Let's Face It - We are the Kings of November and December

[freshmanjs]

How can 2011 not be in the category of "excellent point guard play"? By the time the tournament rolled around we had the NBA first pick and a leading NPOY candidate who both played PG. And in 1997, our PG had 5+ apg, a 3 to 1 a/to ratio, shot nearly 40% from three and had 2.5 spg (and would win the NDPOY a year later). That's not "excellent"? As much as people like to rag Greg Paulus around here, he was 1st team all freshman in the ACC and a 2nd team All American freshman in 2006. Those three and 2000 (Jason Williams had 6.5 apg and 2.4 spg, though he did have a lot of turnovers) belong on the "excellent" list at least as much (if not more) than 1999 (Will Avery had 5.0 apg, 1.5 spg, and an a/to ratio less than 2). I think a lot of times how the team finished the season colors our later perception of the excellence (or lack thereof) of its play.

I don't want to get into another knock down/drag out with you, but I don't see how this categorization tells us very much.

my mistake on 2011. absolutely should have been in the other group. just an error. i really do think the pg thing makes a difference though. look at how much better we are this year when quinn cook plays well (and compared to last year).

it's funny how any hypothesis about why duke is underperforming relative to seed must be imaginary or rooted in selection bias, but a hypothesis that tries to show that duke is actually not underperforming is worth testing (despite being at least as far off the mark, as shown by the data).

[Troublemaker]

Here are my thoughts on this:

• The thread title doesn't give us enough credit. We've got to at least be the kings of Jan, Feb and half of March, too. I seem to recall lots of ACC regular season and tournament championships (which Nov/Dec wins don't count towards ) even when we don't advance as far as we'd hope in the NCAAs. Also, as pointed out in the all the "playing to seed" posts, we've received lots of 1 seeds. Nov/Dec DO count towards achieving those, but it'd be quite a feat to receive 1 seeds by playing well for only two months.
• But, yes, the playing to seed thing. We "perform poorly" there. But here's the thing. We rack up so many 1 and 2 seeds that it's statistically likely that we would "perform poorly" at playing to seed. No other program receives those high seeds at the rate that we do (because of our Dec thru mid-March prowess, as mentioned). In the past 15 years, we've been a one or two seed 13 times, including being a one seed 10 times, including a streak of 8 one seeds in 9 years. Incredible. If any other program received one or two seeds at the rate that we do, especially all the one seeds, that program would likely "underperform," too. Why? Because the odds are typically against any one team winning 3 or 4 consecutive games against quality competition.
  Check the Vegas odds after the NCAA tournament brackets are revealed. The 1 seeds are almost always "plus money" to reach the Final Four (meaning they are regarded as less than 50/50 shots to do so). I've seen many a one seed fall into the +150 to +250 range to reach the Final Four (meaning they are given a 29% to 40% shot of doing it [and keep in mind, Vegas has to make money, so they will overestimate the team's chances]). To put it simply, it's much tougher to "play to seed" when you receive mostly ones and some twos.
• NSDukefan stated above that Duke has sometimes been overseeded relative to talent. I agree with this. We tend to overachieve in the regular season relative to our talent. Why? As others have mentioned, Coach K has his teams reach their ceilings much sooner than other coaches. He is the quickest coach to properly tweak his offense and defense to match the talent available in any particular season. Also, I really do believe as a program we play harder game-in and game-out than other programs. We just put so much energy, effort, and passion into each and every game (with some exceptions, of course). In fact, I would accept theories that we could do a better job managing our energy during a season. That maybe we'd be fresher for March if we didn't play so hard every single game. But that's just a theory, and I would think the coaching staff could scale back practices to offset the amount of energy our guys expend each game.
  Getting back to the point. Yes, we peak early, i.e. we reach our ceiling early. But when people say we peak early, almost everyone is imagining a negative parabola, right? You're picturing this in your head: http://sprott.physics.wisc.edu/fract...gs/fig1-03.gif , meaning we reach our ceiling and then we start playing below our ceiling. I, on the other hand, believe our seasons generally look like a fast-rising plateau curve: http://www.devchakraborty.com/images/FROC%20Curve.jpg , meaning we reach our ceiling early and then we MAINTAIN it. But, because other teams are hitting their ceilings later in the season, it seems like, relative to them, that we are slipping from peak when, in fact, we are just maintaining something we had achieved earlier than them. Generally speaking, I would say we plateau more than we parabola, but it always feels like a parabola because everyone else hits their ceiling later.
• If a team parabolas because of injury, I don't find fault with coach/program, and we've had bad luck with injuries recently.

