Thursday, April 21, 2011

Explaining Vanderbilt's Tournament Conundrum: A Statistical Perspective


If you don't particularly like statistics and think that they aren't relevant to how a team performs, this probably isn't for you. I acknowledge that basketball has certain intangibles and a great deal of luck that make it an impossible game to predict fully. However, with all of that said, I find arguments not based on hard data to be personally unsatisfactory. Statements like "Vanderbilt will lose in the first-round again this year because they've lost in the first-round the previous two years, book it" and "Don't get too excited if you're a Vanderbilt fan, most of their guys have already made their big leaps, they won't get much better" (courtesy of Dan Wolken, a VANDERBILT alum) are based on nothing but conjecture. They could just as easily go the other way as they could that way. My desire is to find the patterns that made those ridiculous speculative statements, which the declarant's no doubt chuckled about with their "friends" while sipping Appletini's and ensuring that their hair was "just right" before making their way back out to the dance floor for another rendition of "him? me? me? him?", sadly, quite factual. Ok, I admit that not everyone who picked against Vanderbilt spends their weekends that way (President Obama....or is he?), but my baseless speculation has about as much merit as theirs that Vanderbilt will be a first-round knock-off from here to eternity, and besides, it's my blog and I'll say whatever I darn-well please!

Also this is really long so if you can only tolerate individual websites in 1 minute increments you might want to stop reading now.

How I got started working on this

A while back I promised Bobby that I'd write a post comparing last year's Vanderbilt team to our sweet 16 teams of yore, as well as the more recent flame-out variety we have become all too familiar with. I went back through the numbers of the previous seasons using Ken Pomeroy's database of his own self-created statistics, which are intended to re-weigh traditional basketball statistics in a manner that more effectively explains a team's identity. Additionally, I went through our tournament box scores, both wins and losses, looking at the score result and trying to identify the statistics for both the Dores and their opponents that led to that result.

Endeavoring to Parallel the '10-11 Commodores with the '03-'04 and '06-'07 Commodores

My first hypothesis was that the '11 Commodores were a better offensive team than any of their predecessors. While it is true that this years team was one of the better offensive groups, it wasn't necessarily the best. Comparing the adjusted offensive efficiencies (points scored per every 100 possessions adjusted for strength of schedule and location of games) and the effective field goal percentages (field goal percentage weighted for increased value of 3-point shots), the '04 Commodores scored about 113.9 points per every 100 possessions and scored on 51.9% of their possessions. The '07 Commodores were at 117.3 per every 100 possessions and scored on 53.1% of their possessions. The '08 team was similar to the '04 team, scoring 113.6 points per every 100 possessions, but they scored on 54.1% of their possessions, the highest of all the teams. The '10 team's stats were 113.7 and 52.6% respectively. The '11 team was our second most efficient squad, scoring 116.5 points per 100 possessions. They scored on 52.8% of their possessions. The point, however, is that they were comparable to the '04 and '07 squads who made the sweet 16, but then again, so were the two first-round flame outs of '08 and '10, both of whom had higher-seeds in the tournament.

My second hypothesis was that the present squad was actually a better defensive team in terms of the quality of opponents shots, but that we were allowing teams to get offensive rebounds and take more shots. This led to greater point totals per 100 possessions for the opposition. The '04 Dores were a very solid defensive squad, allowing 91.3 points per every 100 possessions, and holding opponents to 47.6% from the field. The '07 Dores were worse defensively, allowing 94.0 and 50.6%. The '08 Dores were worse still at 96.4, but they were marginally better in terms of eFG allowed at 49.7%. The '10 Dores were very similar to the '07 Dores at 93.8, but they were actually better than the '04 Dores in terms of eFG% allowed at 47%, which would indicate a defensive improvement.

In '11, our adjusted defensive efficiency spiked to 97.8, meaning they allowed 4 more points per 100 possessions than the previous year, but they were actually the best in terms of defensive eFG% at 45.7%. Additionally, the team was the best by a wide margin in terms of 3-pt. % allowed, as opponents only connected on 30.6% of their threes. No other group was under 33%. How do you have the BEST numbers in terms of field-goal percentages and allow 4 more points per 100 possessions, the most of any Vanderbilt tournament team of the past decade?

As I said, I thought this had to be because of OR% allowed. Well, with the exception of '04 where we were in the top 25% of teams in the category, Vanderbilt has been just about at the D-1 average in terms of offensive rebounds allowed. But the '11 squad, who returned us to the league average, was actually much better than the '10 squad, who might go down as the worst box-out team in Commodore history. In short, while we were still allowing a lot of pure offensive rebounds on strictly the stat sheet, we were actually cancelling them out by getting a lot of defensive rebounds. In other words, we did a pretty good job of ending possessions despite the offensive rebounding numbers. This is where my theory on OR% was somewhat torpedo'ed, but it explained why we could win the rebounding battles in a lot of games and still allow a lot of offensive rebounds. It meant that even though teams were getting second chance looks from offensive rebounds, we were offsetting those with our defensive rebounding. That being said, offensive rebounding was certainly a factor that contributed to our troubles, though it wasn't the sole factor.

