Statistical Analysis. Humor. Knicks.

Wednesday, October 22, 2014

return on investment – allan houston….

mike asked me to write a column or two for his knick blogger – while he was away sunning himself on some remote beach – concerning anything about pro basketball. i’ve never been a knicks fan (being from cleveland, ohio), and the closest i’ve ever gotten to a knick was to play pickup ball against long time knick charles oakley many moons ago (who went to high school here in NE ohio and who stuffed many of my jumpers). but i do have a unique perspective on the nba in that i analyze the game through statistical analysis using computer simulation of nba games, and would like to do so here for something concerning the knicks…

its not hard to get the computer to play basketball – the key is just to get it to do so accurately. but after watching many tapes of games, charting things the league never kept track of (this was over a decade and a half ago and before www.82games.com, whom i wish was around back in 1990), and fiddling around with the numbers, after 15+ years i think i’ve got it down. the key is to derive from the stats how to rate players for how often they handle the ball on offense, and then to rate them for how often they shoot, pass, get fouled, and turn the ball over per time they do handle the ball on offense. the first i call a player’s possession factor (his touches/minute), the second his player attributes. once you can do that getting the computer to simulate the actual playing of a game, i.e. re-creating every shot, pass, rebound, assist, steal, turnover, blocked shot, etc, is relatively easy, and to get the computer to play hundreds or thousands of games takes little time. rating player defensively was alway tough, but thanks to www.82games.com, that part of the process has been made much easier too…

because i look at the league differently than most people do, i often feel i have an advantage over those who do not use computer simulation, in particular those in the league responsible for signing players, especially to mega-dollar long term contracts. in hindsight we can only shake our heads at the contracts given to players like stanley roberts and calvin booth, players given big money but who had played little time in the league prior to their contracts being signed, and wonder what were those teams thinking. but many players in the league have gotten long term contracts at top dollar, and the question is are they worth it?…

one contract i have always wondered about was the one given to allan houston. in july of 2001, after having played five seasons for the knicks, new york gave him a six year deal worth a staggering $100,000,000, and while i’m not sure i’m guessing its all guaranteed. that works out to about $16-$17 million per season, one of the highest annual salaries of any player at the time the contract was signed (and still one of the highest today), and double what he signed for with the knicks after coming over from the pistons following the 95-96 season. the question is – is he worth it?…

the neat thing about computer simulation is that you can take any player and place him on any team, and run as many games as needed to determine if that team is better or worse with that player. its ideal for running “what if” scenarios – like what if the knicks has shaq at C instead of mutombo/mohammed? but its also ideal for determining just how good your players are in relation to other players from around the league – by simply putting other players on your team to replace a certain player and simulating hundreds or thousands of games to see if the team is better or worse….

here are the knicks players from last season (03-04) and their minutes played:

                          default
sub
min min/82 pattern min/g
n.mohammed 1611 20 20 20
k.thomas 2548 31 32 32
t.thomas 2088 25 24 24
a.houston 1799 22 24 40
s.marbury 3254 40 40 40
d.mutombo 1494 18 20 20
o.harrington 872 11 12 12
s.anderson 1947 24 24 12
p.hardaway 2095 26 24 24
m.norris 847 10 8 4
f.williams 714 9 8 8
m.sweetney 494 6 4 4
240 240

the 1st column is simply each player’s minutes played, the 2nd column their minutes played divided by 82 games. the computer simulation model can only substitute for players in increments of 4 minutes, so to get a default substitution pattern for the team i chose the multiple of 4 closest to their actual minutes played divided by 82 games (the 3rd column). do this for each team in the league and you can play entire seasons by computer to re-create what happened in the actual real-life season…

in 03-04 the knicks went 39-43, and based on their statistics they should have gone only 37-45. the above substitution pattern (the default), when 8200 games (100 seasons) are simulated on the computer, averages a W-L record of 38-44. to gauge the maximum impact of a player on a team i typically look at how the team does when that player plays 40 min/g, which is about the maximum playing time per game the best players play in any single season….

using the substitution pattern in the 4th column, i played allan houston 40 min/g, taking minutes away from shandon anderson and moochie norris. however when houston played 40 min/g the knicks W-L record increased just another 2 games over an average 82 games to a W-L record of 40-42. may not sound like much of an improvement for one of your team’s best players to play an additional 16 min/g, but keep in mind in 02-03 allan houston actually played all 82 games and 38 min/g, playing what was probably his best season statistically for the knicks, and the team went just 37-45….

