One Fish, Two Fish, Stat Fish, Jew Fish

Why Cindarellas are Bad

From a literary perspective, I’m sure they’re wonderful. But every time a legitimately underseeded team wins a few games, one of two things has happened:

1) The team has become, for the tournament, a much much much better team than they ever were in the rest of the season
2) The team has merely benefited from the variations in performance that allow upsets to happen

If it’s the first, well then the Cindarella has a legitimate place in the coda of the tournament. If you suddenly go from the #25 team in the nation to a top 10 team, well good on you. The problem is, there’s really no reason to assume this is the case. Why would a team suddenly become wonderful, where they hadn’t been? Why are they hitting more 2s, playing better perimeter defense, getting more boards… Sure, clutch, blah blah, but there’s no reasons that the other teams aren’t responding to pressure the same way. Possible, but it falls into the inevitable category of ‘if this pattern could be produced legitimately by randomness, it’s probably more accurate to assume that its origin is random rather than try to apply some logical reason to it.’

Which brings me to the second, that it’s just a fluke. As long as you’re in the same ballpark as the better team, you should win some (Arizona has 30% against Louisville, Michigan St. has 40%). In the initial 32 games, there are probably 16-20 that have the weaker team having at least a 30% chance of winning. Which really means that 5-6 of those games should have the weaker (presumably underseeded) team winning. And in the next round, that means that 1-2 of those should win their games against better teams. Which means that it’s actually a statistical truth (albeit a shoddily done one) that, in general, there will be 1-2 teams in the Sweet 16 that are there, not because they were the best teams, but because they pulled off random upsets. And of course, there’s a chance again of one of those teams making the elite eight (not a high one if they’re Cleveland State, a much higher one if it’s Arizona.)

In short, there’s no reason to assume that a ‘cindarella’ team is actually good now, and not just ‘lucky.’ Of course, this isn’t luck per se - when David Ortiz hits 2 home runs in a game, that’s not ‘luck’, that’s skill. But, of course, he doesn’t do it every game. Some games he’ll hit two homers, some games he’ll get no hits at all. It would be accurate to say that the overall results are not luck, but the random variation/distribution in his small-sampled performances is. Villanova may be a #15 team, and that’s not luck. But some games they will play like a #30 team, and some games they’ll play like a #2 team (to average about #15), and that distribution is random, not preordained.

But anyhow, why Cindarellas are bad. Mostly, because they create stupidly easy matchups for much better teams later on. Louisville got to play Arizona (39 point win) instead of Wake Forest, last year Kansas got both Nova (15 point win) (instead of Vandy or Clemson) and then Davidson (instead of Wisconsin/Georgetown). George Mason made a very impressive run, (which itself included a game against Wichita St. instead of Tennessee), but it gave 3 seed Florida a 15 point free pass into the championship game.

Cindarellas are a natural consequence of every tournament. They make things interesting. But practically, in general, they are buoyed by fortunate circumstance, and eventually they are not fortunate, and then they get blown away by a much better team. Which is boring.

Lineup Analysis

So. There two games I play more than all others; World of Warcraft and Out of the Park Baseball. OOTP is an absolutely amazing baseball sim, with more features and wondrous things than I could care to mention (though I’d love to if people were interested). Anyhow, I was thinking about my lineup, and about how, abstractly, the perfect lineup would be organized.

So, first, my team (obp/slg) (I play with a fictional league - 32 teams, 8 4 team divisions, top 4 division winners and top 4 other teams make a 16 team playoff - I could tell you some stories):

C: Wes Strickland, 333/497
1B: Norm Moffatt, 411/550
2B: Arturo Marquez, 364/449
3B: Yasushi Ueno, 294/436
SS: Hidekazu Kiyomizu, 361/411
CF: Domingo Vela, 385/426
OF: Carlos Ruiz, 375/545
OF: Alfredo Longoria, 355/551
DH: Carroll Swain, 313/499

As you can tell, overall a fairly strong lineup (actually a bit over 6 runs a game). Some players are obviously good hitters, Moffatt, Ruiz, Longoria… And then you have the weak spots - Ueno’s the weakest link, with Kiyomizu and Swain as the niche spots - Kiyomizu being all on-base focused, and Swain being all power. Here’s the optimal lineup with this roster:

1: Domingo Vela, 385/426
2: Norm Moffatt, 411/550
3: Arturo Marquez, 364/449
4: Alfredo Longoria, 355/551
5: Carlos Ruiz, 375/545
6: Carroll Swain, 313/499
7: Wes Strickland, 333/497
8: Yasushi Ueno, 294/436
9: Hidekazu Kiyomizu, 361/411

Note the major distinctive components of this lineup:

1: Ruiz 5 and Longoria 4, even though Ruiz has almost as much slugging, and notably more on base.
2: Swain 6 and Strickland 7, even though Strickland is clearly better. This is to put Strickland’s OBP in front of the top of the order.

