Tuesday, August 2, 2011

Bests, Worsts of Last 15 Years & What It Means For Future

The 2010 Cornell team that buried three after three en route to the Sweet 16. The 1998 Princeton squad that snagged the league's highest seed (5) in the 64-team era. Or how about the 2003 Penn group led by the dynamic Ugonna Onyekwe?

Which team was truly the Ivy's best over the past 15 years? Which squads had the most potent offensive attacks and played the stingiest defense? Which were the years were the best for the league, and which were the worst? And most importantly, what can the past tell us about the future?


Ask anyone how they'd define the best Ivy team over a certain period of time, and you'll get any number of different answers.

Many will point to tournament success, if any. Others will cite an undefeated league season. Still others will go a layer deeper, making an argument from the individual player level.

Those are fine arguments to make but are not particularly useful as uniform metrics of comparison over a vast number of years. Tournament success is one of the shallowest sample sizes, but most public, you can look at, making it difficult to judge the true measure of a team. Undefeated Ivy runs help from the sample size perspective, but fail to account for league quality, which leaves Penn's 2003 14-win campaign equal to its one three years earlier, despite the fact that the former came against the second toughest league in the last 15 years while the latter came against the 13th.

While arguing from the individual player perspective does add some important color, it is in the roll up of those individual players that we find the best basis for argument. That aggregation leads us to the most objective and consistent metric for comparison of teams across years: Adjusted Pythagorean Win Percentage.

The theory behind Adj Pythag is similar to baseball's version. Take a team's total points scored and allowed. Divide by the number of possessions and multiply by 100 (to get the Offensive and Defensive Ratings per 100 possessions). Then, adjust those two metrics based on the opponents' strength of schedule (basically, your opponents Offensive and Defensive Ratings) to get your Adjusted Offensive and Defensive Ratings, which will now be comparable across all teams and years. To get around the Bayes calculation of Adj Pythag, we'll use the Bill James shortcut approach with the accepted college basketball exponents of 12.5.

Top 20 Teams of the Past 15 Years

Princeton's amazing 1998 run leads the way by a significant margin, as the Tigers made a ton of noise in the non-conference slate, knocking off Texas, NC State, Rutgers and Wake Forest (and falling to future Final Four team UNC by eight) before sprinting through a pretty terrible Ivy League.

A 0.930 Pythagorean rating usually equates to a No. 15-20 spot in the Pomeroy Ratings, a significant achievement for a low major conference (such recent teams have included Davidson 2008 and Belmont 2011).

The most recent Ivy Cinderella story, Cornell's 2010 squad, checks in at No. 2. The Big Red had its own fair share of big wins (Alabama, St. John's, Temple and Wisconsin, as well as a five-point loss to then-No. 1 Kansas), but its lopsided losses against Syracuse and Kentucky, as well as its mystifying Ivy defeat at Penn created a drag on Cornell's Pythag rating.

The rest of the Top 10 includes Princeton's other notable teams of the late 90s, including the 1997 Tigers team that fell three points shy of giving Princeton three-straight first round wins, and the 1999 squad that remains the best team in the past 15 years not to win a title. Joining those Tigers squads are a host of Penn teams that all fell in the 50-80 range of the Pomeroy Ratings, but were among those  that went 0-for-7 in first round NCAA games over a span of nine years.

Some notable teams outside the Top 10 include the 2010 Princeton and Harvard squads, which mark the fourth and fifth teams among the Top 20 not to win an Ivy title. (The 2010 version of the Crimson was the only third place finisher in these rankings).  The remainder were Ivy champs that ranged from 80-130 in the Pomeroy Ratings, while making intermittent noise on a national level.

On a macro level, it would seem that the increased publicity the Ivy League has received over the past couple years is off base. Six of the Top 12 teams came from the 1997-2000 era, while just one has cracked that list in recent years. For all of the hubbub surrounding the 2011 Harvard and Princeton teams, those two are looking up at 12 other squads from the past 15 years.

With a little more information, though, we'll see exactly why it's still fair to say that the last two years have been the Ivy League's best run in this span.