So in summary, the '11 Dores were probably the second best offensive team we've had behind the '07 Dores, and they were the best defensive team, except they allowed 4 points more per every 100 possessions, which is a rather large spike. Despite my original thought that OR% allowed was the primary statistic that led to this spike, it turned out that it had actually improved significantly from the previous year, and that it was in line with our successful teams in '04 and '07.

So Why The Struggles of the Past Three Years in the Tournament?

The original purpose of the post was supposed to assuage Vanderbilt fans' concerns about this year by saying the '11 Dores were more like the '04 and '07 Dores and less like the '08 and '10 Dores, who lost in the first round of the tournament, and that we would therefore win against Richmond because of it. As I began to work on the post, I found that premise was ridiculous and would only make me look like an idiot, because it's not called March Madness for nothing.

My over-arching theory was that it wasn't that the '08 and '10 Dores were worse that then '04 and '07 groups, it was that they ended up playing teams who excelled at their weak points. My primary reasoning for this was Murray State. Vanderbilt was one of the worst teams in terms of OR% allowed and Murray State was one of the best in OR%. I also thought that Siena would prove to be an example, as they were one of the best in terms of steal % (steals obtained divided by defensive possessions), and we were really poor in terms of letting teams steal the ball from us (defensive steal %). But when I actually looked at the box scores, it was pretty apparent that those factors were not the turning points in each of those games. In 2008, Vanderbilt had 9 steals to Siena's 10, 16 turnovers to Siena's 17, and easily won offensive rebounding and overall rebounding. The turning point was that Siena hit 45% of their 3's (9-20) to our 20% (4-20). In fact, Vanderbilt attempted 16 more shots than Siena and got drilled by 21. This was an anomaly when comparing the statistical profile of Siena to Vanderbilt's, were Siena was significantly worse in both adjusted offensive efficiency and defensive efficiency. Simply put, the result was somewhat inexplicable. It was literally a worse team getting scorching hot for one game while a solid opposing team went ice-cold. Granted, in '08 the Dores weren't the greatest 3-pt. defenders, allowing opponents to shoot 35% from three, but they were better than Siena in that category and they were one of the top 3-point shooting teams in the country thanks to Shan Foster (as a team, the '08 Dores shot almost 40% from 3, an absurd stat).

Murray State played out a bit more as the Ken Pomeroy profiles would indicate. The '10 Dores, as I said, were one of the worst teams in terms of OR% allowed, and the Racers were one of the best. Vanderbilt was outrebounded 13-7 on the offensive glass and 34-28 overall. They had 10 turnovers to Murray State's 9. Once again, the 'Dores allowed a major 3-point shooting performance to hinder them, as Murray State shot 46.7% from 3 (7-15) against the 'Dores 33.3% (4-12). Murray State also attempted 10 more shots than the 'Dores (60-50), and yet they only won by 1 point: the most gut-wrenching buzzer-beater I've seen in my life. Still, Murray State was known for putting people on the line and we were known for getting to the line (and making our shots), and Vandy only shot 58% from the line on 29 free throws. Of course, I chalked that up to the fact that most of the team aside from Beal, Ogilvy, and Walker had never played in an NCAA tournament before, so I thought this was a reasonable playing out of the scenario.

Naturally, the converse has to be true as well, that the '04 and '07 teams received favorable match-ups against their opponents that led to their success. For the most part, this theory held true, at least as to first-round games, but there was always one anomaly that just seemed to creep out. For instance, a Western Michigan team that was a decent 3-point shooting team going a horrendous 2-19, while the 'Dores shot 11-19 in '04, despite the fact that they destroyed us on the offensive glass and forced 5 more turnovers against a team that was pretty solid at taking care of the ball. Or us outstealing a George Washington team that was 10th in nation in steal percentage, coupled with a hot shooting day for us and a cold shooting day for them.

So when I first found out that we were playing Richmond, I looked at the two profiles and I thought that the match-up was favorable because they weren't particularly good at anything that I considered to be weaknesses of ours at the time. Richmond was one of the worst teams in the nation in OR% and they weren't exceptionally fast (301st in the nation in tempo) or inclined to force turnovers based on steal %. They could score, but not at the same clip as Vandy, and they were a little better defensively, but not much better. My thought was that it was enough to make this into a close game, but not so much that Vandy shouldn't have won it. I was going to report the joyous news that we would likely be victorious, but then I decided not to, thanks to one article I read that made me realize my match-up theory, though not completely incorrect, was not the reason for our woes.

Every year, ESPN has a blog called "Giant-Killers", (insider content) which, through its own metrics, applies an index rating to teams in the top-5 seeds and the 12 or lower seeds to determine whether they lower seeds will be "Giant-Killers" and whether any teams in the top-5 would be "Slain Giants". Since I still haven't seen the math they are talking about to create this rating, I didn't really put that much stock into it when they said Vandy was the most likely 5-seed or below to fall. Then I thought back and realized that they had made a similar statement about our team last year (or that Murray State was most likely a giant-slayer). This made me take a look at the statistics they valued most.