So anyhow. I was curious about the values of obp and slg (wondering about whether my lineup was too obp or slg heavy - last year it was much more obp, and I was worried that this year was too slg.) I ran a ton of seasons together and came up with some values. One point of team obp (ie, the difference between .335 and .334) is worth 5.58 runs a season. One point of team slg is worth 3 runs a season. I was curious about how shifting the numbers around would improve/hurt the lineup. So I ran some adjustments, modifying each batter by giving them 20 more obp at the cost of 37 slugging (equivalent numbers), and then giving them 40 slugging and taking away 21 obp (also equivalents.) Here was the result (measured in resultant runs per season), by lineup slot (extra obp result / extra slg result):

1: Vela, +4.86 / -1.458
2: Moffatt, +4.05 / +.486
3: Marquez, +3.24 / -.324
4: Longoria, -3.564 / +2.43
5: Ruiz, -1.296 / +.162
6: Swain, -4.374 / +1.782
7: Strickland, -3.726 / +1.134
8: Ueno, -2.754 / -.324
9: Kiyomizu, +3.402 / -3.24

Some interesting results. The slots where extra obp at the expense of power is desired are: 1, 2, 3, 9. The spots where power at the expense of obp is desired are: 4, 6, 7. The relatively neutral spots are 5 and 8. Taking it to the next level, I thought it would be neat to distill each player into possible points (putting their obp and slg into one pool) and then reassigning them in a way that would build the ultimate lineup (given the skill level the lineup was at). My system was imperfect, but here’s what I came up with.

1: Moffatt: 500/389
2: Strickland: 315/531
3: Kiyomizu: 390/358
4: Ruiz: 300/692
5: Marquez: 400/389
6: Longoria: 300/659
7: Swain: 290/544
8: Ueno: 270/481
9: Vela: 450/313

This new lineup scores about .35 more runs per game, or about 50 more runs a year, or about 5 more wins a year. Of course, these players don’t exist, but I tried to keep their stats vaguely realistic, if caricatured. There are several interesting things about these results. The obvious thing is that extreme players are favored - much better to have one crazy power hitter and one crazy obp hitter than two balanced hitters of comparable skills. The thing that surprised me the most was that Moffatt, the hitter with the most potential points, was shoved directly into the 1 spot, with their points spent on OBP. Here are the implied lineup desires (for an excellent lineup, like the 2006/2007 yankees or 2004/2005 red sox, or some such):

1: Butt-loads of OBP, but also not a lot of power
2: Generally, the best obp but with power - in this lineup, it ended up being different because of the 450 and 500 obps in front of the 2 slot - generally you’d want something more like the original Moffatt - 411/550
3: Mostly good obp, with a bit of power - you want good obp to get in front of the 4, 6 and 7 hitters
4: The power version of the 2 slot - the best power generally
5: like the 3 spot, but with more power
6: all power - don’t care about obp at all
7: lots of power, but with some obp - you do want to get on base for the 1,2,3 after all
8: the crappiest batter, just dump them and forget them
9: like the 1 slot - you want all obp, and not a lot of power

This is all theoretical - what really got me was the idea that you wanted extreme performers. In theory this could be a money saving mechanism, getting ‘worse’ batters that fit into a slot you need filled. Anyhow, i just thought it was neat.

Round 1 Mismatches, or why the RPI Blows

The brackets are announced, and there are the predictable surprises and controversies. I’d like to tour the round one mismatches, but first, I’d like to address a controversy that many pundits have been wringing their ties over.

That would be Arizona v. Creighton. Many pundits (Gary Parrish most notably) have been complaining that Creighton deserved a berth more than Arizona. Their argument is pretty heavily tied to the RPI.

Arizona: RPI 62, vs. top 50 RPI: 6-10, vs. top 100 RPI: 8-12
Creighton: RPI 40, vs. top 50 RPI: 2-2, vs. top 100 RPI: 9-5

It seems like a no-brainer from that. I mean, sure Creighton didn’t play many top 50 teams, but their comparative records against top 100 teams suggests that Creighton is far superior.