Worst 20 Teams of the Past 15 Years

The pattern should instantly jump off the page.

A whopping 10 of the 20 worst teams in the last 15 years came from the same span that produced six of the Top 12. While that should take nothing away from those great teams from 1997-2000, it does tarnish the reputation of the league as a whole during those years. A Pythag Win% below .100 places a team, in today's terms, firmly outside the Top 300.

It should also come as no surprise that eight of the bottom 20 were various editions of Dartmouth basketball. Four more Big Green squads are among the next 10 worst, as just the 1997 (23rd best), 2003 (78th) and 1999 (79th) teams managed to escape the bottom quartile.

Columbia's 2003 team that got Armond Hill fired is still the gold standard for futility. Its 79.3 Adjusted Offensive Rating is the worst in the past 15 years, as is its 27.9% TORate. Adj Pythag indicates, however, that the Lions weren't last by as far as is commonly conceived. In fact, an extra point per 100 possessions (which still would have kept the Lions dead last) offensively was all it would have taken to pawn off cellar ownership to the 2004 Big Green.

As one might expect, Penn and Princeton are the only two teams to avoid the bottom 20 entirely. Penn's worst season was its final edition of the Glen Miller show in 2010 (94th), while Princeton's low point came in 2008 (93rd) - the first year of the Sydney Johnson rebuilding project. Harvard is the only other squad to appear just once and joins the Quakers and Tigers as the only programs to have just one representative in the bottom quartile.


Average Program Ratings, Past 15 Years
Quick: Name the best program of the past 15 years.

The answer will depend on who you ask, but should - for any rational person - be one of two choices: Penn or Princeton.

Adj Pythag does very little to solve that debate. Penn has been the far superior program over the past 10 years, despite its recent tumble, while the addition of the late 1990s to the fray helps Princeton make up ground in a hurry.

The result is a statistical dead heat. Both the Quakers and Tigers have been better than national average programs over the past 15 years, tremendous accomplishments for teams from low-major leagues. While Princeton has had the best season on record, Penn has had more of the great seasons, including seven appearances in the Top 12. The Tigers have the tourney win to fall back on (1998 vs. UNLV), but the Quakers have the superior league winning percentage (.729 vs .700) over that span.

Beyond Penn and Princeton, there is the expected mishmash. Harvard edges Cornell for third based primarily on the quality of the Crimson program in the late 1990s and early 2000s. Of the 11 Frank Sullivan-coached teams in the sample, six were in the top half of league teams over the past 15 years and two more (2006 and 1998) were the first two squads on the other side of the midway point.

Yale slots in just behind the Crimson and Big Red, hampered by the pre-Jones era. The same is true for Brown, though to an even greater extent, as the Bears had three late 1990s teams among the worst seven to take the court as Ivy representatives.

Another decent-sized step behind brings us to Columbia - the only program not to have any of its teams rank among the Top 40 of the past 15 years. Finally, we arrive at Dartmouth. The Big Green's magical 1997 run, which included two- and five-point losses to juggernaut Princeton, is the only Dartmouth representative in the Top 75 of the 120 seasons in the sample. For reference, the Big Green's average Adj Pythag over the span of years in the sample would have placed it 306th nationally last year. Again, that's an average.

If you think that Penn and Princeton have a sizeable advantage now, you should see the first 10 years of the sample. Looking at just 1997-2006, the gap between the Tigers (.679) and Quakers (.674) and Harvard (.261) and Yale (.257), the third- and fourth-place teams over those years, respectively, was 66 percent larger than it is today. While the last five years have still been kind to Penn and Princeton, with each picking up a title, they've also been kind to others, including three-time champion Big Red and the Ivy title newbie Crimson, in a way that they hadn't been in the past.


Five teams over the past 15 years have elicited our awe and boredom, as they sucked the drama out of the Ivy race by marching to an undefeated league campaign.

