Great/Bad Player Games and W-L Record

by wjb1492 on Apr 14, 2009 2:08 PM CDT Comment 21 comments

Really not sure what to call this, honestly.

I've been having this discussion about how consistent (or inconsistent) the Bulls players have been this year, how often certain players show up or fail to show up, and what impact this all has had on the W-L record for the year. This eventually reminded me that we've had a similar conversation at some point in the past, using player averages and standard deviations as a measure of consistency. So, I've wasted the past hour or two compiling some data for the Bulls to see what shows up.

Right now this is all really just descriptive, but I found it interesting so I'm assuming at least a couple other people might as well. :) I may get motivated to do some actual analysis someday.

First, I used GameScores from basketballreference.com as the per game stat. While I don't think it's necessarily the best formula for measuring contribution, it does include a wide variety of stats in a single score and was available for every game this year. I included the guys currently in the main rotation, plus Luol out of interest, but did not include Tim Thomas since he's played very little. These are the average gamescores and standard deviations for the year (Bulls games only) for each player:

Ben 13.59 (7.68) Luol 10.70 (6.50)

Derrick 12.74 (6.60) Tyrus 9.10 (6.57)

Kirk 8.04 (5.99) Joakim 7.85 (5.51)

John 13.50 (6.51) Brad 10.97 (5.10)

So my first thought was that either this didn't work so well as a measure of consistency, or everyone on the team is pretty inconsistent. I may go back and calculate the scores for some of the players with reputations for consistency to see whether it looks substantially different. I wasn't all that surprised for this team, though - we've seen guys bounce between good games and bad all year. I do think using a single stat (eg, points for Ben) might show more consistency for some of the guys, but over all stats categories it has been a pretty crazy year.

While starting with fairly large standard deviations gave the guys a wide range for the "average game" category, I was OK with that. My perception is that the guys on the team have all played reasonably well this year with the exception of a handful of games. The extreme outliers (+/- 2 s.d.) were kind of interesting: Derrick had the most hideous games at 3, while Tyrys and Brad had none. Everyone else had 1 on the season. On the spectacular game side, Derrick and Luol had none, Tyrus, John and Brad 1 each, Kirk and Joakim 2 each, and Ben 3. But remember, this is just in comparison to their individual season averages, not on some sort of objective scale.

And here is how this all shows up in wins and losses:

"excellent" game W-L "average" game W-L "bad" game W-L

Derrick 7-5 32-26 1-9

Ben 9-5 28-25 4-10

Kirk 5-2 22-15 1-5

John 2-2 12-7 1-1

Luol 4-5 16-16 3-5

Tyrus 8-5 24-24 7-10

Joakim 12-3 24-30 3-7

Brad 4-0 9-8 3-2

(No surprise that Derrick failing to show up would cause the team the most problems!)

One thing that massively stood out was how great Tyrus and Joakim have been compared to the beginning of the season. Both had the vast majority of their "bad" games early on and a whole lot of "excellent" games in the second half of the season.

I still want to go through and look at whether the guys show up or not against good vs. bad teams, but I didn't have time this morning. Also, I'm planning to look at combinations of good and bad games by player pairs. If anyone has any other thoughts of what to do with this, jump in.

FanPosts are user-created posts from the BlogABull community, and are to be treated as the opinions and views of that particular user, not that of the blogger or blog community as a whole.

14 recs | Comment 21 comments

Story-email Email Printer Print

Comments

Wow, Luol doesn't

change the game one way or the other!

"Worker bees can leave.
Even drones can fly away.
The Queen is their slave."

by jpchi on Apr 14, 2009 3:01 PM CDT reply actions 0 recs

Good idea for a project.

Rec’d as lots of food for thought.

If I’m understanding your chart correctly, it seems that Noah is our biggest positive wild card — when he has an “excellent” game the chances of us winning are damn good. If memory serves, however, he only has “excellent” games against certain types of teams (or probably more accurately: he fails to have “excellent” games against certain types of teams, namely, teams with a hefty low-post presence). That’s just based on memory and not any in-depth analysis and I would love to be refuted on that point.

by arjoseph on Apr 14, 2009 3:11 PM CDT reply actions 0 recs

Yeah, there's a lot more to be done to add more context to this

I just ran out of time this morning. I want to do another part getting into which teams each guy performs well/poorly against, as well as home vs. road. Also, I think I need to look more at what aspect of the boxscore is bringing the gamescore down or up.