[bedeviled]

You included 2004, which helps Duke's numbers a bit vs. starting with 2005 or 2002.

OMG!! I mislabeled those three charts. The data IS from 2005-2012.Sorry about that. It was late, and I was seduced by the '2004' in the 2004-2005 season. The data, though, is from the last 8 years. Sorry.

[Troublemaker]

One more thing. I agree with Kedsy that, health permitting, we're about to enter another golden age of NCAA tourney success for Duke (although I don't necessarily buy the all-too-neat 3/9 thing).

Our recruiting is in as good a shape as it's been in a long time. The Olympics are finally paying off in that area, and there seems to be more top-level recruits interested in Duke and interested in becoming well-rounded people (my theory on that: the Obama effect).

What I'm saying is, we're going to cease the overachieving/overseeding relative-to-talent phenomenon that we've experienced the past few years. When we get 1 seeds from now on, it'll be because of pure nasty talent, not just from playing hard and peaking early.

In summary: if we get Ryan Kelly back at a reasonable time for him to re-integrate and re-gain conditioning, we will three-peat. I had always thought it was going to be the second championship that was going to be the most difficult, but with all the injuries this season (Seth, Ryan, and now even Mason's left thumb [that's our Big 3!]), man, it's been rough for our guys this season. If we can just be healthy for once going into the tournament...

(Yes, 3-peat!!!).

[freshmanjs]

How can 2011 not be in the category of "excellent point guard play"? By the time the tournament rolled around we had the NBA first pick and a leading NPOY candidate who both played PG. And in 1997, our PG had 5+ apg, a 3 to 1 a/to ratio, shot nearly 40% from three and had 2.5 spg (and would win the NDPOY a year later). That's not "excellent"? As much as people like to rag Greg Paulus around here, he was 1st team all freshman in the ACC and a 2nd team All American freshman in 2006. Those three and 2000 (Jason Williams had 6.5 apg and 2.4 spg, though he did have a lot of turnovers) belong on the "excellent" list at least as much (if not more) than 1999 (Will Avery had 5.0 apg, 1.5 spg, and an a/to ratio less than 2). I think a lot of times how the team finished the season colors our later perception of the excellence (or lack thereof) of its play.

I don't want to get into another knock down/drag out with you, but I don't see how this categorization tells us very much.

we are in the realm of subjective again, but i don't agree at all that greg paulus was as good of a pg as will avery. evidence for my view includes their relative nba draft positions and the fact that paulus did not even start at pg at duke at the end of his career, despite that fact that we had no one else who was ready to play the position. paulus did have good assist numbers when he played with jj, but that does prove he was a great pg in my view.

also, you stated upthread that you knew before the ncaat that the 2005 and 2008 teams weren't very good and, so, not much should have been expected of those teams. i could pull out stats to show that those teams were excellent. that wouldn't make it so, just like you posting assist numbers does not prove that paulus was a better player than avery.

[Kedsy]

my mistake on 2011. absolutely should have been in the other group. just an error. i really do think the pg thing makes a difference though. look at how much better we are this year when quinn cook plays well (and compared to last year).

College basketball is a guards' game. Of course it's better to have a good PG than to not have a good PG.

it's funny how any hypothesis about why duke is underperforming relative to seed must be imaginary or rooted in selection bias, but a hypothesis that tries to show that duke is actually not underperforming is worth testing (despite being at least as far off the mark, as shown by the data).

