Adam Morrison, Meet Barry Parkhill
Posted by Neil Paine on July 20, 2009
Last time around, we ran a study using a system of estimating Win Shares in seasons prior to 1978, and I wanted to touch a little on how that was done. Basically, I ran an OLS regression on all pre-1978 players, based on player stats (plus age & height) from 1978-2007, that estimated their "missing" totals -- turnovers prior to '78, TO/BLK/STL prior to 1974, etc. Whenever team stats were available, I scaled up/down the individual numbers to match team totals. When team numbers weren't available, I had to estimate them as well using the same method, and then makle sure the individuals matched the teams. In other words, the team numbers always superceded the individual totals.
Now, do I think this is the perfect way to estimate pre-1978 stats? No. For instance, we're estimating stats for 1952-1977 (an era with no 3-point shot) based on a 1978-2007 sample that almost completely includes the 3-pointer. We're also neglecting differences in height, playing style, pace, etc. across multiple eras of the game. So there's certainly ample room for improvement. I also know that WhatIfSports has apparently estimated these old numbers as well for their sims, but I have no idea what method they used (maybe I should just ask?). Either way, there will be better ways in the future, but for now, as a quick-n-dirty way of comparing players from pre-1978 with modern ones, I think this method suffices.
Anyway, I was examining the expanded Win Shares database I have thanks to the estimation process, specifically looking at career per-minute WS rates among players with a certain number of minutes played. Since 1952, there have been 1,614 players who have 3,000 or more combined career MP in the NBA and the ABA. Among those players, the best career per-minute WS rates belong to George Mikan and Wilt Chamberlain. And only two players have a lifetime WS/3000 MP rate of worse than -1.00: Barry Parkhill and Adam Morrison.
Who exactly was Barry Parkhill? Like Morrison, Parkhill was a college star. At the University of Virginia, the 6'4" guard averaged 21.6 PPG as a junior in 1972, and was named both the ACC Men's Basketball Player of the Year and the ACC's Athlete of the Year that season as he led the Cavs to only their second postseason appearance ever. His numbers fell off quite a bit as a senior (21.6 PPG/4.5 RPG/4.3 APG/.450 FG% in '72 slipped to 16.8/3.7/5.0/.402 in '73), but the school retired his number after the season, and his college career was good enough to warrant being named to the ACC 50th Anniversary Team in 2002 as one of the 50 greatest players in ACC history.
Also like Morrison, Parkhill was a 1st-round choice in the NBA draft, albeit going 12 picks later (the Blazers picked him 15th overall in 1973). But instead of playing for Portland, Parkhill decided to spurn the NBA and play in the ABA for his hometown Virginia Squires. The Squires had actually drafted Parkhill after his standout 1972 campaign, but the pick was voided by the league because he was not a senior. Upon joining the Squires, Parkhill played horribly in 1974, averaging 4.7 PPG on .371 shooting, and he was reduced to 14.5 MPG on a team for which he could have conceivably carved out a role as a sixth man, had he simply lived up to his billing at UVA. In 1975, Parkhill played almost 2,000 minutes and was essentially just as horrendous as he'd been the year before, and after spending '76 languishing on the bench for the Spirits of St. Louis, Parkhill's career in hoops was done by age 25.
As for the 25-year-old Morrison, he's getting some summer league burn right now and playing well, hoping to recapture whatever is left of the promise he showed as a collegian. And let's get one thing straight: he was a better prospect than Parkhill ever thought about being (3rd overall in a 30-team league >>> 15th overall in a 17-team league). Plus he has the excuse of missing all of his age-23 season with an injury. But it's mind-boggling that we can even make a comparison between the two, given the hype that surrounded Morrison in 2006. In both cases, gaudy junior-year stats (does 28.1 PPG in the '06 WCC = 21.6 PPG in the '72 ACC?) and accolades drove unwarranted expectations and absolutely hideous numbers once they reached the pros. No player has ever played remotely as badly as Morrison and Parkhill did in their first three seasons, so we're sort of charting new territory here: if Morrison somehow manages to turn things around and become a serviceable player at this point, it would be unprecedented in the history of pro ball. And if not? Well, he becomes another footnote in basketball history, going down in the record books alongside the likes of Barry Parkhill.
July 20th, 2009 at 7:31 pm
Interesting read Neil.