Given how incredibly different Ty and Joakim’s seasons look now compared to earlier, I’m thinking about going back and just doing a half season.

Man-slave, bring me my PB&J!

by wjb1492 on Apr 14, 2009 3:46 PM CDT up reply actions 0 recs

That looks like a HUUGE amount of work there

Nice job. It got me thinking about why scorers are so valued in the NBA. While a Jason Kidd can give you quality offense over an entire season with his passing, his ability to change things on a game-to-game basis is limited. On the other hand, APM numbers show Ben Gordon gives us less offense on average than Kidd, but Gordon can single-handedly win or lose games for us with his usage and efficiency.

I wrote an article about the go-to guys on championship teams, and what stood out to me what how abnormally amazing all these players were at passing. The article lacked a lot of support, and so I just came up with a few theories on why that is and left it at that, but now I think I figured it out…

Game Score really undervalues the assist. Hollinger has an assist counting for 0.7, when we should really be measuring ability to pass. Whether Jason Kidd gets 14 assists per game or 7, his ability-to-pass remains the same; either the defense was worse or his teammates just knocked down more shots on that particular day, but assist or not he creates offense on every possession with his passing. I’d love to see a ‘Truer’ Game Score (TGS) that uses AST% instead of assists.

But what that means is players who can pass well will have lower standard deviations of TGS than players who can’t. And so while a player like Kevin Durant would have a sd that’s nearly 60% (similar to Ben Gordon’s 7.5 sd with a 13.5 average), a Dwyane Wade, I’d guess, would have an sd much lower. And so it’s that much harder for a Kevin Durant team to compete with a Wade team over 7 games. Durant can cost his team 2-3 games with bad shooting performances, while a Wade can neutralize a bad shooting day with good passing.

by YaoPau on Apr 14, 2009 3:18 PM CDT reply actions 4 recs

This doesn't really relate to what you're saying, but kind of.

Since we’re talking about the inadequacy of the assist as a stat, I’ll throw another idea on the heap: “ability to pass,” or whatever, should also take into account good passes that don’t immediately precede a shot attempt. Sometimes the best pass is the penultimate one, the one that starts the defense rotating and leads to another pass (and sometimes another) that eventually finds an open man in a high percentage zone. It’s a common mistake of ball-dominant guys to think they either need to take a shot or make an amazing assist a la J-Kidd or Magic Johnson. Unfortunately, almost nobody can pass as well as those guys. Kirk seems to have this problem sometimes. While this will be less interesting in terms of your go-to-guy project (since you want those guys either taking a shot or making a one-pass assist), I think it would be very helpful in a more hollistic, team-oriented, what-impacts-winning analysis (like wjb1492 has undertaken).

by arjoseph on Apr 14, 2009 3:45 PM CDT up reply actions 1 recs

I completely agree

You said what I was trying to say with my whole “it doesn’t matter if Kidd gets 14 assists or 7” thing. Kidd helps the Mavs offense regardless of whether he gets the assist because he makes passes every possession – he’s helping their offense regardless of whether they score. Ben mostly just shoots.

by YaoPau on Apr 14, 2009 3:58 PM CDT up reply actions 1 recs

I think it'd be awesome

You always hear about all the really cool stats teams are keeping that just aren’t released to us, and I’d love to get my hands on stuff like that. I just don’t necessarily want to spend the time compiling it myself. ;)

Man-slave, bring me my PB&J!

by wjb1492 on Apr 14, 2009 4:14 PM CDT up reply actions 0 recs

Hey, if it were any good, you might get hired right away.

Viva la nuance! Reading comprehension rules!!!

by tyger1147 on Apr 14, 2009 4:58 PM CDT up reply actions 0 recs

I really did debate about using gamescore

I didn’t want to spend forever trying to come up with my own formula to combine a single game’s stats into one number, particularly when I wasn’t sure how well this was going to turn out. In general, I really don’t like evaluating a game based solely on stats. It’s one of the reasons I always seem to end up defending +/-, even though I know it has extremely limited use and needs to be contextualized.