When you propose a subjective test and then apply your criteria inconsistently, you should expect doubters.

we are in the realm of subjective again, but i don't agree at all that greg paulus was as good of a pg as will avery. evidence for my view includes their relative nba draft positions and the fact that paulus did not even start at pg at duke at the end of his career, despite that fact that we had no one else who was ready to play the position. paulus did have good assist numbers when he played with jj, but that does prove he was a great pg in my view.

Jon Scheyer didn't get drafted either, but he ended up on your "excellent point guard play" list. So I'm not sure how draft position plays into this.

Will Avery got drafted on potential and unfortunately never reached that potential and never even got a second contract. At Duke he was a really quick guard who played good but not outstanding defense and turned the ball over more than he should have. Greg Paulus was at best an adequate defender who despite that got a lot of steals. But in 2005-06 he also got a lot of assists and was voted to the 2nd team All American freshman team and the 1st team All ACC freshman team. His later struggles shouldn't take away that he was a pretty good PG in 2005-06. Perhaps not "excellent," but all I said was he was at least as excellent as Avery. Meaning, in my opinion, 2006 and 1999 should be on the same side of the ledger, whichever side that might be.

also, you stated upthread that you knew before the ncaat that the 2005 and 2008 teams weren't very good and, so, not much should have been expected of those teams. i could pull out stats to show that those teams were excellent. that wouldn't make it so, just like you posting assist numbers does not prove that paulus was a better player than avery.

Actually, what I stated upthread (in a conversation about expectations) was I personally didn't expect very much from the 2005 and 2008 teams, which is different from saying they weren't very good. And I have never said Paulus was a better player than Avery. I merely argued that putting 2006 on the "not excellent" list and 1999 on the "excellent" list appears to be an inconsistent application of whatever criteria you're using.

[bedeviled]

I think this quoted part is more significant than you are giving credit to.

Word. I included the last 2 charts for fun. Maybe I shouldn’t have put them in a thread that tempts us to conclude more than is there. There are a lot of half-truths and nuances in viewing “expectation” as based on average wins for a given seed. However, as evidenced in this thread, it may be a good representation of people’s subjective sense of how a team performs. Thus, I find it interesting. In case anyone is contemplating this manner of data manipulation, I’ll say a few words based on my present (admittedly often confused) thoughts. If you are not interested, then you’ll need to skip ahead a few posts; this could get tedious (I actually intended just to address freshmanjs, but it seems ridiculous to PM all this stuff). The main topics I want to address are:

• Part I: It is more difficult to achieve expectations as a 1- or 2-seed, which makes Duke look worse due to Duke’s typically high seeding.
• Part II: It is more difficult to meet expectations when a team has repeated high-seeded tournament appearances, thereby also making Duke appear worse.
• Part III: Lower seeded teams get a disproportionate boost in expectation by winning a small number of games while not punished in an equal manner for losing when expected to do so. And the converse, higher seeded teams get a disproportionate decrease in expectation by losing an early game while not being rewarded in an equal manner for winning a game when expected to do so.

I’ve tried to reason out the rationales, but certainly could have made some mistakes or bad assumptions I didn’t realize I was making. Shout out any errors you find.


Part 1:
Hard to achieve expectation as a 1- or 2-seed
I’m talking about the “expectation” as termed recently in this thread – the expectation that a team, with a specific seeding, should achieve the same number of wins as the average number of wins by like-seeded teams. While some will immediately dismiss this measure as poppycock, I think it is an important one to consider. I think it might be a good surrogate variable to describe viewer’s subjective expectations. We have a feeling that 1-seeds should win and 13-seeds should lose. While this feeling is based on real experience, our brain probably does something akin to averaging out our experience to provide us with an expectation. I don’t know; it seems plausible to me. This data point is precisely that – an average of real life experience (historical tournament results) of how each seed has performed.

Since the “expectation” we are talking about is really just the seed’s average number of wins, the %achievement above expectation is actually equal to the %achievement below expectation for each seed. Thus, across the seed’s field, there is as much overperformance as there is underperformance. Consequently, there is the perception, both in basic data presented during tourney time and in personal gut/intuition, that expectation based on averages is justified and valid. However, what a seed field does as a group is not necessarily true for an individual team. A deep run by an 8-seed Butler may skew the average for 8-seeds so that 1. the seed’s number of games won above average does, indeed, equal the number of games lost below the average, but 2. the likelihood that any one team meets the average games won is not 50%.