Here’s the problem. The RPI has some severe issues, and among them are that teams like Creighton are dramatically favored. The RPI’s biggest issue is that it tends to favor good records over tough schedules. Ostensibly, half the formula is the opponents’ records, but let’s face it, Alabama St. went 22-9. Beating them in the RPI is worth a decent boost, since they have such a good record. In reality, Alabama St. is a 200ish team, who just played a stupidly easy schedule. On the flipside, you have teams like West Virginia, who went 23-11, which is worse than Alabama St. did. Of course, West Virginia played one of the toughest schedules in basketball. I guess, in short, the RPI presumes that record indicates quality, but this isn’t accurate much of the time. In particular, the RPI will praise teams with weak schedules that win a lot over teams with tough schedules that lose sometimes.

Let’s compare using the Pomeroy numbers, which are better (in my mind) for two reasons: 1) they use margin of victory and 2) their quality of opponent metric is much more sophisticated.

Arizona: Pom 39, vs. top 50 7-10, vs. top 100 9-13 
Creighton: Pom 67, vs. top 50 1-0. vs. top 100 5-4

Arizona played 17 games against the top 50 (20 games against the top 52), and Creighton played 1. To be sure, Arizona’s record in such games is worse, but given the epic difference in schedule quality, Arizona’s 19-14 starts looking a lot better than Creighton’s 26-7. Furthermore, while 5-4 > 9-14, it is worth noting that Creighton’s record is almost totally against teams in the 51-100 ranking. Arizona’s 9-14 is mostly made up of playing the top 50. So even Creighton’s better record has some suspicion on it. (It’s further worth noting that against teams not in the top 100, Arizona went 12-0, Creighton went 21-3.)

Why the huge discrepancy? Mostly, the teams evaluated the most erroneously in the RPI are the mid-low range power conference teams, and the higher end weak conference teams. A good team (50 ranked or so) will go 20-12 or so in the ACC. A 50 rank team in the WAC will go closer to 26-8. It’s the same team, but with different schedules. The RPI will think the WAC team is better (better record), and reward you more for playing them. If you play nothing but middling teams (.500) from a power conference, you’re playing a pretty damned tough schedule. If you play bad teams in a weak conference (.500), you’re playing a pretty weak schedule. According to the RPI, they’re the same quality team.

But I’ve been going on long enough. Arizona is a team that obliterated weak teams, played mostly strong teams, and lost more than it won. They went 19-13 against what was obviously a tremendously difficult schedule. Creighton played a middling schedule (not a lot of strong teams) and went 26-7. They did well against weak teams, and held their own against strong teams. I can see the argument for Creighton being better, but personally, I’d advocate Arizona being better. Their difficulty of schedule should put them over the top. (On top of the Pac 10 (which has very few weak teams) they played UAB, Texas A&M, Gonzaga, San Diego St., UNLV and Kansas.)

Anyhow.

Round 1 Mismatches (Team, Record, Schedule, Offense, Defense, top 50, 51-100, sub 100)

Boston College (7) v. USC (10)

USC: 21-12, 4th, 62nd, 19th, 8-9, 4-2, 9-1
BC: 22-11, 53rd, 27th, 131st, 4-7, 8-2, 10-2

I’m not entirely sure why this happened. Normally the overseeded team will have a better record. But the two teams have comparable records, but USC has played a much harder schedule. They’ve gone 8-9 vs. the top 50, where BC has gone 4-7. Were I to guess? BC beat both Duke and UNC, where USC has no victories like that (besides a victory over UCLA in the P10 tournament.) But beyond that… If someone can figure out what makes BC a better team than USC, I’d love to hear it :)

Florida St. (5) v. Wisconsin (12)

Florida St.: 25-9, 15th, 95th, 12th, 6-7, 10-2, 9-0
Wisconsin: 19-12, 27th, 24th, 60th, 4-10, 6-2, 9-0

This one is a much more even thing. Where USC is clearly better than BC, these teams appear to be matched appropriately. Florida St. has a better record than Wisconsin, and has played a tougher schedule. So why is this listed as a mismatch? Margin of Victory. Wisconsin is +32 in 25 games against Top 100 teams, even though they’re 8-12 in those games.