July 21st, 2009 at 11:06 am
According to the article at the link, the selection of Parkhill that was voided was actually in 1971, following his sophomore year. I had never heard about this incident before. I'm a little surprised that the pick was voided, because the ABA always seemed to play a bit fast and loose with allowing teams to draft and sign undergraduates. This was certainly true in the ABA's later years (e.g., the Stars drafting Moses Malone out of high school). But even before 1971, ABA teams had already signed Spencer Haywood and Ralph Simpson as undergrads. Maybe at the time of the 1971 draft the ABA was still trying to keep up appearances in this regard, viewing players like Haywood and Simpson as exceptions to be dealt with on a case-by-case basis.
In any event, the Squires drafted Parkhill again in a "Special Circumstance" draft held in January 1973, and this time it stuck. This is the "secret draft" mentioned in the article at the link. In this draft, each ABA team was allowed to select two players, who could apparently be seniors or undergrads. Players selected seem to have then been ineligible for the regular 1973 ABA draft. Why did the ABA hold this draft? The answer is probably just that the ABA liked holding crazy special drafts like this. Anyway, when Parkhill finished college in 1973, the Squires held his ABA rights due to having selected him in this draft.
Besides Parkhill, there was one other player selected in the first round of the 1973 NBA draft who also went the ABA route and would also not make it past the merger: Kevin Joyce. IINM, though, Joyce's career was shortened due to injury (in fact, I believe he was actually on the Pacers' roster during the 1976-77 season, but was on the injured list and did not play in any games). Joyce had one decent injury-free year in the ABA, which I assume pulled his WS rate up to a higher level than Parkhill-Morrison territory.
July 21st, 2009 at 11:22 am
"No player has ever played remotely as badly as Morrison and Parkhill did in their first three seasons,.."
Wow, strong language to describe someone getting 2000 minutes for a really bad team. WS are always driven up or down by a team's level of success.
There have been a dozen player-seasons with a PER worse than Morrison's 7.9 in 2007, w. over 2000 min. Lower the requirement to 1000 min, and there are 156 which are worse. About 3 per year.
Intuitively, one guesses the true 'worst' NBA players get a lot less than 1000 minutes. Last year alone, there were 46 players over 100 minutes w PER < 7.9
July 21st, 2009 at 12:27 pm
Good point, Mike. Still pretty sad that any viable metric could find a player as highly touted as Ammo to be even among the worst ever. I was really pulling for that kid (mostly because I'm a lifelong Jordan fan, and wanted him to have a successful draft... sorta given up on GM MJ at this point), and I hope he gets a chance to establish himself as a meaningful component to a team.
July 21st, 2009 at 1:31 pm
Mike, did you ever consider that a certain player getting major minutes on a bad team might be one of the reasons the team is so bad? You act like 2 players with identical stat lines, one on a good team, one on a bad one, are equal in "success". That's simply not true. Better or worse teammates can certainly make a big difference, but at some point you have to acknowledge that box score stats are not accumulated in a vacuum. If a player keeps playing on a bad team, do you ever think there might be a reason? Team wins are all that matter. Player metrics have to tie in to team wins somehow. A team must be the sum of its players. Winshares do this and I don't know why you keep having a problem with that.
July 21st, 2009 at 3:48 pm
Bruce, not only have I considered that "a certain player getting major minutes on a bad team might be one of the reasons the team is so bad?", I tend to look at the lineup on said bad teams to see who could have gotten more minutes, so that the team could have been 'less bad'.
http://www.basketball-reference.com/teams/CHA/2007.html
The 2006-07 Bobcats were a 33-win team, that by it's point-differential should have been a 30-win team, in a weak conference. They're a 27-win team in an average strenth-of-schedule.
Coach Bernie Bickerstaff allotted the minutes, and he could have given fewer to Morrison and more to Matt Carroll, Brevin Knight, Walter Herrmann, Derek Anderson, Jeff McInnis, and/or others who were injured, clinging to the tail end of their careers, or otherwise poor alternatives.
Gerald Wallace, Ray Felton, and Emeka Okafor were the only good/decent players on this team, and they each missed 6 to 17 games. Someone had to play, and there wasn't much to go with. Morrison was at least available for 78 games. Without him, I am sure the team wins fewer games.
No player on a roster can be "one of the reasons the team is so bad". If that were so, the team would just cut him and run with a short roster. A below-average player (or even a severely below-avg player) is one of the reasons the team isn't even worse than it would otherwise be. Average players don't magically appear in the lineup when you cut below-average players. If that were the case, any team could become average, instantly. In reality, many teams would love to be average.