Example for illustration: Suppose that, in the 32 teams seeded X from 2005-2012, six of those teams won the Championship (winning 6 games apiece), but the other 26 teams failed to win their opening game. The average wins for X-seeds would be 1.125 (36 games won / 32 seed appearances). Thus, according to this measure, you “expect” each X-seed team to win 1.125 games. Those that win 0 or 1 games are labeled ‘underachievers’ and those that win 2 or more games are ‘overachievers.’ The teams that won 6 games have achieved 533.33% of expectation (6/1.125) while those that lost their opening game have %achievement of 0% (0/1.125). Since, expectation is 100%, the winners overachieved expectation by 433.33% while the losers underachieved expectation by 100%. At first glance, it appears that the teams are rewarded more handsomely for wins than they are punished for losses (433 vs 100). Mathematically, though, the performances must balance out based on the odds. Accordingly,
433.33 * 6 (overachievement multiplied by the number of overachieving teams) = 2600 100 * 26 (underachievement multiplied by the number of underachieving teams) = 2600. BUT, VERY IMPORTANTLY, it is incorrect to think that any team that gets an X-seed should win 1.125 games. In fact, the likelihood is much greater that an X-seed doesn’t win any games at all!! Unfortunately, this way of looking at the data labels 81.25% (26/32) of the X-seeds as ‘underachievers’ when, here, they are performing the most common way for an X-seed to perform.

Thus, we see that the use of seed win averages as an expectation of how many games a single team should win is problematic.

The fluctuation in performance of the seed field can make it “harder” or “easier” for a given team to meet such an expectation. It is less likely (harder) for a given team to meet expectations if that seed’s average is skewed by outstanding individual performances (such as Butler’s deep run). 1-seeds are the most likely seed to advance and win championships, so the 1-seed group pretty much always has teams running past their 3.5 game average. Indeed, when we think about “special” teams with the mentality to make runs deep into the tournament, they are more likely to be high-seeded teams – particularly the 1-Seed, which has won 6 out of the last 8 Championships. The inspirational low-seed runs, on the other hand, are more sporadic and varied in their distribution (Butler (8), G Mason (11), VCU (11), Ohio (13), Davidson (10)).

However, the reverse is also true: one team’s surprisingly early exit will skew the seed’s average lower, resulting in a greater likelihood that any given team of that seed will supercede the seed’s average. As lower seeds aren’t expected to win as many games, there is more opportunity for high seeds to skew the data this way, making it more likely (easier) to meet expectation.

Thankfully, this is historical data, so we can actually look at look at how these factors played out in real life. It turns out that, for this data set, 62.5% of 1-Seeds performed below the 1-Seed average, thus making 1-Seeds seem like underachievers.

% of Teams “Underachieving” Their Seed-Based Average
Seed	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
%	62.5	53.1	68.8	59.4	65.6	43.8	40.6	46.9	53.1	59.4	56.3	62.5	75	90.6	93.8

**It looks like the 2-seed data was dramatically skewed from the general trend by the Duke & Missouri losses in addition to having no Champions during this timeframe. Also of note, I wonder if the effective cap on games (Champions can’t continue adding more wins to the end of their run) hinders the top seed’s percentages from even better mirroring of the bottoms seeds’ in this U-shaped trend.

Anyway, there you have it – the “expectation” is a moving target and more complex than “it’s an average, so half the teams do well and half don’t.” Most importantly, you can’t actually expect the expectation. In fact, in this historical sample, high seed teams are more likely to have less wins than the seed’s average than more. Nonetheless, this might be how our mind thinks, so it is important to understand our inner expectations.