Wisconsin Wins against Top 100: 29, 20, 17, 13, 12, 10, 5, 5, 4, 2
Wisconsin Losses against Top 100: 19, 13, 11, 7, 5, 5, 4, 4, 4, 3, 3, 1

Please note. Wisconsin’s wins are mostly in double digits, with only 4 of 10 wins in the singles. On the flipside, 9 of their 12 losses are in the single digits, with 8 within 5 points. This may sound like an evasion of their comparably poor record, but it’s not. Which would you favor in a neutral game, a team that is 10-0 with their wins by 1 point each, or a team that is 5-5, with wins by 20 each, and losses by 1 each. In general, the 5-5 team will win that match, 70-75% of the time. While there is a limited clutch skill involved in basketball, winning close games is more about luck than skill. In short, Wisconsin has been performing a lot better than their record indicates. Pomeroy has a stat called, rather intuitively, luck. Luck is basically a reflection of how good your record is given your Margin of Victory. High luck means the team won more than the performance would suggest. Wisconsin’s Luck rank is 312th out of 340. Almost no teams had a record that far behind their performance.

Florida St.’s luck rank is 20th. Let me show you what I mean.

Florida St. Wins against top 100: 13, 13, 11, 11, 11, 10, 7, 4, 4, 3, 3, 3, 2, 2, 2
Florida St. Losses against top 100: 23, 14, 10, 8, 8, 6, 5, 3, 3

Florida St. is +19, but they’ve gone 16-9 in these games. So, to recap, Wisconsin is +32 in 20 games, Florida St. is +19 in 25 games. Wisconsin has outperformed Florida St. on paper, but has been dramatically outperformed in actual record.

Do I believe that the Wisconsin is better than Florida St? Maybe. But the point of the article is that this particular 5-12 matchup is nowhere near as dramatic a mismatch as it looks like. Watch out for Wisconsin.

Michigan St. does not deserve a 1 seed.

Seriously. They’re a good team, no question, and they may well win the B10 tournament. But that wouldn’t make them more deserving of a 1 seed than UConn, Pitt or Memphis. This is a team with some pretty freaking brutal losses under its belt. Losing to Maryland by 18! Losing to UNC by 35! Losses to Northwestern and Penn St by 7 and 4 respectively, both at home. And losing to Purdue by 18 away. They’ve won some good games, including good defeats of Kansas and Purdue, but the above losses are very worrisome for a team that, evidently, is considered to be one of the top 4 in the country. Any top team can have one or two of those losses, even great teams lose badly sometimes. But having 6 of them? Losing to UNC is fine, but losing by 35 is a joke. Losing to Purdue away is fine, but not by 18.

Contrast this with Pitt, UConn or Memphis. (Just losses being listed)

Pitt: Louisville 6 away, Villanova 10 away, Providence 8 away, West Virginia 14.
UConn: Georgetown 9 home, Pitt 8 home, Pitt 10 away, Syracuse 10.
Memphis: Xavier 5, Georgetown 9 Away, Syracuse 7 home.

There are a few losses that don’t look good, the georgetowns, the west virginia, the providence…

But Pitt as maybe 2 bad losses (bad losses = either losing to a team that had no business winning or losing by more than appropriate for a top team given the quality of opponent), in Providence and West Virginia. UConn has Georgetown by 9 at home. Memphis has Georgetown by 9 away.

I’ll be the first to admit that Memphis had an easier schedule than the other 3 (the other 3 are in the 13-20 range of difficulty ranking) while Memphis enjoyed a 68th ranked schedule. But any of those teams look way better than Michigan St - it got blown out by good teams, and lost to mediocre teams who don’t rate a bubble berth.

Is Michigan St. a bad team? Come on. They’re in the top 15, no question. But the fact that Pitt and UConn lost *recently* doesn’t make them somehow magically worse. Michigan St. lost more than the above 3, and it lost worse to worse teams. So let’s be reasonable. That two projected #1 seeds lost is remarkable. But to suggest that Michigan St. of all teams should ascend to their place… A bit silly.

Upset-tastic

This Thursday, four different top teams got knocked off of their high horse. Oklahoma was edged out by Oklahoma St. 61-60, in the quarterfinal round of the Big 12 tournament, Kansas was dropped by Baylor 71-64, Pittsburgh was dominated by West Virginia 74-60, and UConn lost to Syracuse 127-117 in 6 (count ‘em 6) overtimes. Pitt and UConn were both projected #1 seeds, Oklahoma was projected as a #2 seed, and Kansas as a #2-3 seed. What do these losses mean, both for their seedings and as reflections of their true quality?

UConn lost to Syracuse. Syracuse is a damned good team, in the 10-20 area, and losing to them 127-117 is no embarrassment, even for a 1 seed. I suspect that between the quality of Syracuse and the epic nature of the loss, I suspect the selection committee won’t hold it against UConn - their 1 seed will be intact.