I've estimated that Morrison added 1.4 wins to what the Bobcats otherwise would have won in 2007. This isn't much, for a 30 mpg player. But it's more believable than -1.5 (WS), whatever it is that implies. Among players with 1000 minutes that year, he ranks 246 of 261; ahead of Damon Jones, Snow, Payton, Farmar, Telfair, Hassell, Jeffries, A Wright, Jaric, Gelabale, Dahntay Jones, Buckner, L Wright, and check out Diawara of Denver.
July 21st, 2009 at 4:59 pm
C'mon, Mike, we're running headlong into this "ability vs. value" brick wall again. Bickerstaff kept playing Morrison not because his past performance was warranting it, but because he expected his future production to improve and meet his alleged talent level. It didn't. Morrison may have been more talented than his backups on Charlotte that year, but that doesn't give him a free pass for his lack of production. Somebody (whether Morrison, the coaching staff, the front office, whoever) insisted that he take 24% of the team's shots when on the court despite producing just 0.91 points for every possession he consumed. Likewise, somebody insisted he play 30 MPG despite horrendous defense. That behavior was detrimental to winning basketball, and he has to be called to account for it. You can't just excuse it by saying, "the coach obviously thought he was better than the guys on the bench." Lack of production is lack of production, and it has nothing to do with the player's perceived talent.
July 21st, 2009 at 5:03 pm
Mike G wrote:
Without [Adam Morrison], I am sure [Charlotte] wins fewer games.
That's quite a statement. How can you be "sure" that Charlotte would have won fewer games had Morrison's minutes been distributed among the other players on the roster? Or are you saying that the Bobcats would have won fewer games if Morrison had been removed from the court and the Bobcats had been forced to play four-on-five? If it's the latter, then I agree. If it's the former, then I disagree.
I’ve estimated that Morrison added 1.4 wins to what the Bobcats otherwise would have won in 2007. This isn’t much, for a 30 mpg player. But it’s more believable than -1.5 (WS), whatever it is that implies.
Why is 1.4 more believable that -1.5? (Negative Win Shares essentially mean that a player was so bad that he was taking away wins his teammates were generating.)
July 21st, 2009 at 6:51 pm
I understand that negative net contributions can theoretically be produced; and that the most egotistical, strong-willed general manager might coerce a given coach to play his boy 2300 minutes, even if he is not among the top 200 players in the league. But this is still a far cry from that player being the worst to ever appear (significant minutes) in the NBA.
Perhaps, when a negative value appears in the WS column, it is incumbent upon the author of b-r.com (or this here blog) to examine just how this number could possibly be reduced, to benefit the team: Who exactly should have gotten more minutes?
Obviously, someone has to be the worst player ever to get 2000 minutes in a year. If he only gets half as many minutes, is he only half as bad? Should he have refused to go in when told to? Would that make him a better/less-bad player?
Such logical contradictions might be a red flag. Points are not proportional to Wins; there's an exponent (14-ish) involved. Why would we suppose 'contributions' (however measured) would be proportional to wins?
July 21st, 2009 at 7:42 pm
I find this to be a logical contradiction:
"Morrison may have been more talented than his backups on Charlotte that year, but ... That behavior was detrimental to winning basketball"
If they don't have better players to take his minutes, then giving more minutes to worse players is more detrimental. They won 30 games; lots of teams have done worse, and it seems you're saying this team should have done worse. But why would they choose to?
July 21st, 2009 at 8:09 pm
A no brainer really, giving more minutes to worse players is going to be more detrimental.
July 21st, 2009 at 8:54 pm
Mike, I see you avoided my question. Once again, you wrote:
"Without [Adam Morrison], I am sure [Charlotte] wins fewer games."
How can you be sure of this?
July 21st, 2009 at 9:21 pm
Justin, your estimate of Morrison's wins added (-1.6) and mine (+1.4) are 3 wins apart. That is just about the difference in their actual (33) and pythagorean (30). So, to be more precise, I am not certain that they win more with him than without (since we'll never know); but I am certain they are more likely to win more games with him than without him.
You may be a 30% shooter from the arc. Someone may tell you to never shoot from there, because that's a low-efficiency shot. But having that shot in your arsenal, as an option, cannot hurt your effectiveness; it can only help. If you overuse it, it will help less. You can't replay the season to see if you'd have done better by never shooting it. You probably shot it when it felt comfortable, to avoid the shot-clock expiring, etc.