[bedeviled]

Part 2:
More difficult to achieve expectations due to repeated high-seed appearances

At the end of the last post, I charted the percent of teams in a given seed that appear to underachieve according to the expectation based on average seed wins. Because none of the seeds’ win averages calculated out to integers, each tournament appearance can be classified as “underachieving” or “overachieving” (making mathematic life easier). Thus, if 62.5% of 1-seeds are not meeting the 1-seed win average, then 37.5% are exceeding the average. Hence, we see that there is a greater likelihood that any random 1-seed will appear to underperform as opposed to overperform. Moreover, by nature of event odds, the more you play, the more certain it is that you will live up the billing of “underachiever.” If we calculate the odds for additional appearances, we get the following (disclaimer: I’m a lover, not a mathematician):

2 Appearances (How often do you look bad (“underachieve”)?)
2/2 Appearances	1/2 Appearances	0/2 Appearances
39%	47%	14%

So, now the likelihood of looking bad is double that of looking good.

3 Appearances (How often do you look bad?)
3/3 Appearances	2/3 Appearances	1/3 Appearances	0/3 Appearances
24%	44%	26%	5%

Well, you aren’t coming out looking gleaming after 3 appearances! Just try to make it respectable; the combined “underachievement” percentage has increased to 68%. Notice that the odds of totally floundering are basically the same as looking good once.
I’ll post one more iteration since Duke was a 1-seed four times during the period.

4 Appearances (How often do you look bad?)
4/4 Appearances	3/4 Appearances	2/4 Appearances	1/4 Appearances	0/4 Appearances
15%	37%	33%	13%	2%

Okay, now take a look at how Duke did and compare it to the chart.

Of course, those weren’t Duke’s only appearances during the timeframe. Unfortunately, the combinatorial odds get overwhelming when looking at 8 games with 3 different seeds and probabilities. I’m not going to list out the different possibilities of performance records, but my best efforts calculate the most likely single scenario for the seeds Duke had as:
Underperforming 3/4 times as 1-seed, 2/3 times as 2-seed, and overperforming as the 6-seed. The chances of that scenario were a whopping 8.2%. Duke’s own performance, using this criteria, was a 2.4% likelihood occurrence. Throatybeard pointed out that a basket here or there would have pushed us to defying expectations, which is interesting given the likelihood of looking bad.

What’s in store for the future? You could look at this:

5 Appearances (How often do you look bad?)
5/5 Appearances	4/5 Appearances	3/5 Appearances	2/5 Appearances	1/5 Appearances	0/5 Appearances
10%	29%	34%	21%	6%	1%

and bemoan that the combined “overall underachiever” status categories now comprise a glorious 73%! Or, you could realize that the past is the past and does not affect the odds of our beating our next tourney opponent. I still hope for a one seed this year and every year thereafter. Be aware, though, that “expectations” may need to be challenged.

[bedeviled]

Part 3:
Low seeds get a disproportionate boost from expectations while high seeds get a disproportionate punishment
Once again, I’m talking about the expectations for a given seeded team to achieve the same number of wins as the average number of wins for that seed. While the last couple posts have pointed out some problems using this approach, I actually think it is still worthwhile to consider and use. Why? Because I suspect it is akin to our real-life personal, subjective expectations. So, maybe such data points and massage can’t clearly tell you about how Duke really performed, but they may help in understanding the perceptions of how Duke performed (both your own and the sports-entertainment community’s).

As an alternative method, it is also reasonable to consider each win as equal, no matter the teams or round – a win is a win. In that sense, even if it lost in the second round, a 1-seeded Duke has clearly outperformed 15-seeded NE Bumble State who lost in the first round. Duke was clearly better in the tournament than they were. BUT, that’s not how we usually operate in this scenario. Instead, we feel disappointment. Perception-wise, Duke did not fulfill its “potential” while NEBS did. Our tendency is to weight wins according to our sense of things. If both Duke and NEBS win in the first round, it is an understandable viewpoint to see NEBS’ victory as more of an accomplishment than Duke’s first round victory. It’s the way we work.