Oklahoma lost by 1, to a quality team. Oklahoma St. is unranked, but the Pomeroy numbers have them as the #30 team, which I buy. Losing by 1 to the 30 team isn’t a big deal. Well, depending on how good you think Oklahoma is. If you think they’re a top 6 team, then it’s a moderately nasty knock. If you think they’re #18 (Pomeroy), it’s less of a big deal. #18 teams lose to #30 teams narrowly from time to time. They’ll probably keep their #2 seed. Another projected 2 seed, Kansas, lost to Baylor, and the 3 seeds aren’t a terribly threatening bunch. Practically, the loss was mostly a product of Oklahoma’s going 3-19 from beyond the arc. Had they performed normally, they’d have won by 8-11. But when you only win by 3-4 a game, it means that when you get unlucky, you lose.

**Side Note**

There is a divergence of opinion between stat people and fans regarding the significance of close wins. Most fans look at a team like Oklahoma, with 3 losses but a ton of narrow wins, and conclude that they’re looking at a great team that pulls together in tight spots. A stat person looks at the close wins, and figures that had chance gone slightly in the other direction the wins would have been losses - that a team that goes 10-0 in 1 point games should be more realistically treated as a 6-4 or 7-3 team. I’d like to think Oklahoma’s loss is a reflection of this. But of course, one game isn’t reflective of anything :)

Kansas lost to Baylor by 7. That’s a hard loss to a #51 team, for a team that was a projected #2 seed. This loss has a damned good chance of costing Kansas their shot at a 2 seed. If any of the projected 3 seeds runs the table on their tournament… It’s just a really disappointing loss - top teams will lose top #50 teams every now and then (paging North Carolina) but it’s still a nasty blow. It was a fair loss too, not 3 pointer luck.

Pitt lost by 14 to West Virginia. Losing to anyone by 14 when you’re a #1 seed is a huge knock. But West Virginia is unranked, projected as a #7 seed. Losing to a team like that at all is rough, losing by 14… But West Virginia is nowhere near as bad as all of that. UWV has played the 8th hardest schedule in the NCAA. Pomeroy has them at #9, and while that may make some people scoff, to have a 10 loss team that high, they are substantially better than popularly conceived. Even losing to a #9 team by 14 is huge. West Virginia got to the line 27 times, and made 25 of the resulting free throws. Will it cost Pitt a 1 seed? I don’t think so. The loss was rough, but does it drop Pitt lower than Memphis and Louisville? I doubt it, but it could happen.

As for the winners? Baylor’s resume is too weak for a seed, even after a good win over Kansas. West Virginia? I’d like to think that this win will bump them to a 5-6 seed, but that’s probably wishful thinking. If West Virginia drops both Syracuse, aaaaand the winner of Louisville / Villanova (which could happen), that would definitely do it. But we’ll see :) Oklahoma St.’s path to a higher seed lies through Missouri, a road that I do not envy. Syracuse is projected as a 6 seed, but they deserve better. The win over UConn will help a bit, but it’s West Virginia they need to beat.

Watch the end of the conference tournaments. We’ve got underdogs aplenty, and seeds are up for grabs :D

2009 Player of the Year

After my bold declaration about DeJuan Blair’s superiority, I had a moment of hesitation. Blake Griffin is an excellent player, and I hadn’t truly run the numbers on the two of them. And you know me, I hate making assertions without numbers to legitimize my iconoclasm. So I sat down with the two of them, to try and figure out their values. (It truly is a two man race, there isn’t anyone else with their performance levels. Hansbrough is certainly competitive on the scoring end, but his rebounding is firmly on earth while these two are soaring in the clouds.)

So. Comparing Griffin and Blair. Griffin is a much better scorer. He shoots extremely efficiently at 1.27 points per shooting possession, and he takes a huge number of shots. His shooting is worth 120.69 points over an average player. Griffin is an excellent offensive rebounder, and a spectacularly good defensive rebounder - his rebounds are worth about 170.7 points above average. Blair puts up fewer shots, and does so less efficiently, and his scoring is only worth. 62.49 points. His rebounds, on the other hand, are the best in basketball. The man almost gets more offensive rebounds than the next two closest players combined. His rebounding is worth 215.2 points. Put these numbers together and Griffin comes out on top, but this is missing the major remaining variable - Griffin turns the ball over much more often, because Blair turns the ball over incredibly rarely. That difference is worth 30 points for Blair. Take those numbers, shake them to adjust for possession numbers, and the differences are, as they say, quite similar. Over a 30 game season, Blair is worth appx. 9.13 wins above average, and Griffin is worth 8.97 wins. (This is for a team that otherwise would be 15-15; a better team would make their value worth fewer wins comparatively.) Tell me that isn’t a little mind blowing - take a completely average team (say Witchita St. or something) and give them Blair/Griffin’s rebounds, turnovers and scoring, and that team becomes Brigham Young / UCLA / Xavier / Clemson, a serious top 25 team at 24-6. Phrase it from the flip side - take away Blair from Pitt, and Pitt plummets from a top contender to being Texas level (high seed but not a serious contender) - if Oklahoma lost Griffin they’d be at Wisconsin/Temple level (probable bubble team). Overall, however, Blair does come out a bitty bit on top. Add in the fact that Pitt has played better teams (19th SoS as opposed to 47th) and the numbers definitely favor Blair, very slightly.