A low-% shot is clearly better than no shot at all. Sometimes there's a better alternative, and sometimes there's not. A weak player is better than a weaker player. Someone else, given more minutes, may suddenly blossom. I don't see anyone on that '07 Charlotte roster who, even in retrospect, was going to do that.
I named all the Bobcats' players down to 1100 minutes; after that, Jake Voskuhl, Primoz Brezec, Alan Anderson ... I mean, this is desperate territory. You haven't suggested anyone who can come up with better minutes than what Morrison offered.
So to just say, "He's so bad, he's worse than nothing", without offering any alternative that's better, doesn't seem to be especially useful advice. Unless there's some clue in there as to what's not quite right about Win Shares.
July 21st, 2009 at 9:56 pm
You only find it to be a logical contradiction because you refuse to understand the difference between "ability" and "value". Morrison performed with a certain rate of production that year, production that happened to be strongly associated with Charlotte Bobcat losses. Bernie Bickerstaff operated under the belief that Morrison's "true talent level" was a higher rate than his actual production, and gave him ample playing time in the hope that his rate of performance would eventually catch up to what Bickerstaff perceived his "true rate" to be. Bickerstaff's beliefs about Morrison's true talent fall under the auspices of "ability". Morrison's actual on-court performance falls under "value". Due to imperfect information there is a difference between the two things, and in this case Bickerstaff appears to have misjudged Morrison's fundamental ability level.
July 22nd, 2009 at 9:04 am
Mike G wrote:
I named all the Bobcats’ players down to 1100 minutes; after that, Jake Voskuhl, Primoz Brezec, Alan Anderson ... I mean, this is desperate territory. You haven’t suggested anyone who can come up with better minutes than what Morrison offered.
You left off an important name: Walter Herrmann. Herrmann shot 52.7% from the field and 46.1% from three-point range while averaging 19.5 minutes per game. If the Bobcats had swapped Herrmann's playing time with Morrison's, I believe they would have won more games.
July 22nd, 2009 at 9:57 am
I did mention Herrmann, but misidentified him as having >1100 minutes (he had 936). He was only there for 48 games, the last 17 of which were all 24-44 minute gigs. He started the final 12 games, and the team went 7-5.
Subsequent to that rookie outburst, he's had very little to commend him. I don't know about his defense, or where his shot went. The team no longer wanted him in '08, and the Pistons haven't found any use for him.
I guess I don't distinguish between value and ability (talent), because a skill that doesn't apply to winning isn't a relevant ability. A guy may shoot 96% FT, but if he can't get to the line, it's irrelevant. There's no basketball-game ability in that. It's like saying the ability to spin a ball on your fingertip has some potential; but it doesn't.
Anyway, it's generally good to stick to your guns, and put 'the power of language' to your chosen stats. Sometimes that will keep the door open for the stats' evolution.
July 23rd, 2009 at 12:09 pm
Hey Neil, this is completely unrelated, but I had nowhere else to ask... But can you possibly do a study on what are statistically the greatest defensive teams of all time and the greatest offensive teams of all time?
I'd be very interested in seeing how today's era matches up with that of yesteryear.
July 24th, 2009 at 11:19 am
Neil,
Interesting results with Parkhill and Morrison. I was even more intrigued that Mikan and Chamberlain finished 1-2 in WS/minute. Mikan's dominance is so underrated! Thanks for explaining your methodology of calculating win shares for the players from the past. Anytime you want to share your results for the 1614 players you have analyzed or even the best 100 or so, I think I speak for everyone when I say, I'd love to see them.
July 28th, 2009 at 5:58 am
I am tired of seeing sports-statisticians using OLS as if it were the ultimate tool for serious analysis. In fact is really flawed since it fails to account for what it is called endogeneity. Notice that no matter how many variables you actually observe, there are many others that are not observable/measurable. That is why a guy who has never seen playing a guy cannot tell whether he was good, very good or great by his statistics. Whenever those unobservables (say, intangibles) are correlated with other performance measures, OLS fails to be "consistent" - that is, produces bad estimates. Look at your rankings in HOF probability. Manu Ginobili shows up with a tiny 5% which may well decline over the years. To me, Manu Ginobili merits more consideration since he was key in 2/3 rings and is a superclutch player. He chose to be sixth-man and reduced minutes for the good of his team. Being a good teammate makes you worse statistically but it is positively correlated with performance. Starbury or Baron Davis have, accordingly much higher probability - and none of them stands a chance in my opinion. Other methods like GMM are more suited and statistical analysis should improve in that way