"Giant-Killers" has identified three criteria that most teams who have pulled off 5-seed or below upsets have had:

  • High offensive-rebounding percentage

  • Low turnover rates and high rates of opponents turnovers

  • High 3-point scoring as a proportion of all points scored
Of course, "Giant-Killers" uses these metrics to identify teams that are seeded 12 or below and can pull upsets. All three of Vanderbilt's early exits have come at the hands of teams who exhibit one or more of these factors (Siena had #2, Murray had all three, and Richmond had #2 and #3). But the biggest story is not the existence of all three of these factors in the teams that have knocked us out in the first round the past three years, it's been our fundamental susceptibility to all three of those factors since 2008.

The biggest problem is that we have been one of the worst teams in nation in terms of defensive TO% since 2008. As stated, this is part of CKS' philosophy that it's better to force opponents to take bad shots and then get the rebound than to gamble on trying to get a steal. But even with that philosophy, there's NEVER been a worse CKS team at forcing turnovers than this one right here, which is shocking when you consider the massive improvement in athletic talent. This isn't just about stealing the ball though. Combined with our lack of steals is the fact that our defenses simply don't pressure to the point that they create unforced turnovers. This means that every possession an opponent has they are likely to get some sort of shot, good or bad. Come March, this philosophy is terrible, where lower-seeded teams are going to gamble to try and create more possessions for themselves. The '04 and '07 squads were over twice as good in this category, which still put them just above average.

An additional problem is that Vandy gets swatted, like, a lot. As "Giant-Killers" notes, teams that are in the top-5 seeds typically have much greater talent, and thus their main goal is to "stay the course", score efficiently inside, and not give up a lot of turnovers. Therefore, teams that allow themselves to be blocked at high-rates are susceptible to being knocked off, because they are not scoring efficiently inside. Vanderbilt was blocked 5 times by one guy and 9 times overall against Richmond, which held true to their issues in this category. However, something that goes hand-in-hand with this has been every Vandy fans frustration with our inability to finish around the rim. Without question, Festus was our best "finisher" last season, because most of the time he could just jam it or use his very improved and very useful hook shot. But guys like Taylor, Goulbourne, and Rod just never seemed to hit the point-blank shots that are so crucial to winning games.

Based on what I observed, OR% was actually the third most-indicative factor in looking at our tournament failures. It was not very good down the stretch this year, as I observed in SEC play, but it was not so bad that it couldn't be overcome, and in fact had been overcome by the team during the season, as well as by the previous editions. Most of our tournament success also came while being outrebounded.

Summary: Does this mean we're without hope for the next year?

Absolutely not and whoever is actually getting paid and makes that observation is an idiot. All it means is that, next year, we know what it is we need to look out for, since a team's profile can change drastically from year-to-year. None of the above flaws are permanent by any means. That's not to say that we'll experience a miraculous reversal where we'll suddenly be the best offensive rebounding team in the country, or we'll suddenly be running track meets and leading the country in steals, but it is to say that they are areas that can see vast improvement as the players enter their 4th year of playing with one another. Keep in mind that last year's team did not have the continuity that this year's squad will have. They did not have real game experience with their lineup, and in fact had to change mid-year at the 4 due to injury. They lacked bench depth because of inexperience. Finally, some guys just didn't have as good a year as they would have liked. A year is a long time for things change, for the parts to gel. The 'Dores have an opportunity to hit the ground running for the first time since the Byars year of 2007, so it is perfectly conceivable that, even with minimal-to-moderate improvements in these three areas, they will be less susceptible to upsets, and consequently, more consistent down the stretch. Throw in what should be an improved bench that will take some weight off our big three, and you've got a team who is yet to do it's greatest damage.


Douglas James said...

While I am super excited about next season the one thing about this analysis that worries me is that it would seem those issues come down to coaching and if CKS hasn't focused on those things in the past 10 years why would it be his focus this season. I guess what it comes down to is matchups and if Vandy gets favorable ones (which to be honest is important for everyone).

The positive aspect is the higher seed you are the less likely you are to have an unfavorable matchup (the last few years not being the case). I might be getting alittle ahead of myself but Vandy should be a 3 seed and probably even a 2 seed next year if things fall right. That means godd chance for favorable matchups all the way to the elite 8.

Not sure what my original point is........

Anonymous said...

Could you repeat the whole thing, I missed the middle part.

Stanimal said...

Actually Doug, your statement about Stallings' focus isn't 100% accurate. As his coaching resume shows, this group is his worst at turnover %. In actuality, his previous groups did not purposely veer away from creating turnovers. My point is that this comes from a passive approach to defense that the team has had over the past couple of seasons. I don't think Stallings is particularly pleased with that aspect of the team's performance. It's not that he doesn't want to force opponents to turn the ball over, he just doesn't want to do it at the cost of gambling and getting out of position.

As to offensive rebounding, he tells his team constantly about how important it is to be in position when the ball goes up. That can be seen from his discussion on Vanderbilt All-Access that was on ESPNU.

The bigger concern, IMHO, is whether these protestations will "sink in" finally.

AspenVU said...

That was a great read and analysis - well done!