The list includes teams you immediately think of when pondering Ivy dominance. There were the 1997 and 1998 Princeton teams that gave No. 5 seed California and No. 4 seed Michigan State scares as well as notching a tourney win over UNLV. There was the 2003 Penn team that was one of the most balanced of the AI era, but fell flat in the tourney against Oklahoma St.

But also among that group of five was a relatively forgettable 2000 Penn team that hung around with Illinois as a 13 seed and the 2008 Cornell squad, which was one of the worst Ivy champions of the past 15 years.

The chart above takes the Average Adj Pythag for all the league teams in any given year and then ranks each of the 15 years against each other. This adds some useful color to the debate. Three of the five undefeated campaigns came against leagues which rank in the bottom third (1998, 2000 and 2008). While the 1998 Princeton team would have had a great shot snapping up all 14 wins any year, the 2000 Penn squad probably benefited from the weaker competition and the 2008 Cornell group would struggle to repeat their undefeated run against any league.

The 2003 Penn team clearly deserves a massive amount of respect. It managed to do against the second-toughest competition what the great 2010 Cornell team couldn't do against the third - survive the Ivy slate unscathed.

The 2002 to 2005 Quakers run of three titles in four years becomes much more astounding when you consider the quality of those leagues over that span. While 2010 and 2011 have a jump start on snatching away the title, the 2002-2005 seasons are the gold standard over the past 15 years and potentially of the AI era, as the league was much tougher top-to-bottom than it had been in a long time.


Much like the debate over the top team from the past 15 years, the best offensive shapes up quite the same way: Cornell 2010, Princeton 2008 and a lot of also-rans.

Cornell gets the slight edge here, but the margin is essentially insignificant. The difference between those two squads and the rest of the list is what is most impressive, as 1998 Princeton was as far ahead of 2003 Penn as the Quakers were ahead of No. 19 on the list (2008 Brown).

Every Ivy team among the Top 20 best offenses of the past 15 years won at least 10 games during league play. That statistic will become more important as we flip to the defensive side of the court.

Very few of the worst teams should come as a surprise to anyone, least of which the "leader," 2003 Columbia. As we'd expect, the 1997 to 2000 era is the biggest offender here with half the teams on the list and the three worst teams over the past 15 years are responsible for eight of the 10 spots.


At the very top, the weighted defensive rating rankings dovetail nicely with the common perception of the league's best teams over the past 15 years. Even that 1997 Dartmouth team, despite being particularly punchless on the offensive end, came incredibly close in the Ivy race against a far superior Princeton squad.

The list begins to fall apart outside the Top 10, though. Included in the 11-20 range are two teams which won just seven (Harvard 2002) and eight (Yale 2003) Ivy games and two squads that had Adj Pythag rankings which were lower than .500 (Harvard 2002 and Harvard 1997).

Having some of the worst Ivy defensive seasons weren't as penal as their terrible offensive counterparts (bottom five defensive teams pulled down 12 total wins versus eight for the worst five offensive squads), but the bottom 10 and 20 on each list had, in general, a similar lack of success (3.7 league wins for the worst 20 defensive teams and 3.6 wins for the worst 20 offensive teams).

Some teams managed to overcome their sieve-like defense, as four of the 10 teams on the list avoided double-digit losses, including 1998 Harvard, 2007 Harvard, 2010 Penn and 2010 Brown.


Top Adj Pythag Differences Year-Over-Year

The dramatic turnaround that we all dream about for our teams is, indeed, possible. But likely? That's a different story.

Only 10 of the 112 teams (9%) in the last 15 years have mustered a Adj Pythag increase of more than .200 on a year-over-year basis. The nine percent likelihood is dependent on another very important variable, however.

Of the Top 20 biggest movers on the chart to the left, only two of the prior year teams posted an Adj Pythag of over .400 (No. 4, 1999 Penn and No. 17, 2010 Cornell). The remaining 18 teams showed vast improvement over squads that were pretty terrible (almost half had an Adj Pythag of under .200 in the previous season).

Thus, while significant improvement is somewhat possible, the likelihood of it occurring is seemingly inversely proportional to the quality of the prior year team, something which meshes with the simple statistical tenet of regression to the mean.