One thing though. One thing the calculation doesn’t consider is the value of the shots Griffin takes - every shot he takes has value not only for its efficiency, but also for the fact that his teammates are able to take higher efficiency shots. It is this particular value that I haven’t calculated because it is, frankly, a pain to mess with, and today is not the day for me to slay that particular monster. Griffin’s value from this contribution is much higher than Blair - Griffin takes many more shots. How valuable is this? Not sure, but I’d honestly guess that the difference is enough to make the numbers favor Griffin again. Add in the fact that Griffin plays a lot more minutes than Blair… I guess it’s hard not to give the tiebreaker to Griffin. Of course, at this level, if there’s a substantial difference in defensive performance, whoever’s better would probably be the most valuable.

In short, there’s not really a good answer here. They both deserve it. My instinct would be to grant a slight edge to Griffin. That’s bearing in mind, however, that Blair is an offensive rebound way way way beyond anyone we’ve seen in the last 5 years. He gets more offensive rebounds than some teams. And he deserves a ton of attention for that. (One other tidbit, on average teams Griffin would be more valuable, because the value of an offensive rebound depends on how good your offense is at shooting, so Blair’s offensive rebounds at Pitt are 20% more valuable than they would be at an average school. Just saying, it would shift things a bit.)

NCAAB Player of the Year

I don’t want to get sidetracked from the above analysis, but I was subjected to a rude awakening by someone who posited that the PoY will not be DeJuan Blair.  This, to me, seems beyond belief.  We’re talking a player that grabs more than a third  again the offensive rebounds of the next closest player - in the last 4 years.  He’s in another world.  Oh, and Oklahoma is seriously overrated.  Just saying.

Celtics Defense - Who is Pulling their Weight?

The Boston Celtics last year had the third best defense the NBA has seen since the early 70s, and this year they are performing at the same level. Who is responsible for this? Garnett has a well deserved reputation for defensive excellence, Rondo and Perkins look pretty good given their youth, but both Ray Allen and Paul Pierce had worn the ‘talented scorer that doesn’t try on defense’ mantle until very recently. Yet somehow they’ve combined for a historically efficient defense.

Player defense in basketball is hideously fickle to work with - where team defense is very easy to calculate (indeed, as easy as offense) player defense is a genuine struggle, most notably because there are almost no worthwhile defensive stats to work with. Attempting a way around the problem, I came at things backwards. Instead of trying to figure out Paul Pierce’s defensive contribution through his stats, I figured it out using the stats of the players he was guarding. If I know how much value LeBron contributes on average (if anyone’s interested in my value assessment formula i’d be happy to go into it), then I can compare it to how much he contributed while covered by Pierce, and assert that the difference is in Pierce’s defense. It’s an imperfect method, but it is reasonably accurate in broad strokes. Let’s go… into the numbers (queue the Wayne’s World transition music…)

Rajon Rondo

Strengths: An excellent shot preventer. Opposing PGs lost .55 points a game off their 2 pointers, and .57 points a game off their three pointers. In aggregate, opposing PGs average .08 points less per shot than normal (this counts free throws as well.) He forces a lot of turnovers, creating an additional .38 turnovers in the men he guards per game. He denies his opponents good passing opportunities, at .74 assists per game. He also forces his assignment into more fouls than normal, at .54 fouls per game.

Notstrengths: About average at stopping penetration and preventing free throws - opponents got an average amount of free throws off him. Poor at stopping offensive rebounds on his assignment, to the tune of .12 rebounds above average a game. That’s about it.

Total Value: 1.98 points per game.

Ray Allen

Strengths: Monster shot stopper. Opposing SGs lost .48 points off their 3 pointers, but .91 points off their twos. Good at stopping penetrations, .3 points off of FTs less per game. Opponents get .8 fewer assists per game. His assignments lost a staggering .12 points per shot.