So, is the weighting we ascribe to wins and losses disproportionate? Let’s go to the data to look at the boost/decrement in the % of expectation achieved for wins and losses. First, I’ll address whether the low-seed team’s boost is disproportionate to its punishment (and high-seed vice versa), and this will kind of expand to the whole tournament field. In my Part 1 post, I used an example and purposely included information on how the gain in %achievement (of expected wins) equals the loss for a given seed. As another simple example:
Consider a team that is expected to win 1 game. If they win 1 game, they achieve 100% (1/1) of expectation. If they win 2 games, they achieve 200% (2/1), but achieve 0% (0/100) if they lose the first game. So, a game won over the expected results in a achievement gain of 100% (200-100). This is balanced by a loss below expectation, which results in an achievement loss of 100% (100-0). This is the predominant trend of the data – every game won over the expected is balanced by the same number of corresponding losses under the expectation. Seems pretty fair, right? BUT, what about game 3 in this example?!? That’s just playing with house money, isn’t it? The corresponding loss would have to happen before game 1…and there is no game –1!

Indeed, there are significant barriers in the game-by-game-balancing pattern. Seeds 13, 14, and 15 (seed 16 data is mathematically invalid due to having a 0 win average) get MORE than 100% boost in perceived achievement per game despite only being able to lose 100% of expectation in the first round (15-seeds achieve a staggering 1600% of expectation if they win one game! – yet that team can only lose 100% of achievement at most). On the other end of the expectation spectrum, 1-seeds can never actually gain an additional 100% of expectation despite being able to lose 100% of expectation if they lost in the first round. This is because they are already expected to win 3.5 games while the tournament caps out at 6 games (6/3.5 = 171%). Yet, if a top-seed loses early, they lose the achievement points from all the subsequent games they were expected to compete in...so much so that even when they are winning the opening games, they still technically have a <100% achievement rate. What gives?! How can this possibly be fair? Low seeds get huge boosts which they can’t back up with losses! Top seeds can lose everything with an early loss but can’t make an equal gain even by winning the Championship!

If it is so unequal, how does it work out to make nice, pretty averages for our expectations? Well, as I said, there is some balance on a game-by-game basis for a particular team. What ultimately balances the numbers, though, is the play across the entire seed field. Take, for instance, the 15-seed that achieves 1600% with a first round win (an increase of 1500% from the 100% expectation). These values are based on the odds, a consequence of the fact that there are only 2 teams who have actually completed this task (shhhhhh!). Thus, together, those two teams excelled 3000% above expectation. This is balanced by the other 30 15-seed teams that did not achieve the expected 0.06 wins. Each of those 30 teams lost 100% of their expectation, for a total loss of 3000%. So, you see, the whole seed field picks up the slack so that the numbers work out perfectly. Unfortunately, single teams are the beneficiary or brunt of these number shifts….which, ultimately, are perceived achievement shifts. Again, we see that perceived achievement based on meeting the seed average is a complicated matter that depends as much on the other teams and math as on the performance of the team being examined.

You may have already noticed another inequality in the ability to boost/lose perceived achievement: in addition to meeting barriers regarding how much a low-seed can lose and how much a high-seed can win, the value of each win/loss differs according to seed. Lower-seeded teams have a greater change in achievement percentage for each game played. Compare the earlier example (team is expected to win (1), but wins (0 or 2) for loss/gain of 100%) to a team expected to win (3) but wins (2 or 4) (loss/gain of just 33%). The difference is due to expectation. Is this fair? Should a high-seed team have to put together numerous wins to reach the same perceived achievement that a low-seed team gets with a single win?

Well, this is just one way to look at performance. Perhaps other ways, like taking each win at face value, can give a different perspective on outcomes. But, if you are conscientious about assumptions and expectations, I think this is an extremely interesting tactic. For specific purposes, I think these disproportionate limits and boosts/deficits in perceived achievement are, not only fair, but also very appropriate. First, the numbers are based on the historical data. This is not conjectured expectations or wishes. The data is what really happened. It balances out this way because of the frequencies that events actually occurred. Only two 15-seed teams have won. Given the unlikelihood of a 15-seed winning, I think they did way-overachieve based on expectations. And, I think it’s reasonable that the perception of Duke’s achievement took a big hit due to last year. Was Duke a good team? Yes. Were they as good as other teams that won in the tournament? Yes. Were they as good as teams that were “overachievers?” Most certainly so. However, 93.75% of 2-seeds won their opening round.

