On the other end of the spectrum, teams near the bottom like Pitt and Illinois generally tend to have more of an emphasis on post play and the midrange game.

So, what is the difference between individuals and teams? For one, though certain players at certain positions may be able to get almost all of their shots at the rim, it is almost impossible for a whole team to do so consistently. Thus, when faced with having to find a secondary option for scoring, the 3 will almost always prove the wiser option than the post or midrange game.

While shot selection is certainly important, it is also interesting to look at the players who outperform their expected points average the most, which is essentially the players best able to make shots most other players can’t make. Interestingly enough, the top player in this category also happened to be the #3 pick in this year’s NBA Draft, Jahlil Okafor. Some other notables follow closely behind in the top ten, including All-America candidates Brice Johnson, Jakob Poeltl, Kyle Wiltjer, and another top ten NBA pick in Frank Kaminsky. One other thing that stands out is that three of the top eight all happen to be from Gonzaga, which makes it seem as though offensive system may have some sort of effect on this category.

Not surprisingly, when looking at teams as a whole, Gonzaga is right near the top in outperforming their expected points, behind only Northern Iowa. As expected, this is the category where there tends to be the strongest divide between the perennial powerhouses of college basketball and the smaller programs, which makes sense because of their ability to bring in the top recruits year after year.

When looking at expected points and actual scoring together, it becomes clear why shot selection is so greatly stressed throughout the game. There is a strong correlation between teams that take high percentage looks and teams that convert at an efficient clip, just as one would expect. While some teams like UNC have the ability to take tough looks and still score at a decent efficiency, the fact remains that teams that take better shots tend to be more successful. There will always be exceptions to the rule (see Curry, Stephen), but it does not change the fact that the importance of working to get high percentage looks cannot be understated.

This whole concept holds even more importance on the defensive end. Unlike on offense, the defense has no control over whether the opponent makes a shot, rather they just have the ability to try and make it as tough of a shot as possible. Thus, looking at teams that are able to force tough shots for opponents is a great way to measure the effectiveness of a team’s defensive performance. Without player tracking, we cannot look at how close the defenders were, but given what is available, seeing from where shots come can certainly shed some light on the premier defensive teams in college basketball.

Looking at the data, teams like Purdue, who pride themselves on being smart players who leave their heart out on the court every night, tend to be at the top, while underperforming programs like St. John’s find themselves struggling to force tough shots for their opponents. While it certainly doesn’t hurt to also have two 7-foot rim protectors like AJ Hammons and Isaac Haas, as athleticism certainly plays a significant role on defense, you can see in the data that it takes more than one or two guys with freakish athleticism to make it tough for opponents on the defensive end. Looking at some of the teams near the top, it becomes clear just how big of a role effort and grit play in whether or not teams are able to get stops on the defensive end.

While on the offensive end there are numerous different stats to evaluate teams, there has never been much to evaluate defense objectively, other than simple stats like points allowed. However, in no way does that account for whether an opponent just happens to have either a good or a bad night. Looking at shot selection allowed gives a whole new perspective on how to evaluate team defensive performance.

By no means does shot selection tell the whole story when referring to either offensive or defensive performance. However, given the current state of data available in NCAA sports, it adds a whole new dimension to the traditional box score.

The post Shot Selection in College Basketball first appeared on 'CATS STATS.