Nonstrengths: About average regarding rebounds, slightly subpar at forcing turnovers (.11 fewer than normal per game) and his assignments commit no more fouls than normal.

Total Value: 2.06 points per game.

Paul Pierce

Strengths: Good shot preventer. Assignments lost .3 points per 3 pointer, and .64 points per 2 pointer. Decent at preventing opposing FTs, at .11 FT points per game. His assignments put up .08 fewer points per shot. He is good at preventing opponent offensive rebounds, at .22 per game. His assignments get .3 assists per game less than normal. His opponents commit .36 more fouls per game.

Nonstrengths: Slightly below average at forcing turnovers, at .09 fewer per game. That’s about it.

Total Value: 1.52 points per game.

Kevin Garnett

Strengths: Monster shot preventer, at 1.06 fewer points on 2 pointers, and .4 fewer points on 3 pointers. Opponents put up .62 fewer assists per game. Assignments average .1 fewer points per shot.

Nonstrengths: Slightly below average at stopping Free Throws, at .11 FT points per game above average. Average at Rebounds, Turnovers and Fouls.

Total Value: 1.75 points per game.

Kendrick Perkins

Strengths: Decent shot stopper, at .25 points on 2 pointers and .09 points on 3 pointers. Quality at stopping FTs, giving up .24 fewer per game. His assignments score .04 points less than average per shot. He is amazing at stopping rebounds, at .75 fewer per game.

Nonstrengths: Average with fouls and assists. Slightly below average at turnovers, forcing .11 fewer a game.

Total Value: 1.33 points per game.

Thoughts

The interesting thing is the ranking:

Allen 2.06
Rondo 1.98
Garnett 1.75
Pierce 1.52
Perkins 1.33

It should be noted that, of course, this doesn’t take into account the effect of Garnett and Perkins on the perimeter players. With helping bigs, all three can play much closer defense than they might normally do, which would allow them to dramatically lower their assignment’s shooting efficiency. All five players are within .7 points of each other - is this an indication of the inaccuracy of the method, that the effects of the good players on the defense are somehow aggregated? I say no. This is one of the best defenses in history, it stands to reason that each player be a substantial contributor. Surely, a good amount of the smalls’ shot prevention should be credited to the bigs - this isn’t a true indication of personal value. But it does tell us some things.

An interesting bit of data is how often opponents shot on each player. Both Allen and Rondo had their assignments take one less shot a game against them than normal. Perkins had opposing shots drop by .34. Pierce didn’t see his shots taken drop at all, but opposing PFs actually took .43 more shots per game than normal. It’s interesting that Rondo and Allen, who both were extremely effective at reducing shooting efficiency, had fewer shots against them, but Garnett, who was even better at shot defense, got an increase in shots against him. How weird is that?

It’s worth noting that this team does not prevent offensive rebounds particularly well (besides Perkins), nor do they force turnovers very well (besides Rondo.) What they do incredibly well, is shut down opponent shooting. Even Perkins, who is the weak link on this defense (though he’d likely be an anchor in another) slows down opposing shooters, and Garnett, Rondo and Allen are smothering.

The data put out isn’t ironclad, but it’s certainly interesting. :)

I have a lot of data here - who each player struggled against, dominated, how they fared against top tier competition, etc. I just didn’t want to vomit numbers everywhere and lose focus. If any further information would be helpful in any way, I probably have it somewhere, so let me know :)

Starbury Stats

Well, I can’t make substantive comments on Marbury’s defense.  (I mean, I could, but I’m not going to mess with it today.)  The 82 games +- numbers on Marbury’s defense are not favorable though - over his most recent 3 full seasons (05-07) the Knicks allowed from 4, 1.4, and 5.4 more points per 100 possessions when Marbury was on the court.  Exactly what that means is uncertain, but it sure as hell isn’t good.  Offensively, I can offer more info.

Over the past 3 seasons, Marbury’s offense has cost the Knicks anywhere from 1-2 wins a season as compared with an average starting PG (think T.J. Ford level offense).  He’s been a sub-par shooter, with 3 pointers unremarkable and 2 pointers poor.  He does generate above average points with free throws.  He dishes assists badly and commits fouls more than average.  He rebounds about average.  His one saving grace is that he commits turnovers rarely.  The major hole in his game is the shooting and the poor assists.