better yet why not put that 02-03 allan houston on the 03-04 knicks and play 8200 simulated games to see how much better they play? i did just that and the team’s W-L record improves, but by just another half a game over an average 82 season, to 40.5-41.5. thus houston’s best season statistically of 02-03 wasn’t much better than his 03-04 campaign, just half a game over the span of an entire season…

so let’s see if the $16 million/yr man is worth his salary. the best way to judge that IMHO is to “trade” a number of SGs from around the league to the 03-04 knicks, replacing allan houston for 40 min/g at SG, and playing enough simulated games (in this case 8200, or 100 simulated seasons) to see if they improve the team’s W-L record or not, moreso than what houston does. typically i’ll try this with some of the best players in the league at that specific position (in this case SG), and also some of the worst, to see who the player in question (houston) plays like more. three of the best SGs in 03-04 were kobe bryant, tracy mcgrady, and ray allen, a few of the “worst” statistically were david wesley, dion glover, kendall gill, and dajuan wagner, and i also ran vince carter and lebron james for comparison. here are the results:

            average          scoring
player W-L pts/g FG% reb/g ast/g st/g to/g bs/g touches/min
mcgrady 02-03 53-29 29.7 .553 6.4 4.9 1.6 2.3 0.8 1.7
mcgrady 03-04 45.5-36.5 26.6 .517 5.9 5.2 1.4 2.5 0.6 1.6
bryant 02-03 51-31 26.7 .540 6.4 5.1 2.1 3.1 0.8 1.6
bryant 03-04 49.5-32.5 24.7 .539 5.7 5.1 1.8 2.6 0.4 1.5
r.allen 03-04 45.5-36.5 21.8 .558 5.3 4.6 1.3 2.7 0.2 1.4
carter 03-04 42.5-39.5 22.5 .494 5.0 4.7 1.2 3.0 0.9 1.5
l.james 03-04 41-41 20.4 .480 5.6 5.7 1.6 3.3 0.8 1.6
houston 02-03 40.5-41.5 21.9 .555 3.0 2.6 0.7 2.1 0.1 1.0
houston 03-04 40-42 19.7 .533 2.8 2.1 0.8 2.2 0.0 0.9
glover 03-04 38-44 15.0 .444 6.2 3.1 1.2 2.4 0.5 1.0
wesley 03-04 37-45 16.8 .478 2.8 3.4 1.4 1.9 0.3 1.1
k.gill 03-04 36.5-45.5 14.6 .438 5.3 2.4 1.7 2.2 0.4 0.9
wagner 03-04 33-49 14.5 .436 3.1 2.6 1.3 2.0 0.4 1.0

scoring FG% is simply (2pters + 1.5x3pters + FTM/2)/(FGA + FTA/2)…

i also ran kobe bryant’s and tracy mcgrady’s 02-03 stats because while they were the two best SGs in the league in 03-04, they were also each better statistically in 02-03 than they were in 03-04…

as you can see the difference between the best and worst players at a single position can be upwards of 15 to 20 games in a single 82 game season, in this case tracy mcgrady’s stellar 02-03 campaign (where he led the league in scoring with 32 pts/g on a very good scoring FG% of .553), compared to some of the league’s worst SGs. but what is also evident is that allan houston’s performance statistically is just a few games better over an 82 game season than that of some of the league’s worst SGs, and far worse than the league’s best SGs….

now don’t think that being just a few games better than some of the worst players at your position is insignificant – if all of the starters on your team were 4-5 games better than the worst players at your position, that means your team would theoretically be about 20-25 games better than the worst team in the league. if the worst team in the league wins 15-20 games out of 82, that means your team would win somewhere between 35-45 games (which btw is what the knicks won in both 02-03 and 03-04). but again its nowhere near as good as being 10-15 games better than the worst players at your position…

when allan houston signed his huge contract extension after the 2000-01 season, he had played for the knicks for 5 seasons, yet in those 5 seasons he averaged less than 18 pts/g (less than 19 pts/g in the playoffs) and never averaged as much as 20 pts/g in any single season, was an average defender (at best), was a poor offensive and average defensive rebounder, a player who got few steals and few blocks, and a shooter who hit a good 40% of his 3pters but just 46% of his 2pters (less than the league average of .468 on 2pters during those 5 seasons). his scoring FG% during that 5 year span was .533, which is good, but which is less than 2 percentage points above the league average of .515 during that same time period. to his credit he missed only 5 games out of the knicks 378, but is that worth $16-$17 million/yr for six years?…

notice that lebron james on new york wins just as many simulated games as allan houston does (41 to 40) but shot much worse, a low scoring FG% of just .480 (03-04 league average was .508) compared to houston’s .533. yet he rebounded twice as good, almost tripled houston’s ast/g, doubled his st/g, and blocked far more shots. bottom line is that in his 5 seasons as a knick before signing his extension allan houston was a very good shooter but contributed little else to his team…