	% of Teams winning a game in each round				2005-2012
Seed	Rnd64	Rnd32	Sweet16	Elite8	Final4	Championship
1	100	93.75	71.875	37.5	28.125	18.75
2	93.75	65.625	46.875	21.875	9.375	0
3	90.625	59.375	31.25	9.375	6.25	6.25
4	75	40.625	12.5	12.5	0	0
5	59.375	34.375	12.5	9.375	3.125	0
6	56.25	25	6.25	0	0	0
7	59.375	18.75	6.25	0	0	0
8	53.125	3.125	3.125	3.125	3.125	0
9	46.875	3.125	0	0	0	0
10	40.625	15.625	3.125	0	0	0
11	43.75	15.625	6.25	6.25	0	0
12	37.5	18.75	0	0	0	0
13	25	6.25	0	0	0	0
14	9.375	0	0	0	0	0
15	6.25	0	0	0	0	0
16	0	0	0	0	0	0

The numbers are what they are. There is some validity to the unequal treatment because it is based on the odds of the actual outcomes. And, to me, I think it was reasonable to have an expectation that Duke should win. Secondly, important to me, this type of analysis makes some sense gut-wise. Averaging out past experiences may be how we develop our subjective sense of achievement. And, in this mode of thinking, I tend to think an 8-seed Butler has achieved more in a 3rd round win than a 1-seed Kansas, specifically because I do compare results to internal expectations.

This type of data set analysis emphasizes our internal expectations – low seeds have more achievement riding on a per-game basis while top seeds, due to the dependent nature of rounds, must string together wins to match our expectations. It’s a fun way to measure the shock-factor of results. And, there is some validity behind it, too, as the numbers are calculated from historical outcome. However, one must realize that the interpretation is based on expectations which are not as solid as we assume them to be (see Parts I and II also). Basing expectation on seed averages can be misleading for reasons beyond subjective seeding. And, it is only one way of looking at the data. It is not Truth. On the contrary, we have seen that there are many nuances involved with understanding what the data set really means.

[Kedsy]

Thanks, bedeviled. Really interesting analysis and obviously a LOT of thought and work put into it. I'm giving you a big round of applause from my desk chair.

It's fascinating that the most likely result from having four #1 seeds is to "underachieve" three times and "overachieve" once (exactly what we've done between 2005 and 2012). Also, it makes our period between 1986 and 1994 seem that much more amazing, as (regardless of seed) we overachieved our seed expectation 8 times in 9 years.

[Billy Dat]

Today's "Pat Forde's Take on Duke" thread, which is thematically similar to this one, made me think of an added wrinkle that I am not sure was explored in this thread - one that is harder to quantify but may be part of the trend.

On the twice-weekly ESPN College Basketball podcast featuring Andy Katz and Seth Greenberg, they recently interviewed Gonzaga HC Mark Few. Few was discussing the reality of being every WCC team's Superbowl because of the stature of the Gonzaga program and the potential for the win to improve the opponents at-large NCAA chances. Few said it takes a huge toll on the players emotions to "get up" that high for every game. Greenberg correctly pointed out that his teams never had to play that role, but that he respected the burden that it placed on teams like Gonzaga and Duke. He said that it was very hard to play that way and not end the year emotionally spent.

When do we think the full impact of Duke being every other team's Super Bowl, especially the level of hostility and anger from opposing fan bases and even casual fans, really took hold? K always says that the tide turned in 1992, but I think the graduations of Laettner and Hurley, and the dip the program took from 1995-1997, kept things neutral for a while. I feel like the start was the 2001 title team with Billy Packer and Gary Williams, let alone the entire Maryland fanbase, fanning the flames. Then, once JJ arrived, it fully flowered and really hasn't stopped. That kind of dovetails with the "Kings of Nov/Dec" era identified at the start of this post.

Maybe Duke "under performs" in the NCAAs because, since roughly 2001, there are more people then ever focusing their collective negative energy trying to bring about that very outcome.