The Adj Pythag year-over-year history is ominous for most of the 2011-2012 Ivy title chasers. Assuming that Harvard at least maintains its 2010-2011 production, the only team within .299 of the Crimson last season was Princeton, which loses two All-Ivy caliber players.

While a 14-game sample size is small enough to allow a team with an inferior Adj Pythag to claim the title over a squad with a superior rating, such an upset has only occurred once (2004) when the gap between Adj Pythag ratings was greater than a razor thin .020 margin. Thus, for the other six to grab the title over Harvard, they'd probably have to make up most of that .299 or greater margin, which would entail an extremely rare level of improvement for an Ivy team.

Using the Adj Pythag change data, we can create a thumbnail approximation of where teams will end up in 2011-2012 based on last season's finish. While this approach is far less scientific than the previous 2011-2012 projections posted on this blog, it can be used as a sort of reality check to ensure that the results of that analysis were reasonable.

Let's walk through the reasoning for each of the adjustments by team.

Harvard: Of the 13 teams in the sample to have an Adj Pythag of over .700, just three advanced the next season and all three were gains of under .010. Sitting just outside this subset is 2009 Cornell's team (.696), which advanced by 0.172, so a large climb is technically possible. Harvard does match up well with the teams that climbed further in the following year, but convention wisdom indicates that gains should be expected to be modest.

Princeton: Many folks around the league expect the Tigers to fall on hard times after losing All-Ivy guard Dan Mavraides and POY candidate Kareem Maddox. While it's hard to disagree directionally, the expected magnitude of the fall might be a bit overstated. Only eight percent of Ivy teams have experienced a year-over-year drop in Adj Pythag of more than .150 and many of those were teams that lost a majority of their starting lineups (2004 Harvard, 2006 Princeton, 2008 Penn and 2011 Cornell). The Tigers lost a lot, but still have remaining pieces beyond what those teams had, moving them up to the fringe of the curve, rather than placing them among the relative outliers.

Yale: With Greg Mangano on the verge of making the World University Games team (he's in the final 14, from which just two more will be cut), only one decent graduation loss and a good freshman class coming in, the Bulldogs have the look and feel of a team that is primed for a breakout season. An outlier-type increase might not be supportable, but Yale should push the bounds of what is considered within the realm of normal, which could put them all the way to the fringe of the Top 100.

Penn and Cornell: For teams hovering around .400, the average historical change the following year is basically zero. Each of the Quakers and the Big Red have the same reason to be optimistic (good guard play) and the same reason to be incredibly nervous (gaping holes in the post) about the upcoming season. Given that it's not obvious which way either team will move we'll put both teams in a holding pattern until there's evidence to move them one way or the other.

Columbia and Brown: You might suspect that the given bump of .020 is mere regression to the mean. On a school specific basis, however, regression to the mean would actually cut the other direction for each. Rather, given that Columbia didn't lose much of anything other than rebounders who were atrocious offensively and given that Brown has put together three straight solid recruiting classes, both teams could be expected to keep climbing, though modestly.

Dartmouth: Regression to the mean and a healthy dose of freshman talent is the recipe for an expected rise by the Big Green during the upcoming season. Dartmouth hasn't had an average Division I player since star swingman Alex Barnett left after the 2009 season, but the trio of David Rufful, Jabari Trotter and R.J. Griffin might at least be Ivy replacement level. The team still has no size up front (though it will have a bunch of combo forwards to slot in from its freshman class), but it is at least bringing in average Ivy recruiting classes again, which should be enough to bump it up the charts a bit.

The thumbnail approach seems to justify most of the projections in the initial draft, but suggests that Yale might have some more room to rise than previously anticipated (139th in the initial vs. 112th by the thumbnail approach). At the same time, the expected gains for Brown (217th inital vs. 239th thumbnail) and Dartmouth (294th initial vs. 308th thumbnail) might have been a little too robust.

Having the historical information, however, can go a long way toward helping create bounds for expectations based on previous results and can provide a strong base from which to make educated guesses about future projections.

No comments:

Post a Comment