In 2005, it was a totally different story.  That year his offense was worth nearly 4 (count ‘em, four) wins over average to the knicks.  He was a decent shooter then (3s were bad, - 2 pointers were quite decent), but his real scoring value was from his free throws - the only players this year who get more per minute out of free throws are Devin Harris, Chauncey Billups and Chris Paul.  His rebounding was about the same, but everything else changed.  He committed fewer fouls, committed slightly fewer turnovers but his assist rate went from ass-terrible to respectable.  In short, when on his game (way, way, way back when) he was a top tier offensive PG, in the Steve Nash/Jameer Nelson/Chauncey Billups level.

What does this mean?  Well, it depends on how much of it is a product of his environment.  But the falloff in 2 point effectiveness and free throw efficiency suggest that a lot of his game was penetration, and that in the past 3 years his athleticism has deserted him.  Assists…  well, who knows, but they don’t look good.

How does he compare to Eddie House?  About the same, probably a little better overall.  House is a monster shooter from the outside, but beyond that is inferior to Marbury in every way.  Assuming what Marbury has done is what Boston would get, and assuming his personal issues wouldn’t cost Boston anything…  unless they’re buying cheap, I’m not really seeing it.  His offense doesn’t really justify much of an expenditure, and his defense is, from a casual examination, substantially worse than what Boston could get elsewhere (or from House.)  And then let’s add in the fact that the man has been riding pine since January of ‘08…

That said, if the price is right, I can see the rationale for a conservative gamble.  Worst case scenario, he blows, they’re out however much, and he doesn’t play.  But best case, he recaptures even some of his ‘05 form, and becomes a substantial contributor off the bench.  If the money is right (and the guy isn’t a cancer), this is a shrewd move.

(Say it ain’t so) Starbury to the Celtics?

Josh:

John, I’d almost convinced myself that this story was dead, but yesterday ESPN reported that the Knicks had negotiated a buyout with Marbury and that he was already in talks with Boston. This makes me slightly ill to my stomach, as the Celtics are without question my second favorite team in sports. I’ll try not to go into all the my-god-what-an-asshole reasons that Starbury is an awful fit for my team, and do my best to analyze it from a chemistry perspective, which I think is really the proper analysis of why this is a TERRIBLE idea.

This is a Celtics team that is built on defensive intensity and, as is almost a given, a very tight team cohesion. This was one of the reasons I was so sad to see James Posey leave - he was both an excellent defender and an excellent teammate/glue guy. I don’t think it’s going to surprise anyone when I say that Marbury is something like the complete opposite. He is selfish, prone to fights, makes cliques in the team, and his attitude in adversity is pitiful. He’s a quitter, John, a whiny selfish quitter. We’ve also been down the KG and Marbury road before, and a combo that could have been one of the better inside-outside games of the last decade ended in a bitterness that stayed with that organization until basically last year when they cleaned out their roster. Obviously he’s not entirely to blame for the implosion of the Knicks franchise, but he is one of the people who was responsible. I do NOT want truck parties on Causeway.

There are some other arguments to make too. A lot of Celtics say that the most important player on the floor each night is Rondo, and I’m inclined to agree. The way they run their offense, it is essential for him to make good drives to the hoop, finish well and kick out well. He’s a tremendous player when his game is on; not quite at the Chris Paul/Deron Williams/Derrick Rose level, but only one small step below them. The problem is, his ego can be a little fragile. When the team is in a bit of trouble, he is not the guy who is going to have the breakout game to snap a losing streak. He’s also not the one who’s going to turn around a post season series on his own. That’s not his game. One of the most important priorities in Boston has to be keeping him confident and loose. When Sam Cassell showed up last season and started eating up Rondo’s minutes, Rondo really struggled in a way he hadn’t all season long. And it wasn’t rust, it was that his game got too tight. He forced shots, he forced passes, and it got ugly. And that was with an all around great teammate like Sam Cassell. What will it be like when Marbury starts bitching about playing time? What about when Rondo has a stinker (everyone does) and sees Marbury arguing for more playing time? Ugh. Team chemistry is SO important in winning franchises (ask the Spurs).

As for the argument that they need another PG, I do understand it, and it makes sense. Eddie House, who I absolutely love, is not a smooth ball-handler and really has trouble against a strong press. He’s a lot better as the swingman, which is where the Celtics have used him at times this year (with Rondo still at point). Also, he works with the offense running through Pierce or KG, as he is quick enough to elude most defenders and has a good enough perimeter game to make those stop-on-a-dime threes. I’d love to have a proven scorer and veteran coming off the bench to back up Rondo and put House as the 2 in the backup team; I think that puts the Celtics right back on top of the East. I just wonder about what it does to the front line of the team.

Any statistical analysis to offer? Can you make me feel better about it? Do I have to cheer if he’s wearing the green?