In this case, I decided I could live with it just to save the effort of compiling my own stats. Besides assists, I don’t really like that the formula includes fouls, either, which are sometimes a very good thing but treated as uniformly bad in this formula. I decided to go with the fiction that all of this will be roughly even over the course of the season, as far as determining which games are significantly better or worse than the player’s average. I figured this was better than just going the old pts+asst+reb formula, since at least it takes shooting efficiency into account.

I’d be totally up for using a different formula, but I’m not too creative at coming up with stuff like that – better at critiquing what’s already there than contributing something original! But if you’ve got some ideas, I’d be thrilled to go with something that better captures a player’s overall game.

Man-slave, bring me my PB&J!

by wjb1492 on Apr 14, 2009 4:04 PM CDT up reply actions 0 recs

I don't think Game Score is a bad choice for something like this.

It works perfectly fine for something like this. Especially when you’re just separating good/average/bad games for individual players. Any of the linear weights formulas out there would do that reasonably well with similar results, and Game Score is probably one of the better ones. Thanks for putting this together.

by Scotter on Apr 14, 2009 7:06 PM CDT up reply actions 0 recs

Rec - great job!

by chgobr on Apr 14, 2009 7:43 PM CDT reply actions 0 recs

Joakim as the guy with the lowest standard deviation

of the guys that have been here all year doesn’t surprise me at all. He’s been our most consistent player from game to game. His defense has been less consistent than his offensive numbers, but this wouldn’t be measured using GameScore. He’s getting better and better on defense though. The whole team seems to have looked better on defense the last four or five games. They’ve had a bit of bum luck as they have forced teams to take bad bad shots and for a few games teams were just hitting those tough shots. They really locked down Detroit last night though.

Fire Vinny.

by fundamentallysound on Apr 14, 2009 9:44 PM CDT reply actions 0 recs

Ehh, he has the lowest standard deviation

Because he has the lowest average. His standard deviation is 70% of his average. To me that means you don’t know what you’re going to get out of him, game-score wise at least.

by YaoPau on Apr 14, 2009 10:02 PM CDT up reply actions 0 recs

Oh right, that's true. I hadn't even thought of that aspect of it (the % of average s.d.)

shows what I know.

Fire Vinny.

by fundamentallysound on Apr 14, 2009 10:04 PM CDT up reply actions 0 recs

I think he's really affected

by the horrible first half/good second half part of this, as well as Tyrus.
Would be interesting to run the numbers in halves: pre- and post- All Star break.

Joakim Noah: Better than you.

by Prevenge on Apr 14, 2009 11:33 PM CDT up reply actions 0 recs

I followed wjb1492's lead

and looked at game-by-game deviations, except instead of using Game Score I used statistical +/- (APM) and held AST% and OREB% constant, so only TS%, USG%, and TOV% fluctuated from game to game.

Before January 25th, Joakim’s stats:

19.4min, 4.8pts, 6.5reb, 0.9ast, 0.6stl, 1.3blk, 0.9tov, 3.0pf, 0.527 TS%, 19.3 TOV%, 12.67 USG%
PER36: 8.8pts, 12.0reb, 1.6ast, 1.1stl, 2.4blk, 1.7tov, 5.6pf, 0.527 TS%, 19.3 TOV% 12.67 USG%
Estimated Average Offensive Statistical +/-: -3.41
Standard Deviation: 11.77

Since January 25th, Joakim’s stats:

30.3mpg, 8.9pts, 9.2reb, 1.9ast, 0.6stl, 1.6blk, 1.2tov, 3.0pf, 0.628 TS%, 15.6 TOV%, 12.3 USG%
PER36: 10.6pts, 10.9reb, 2.3ast, 0.8stl, 1.9blk, 1.4tov, 3.6pf, 0.628 TS%,15.6 TOV%, 12.3 USG%
Estimated Average Offensive Statistical +/-: +1.05
Standard Deviation: 6.51

Surprisingly, Noah’s rebounds, blocks and steals have dropped since January 25th, but look at what he’s gained. His TOV% is 4% better, his PF36 went from 5.6 to 3.6 allowing him to stay on the floor longer, and his TS% went from crappy to one of the league’s best. But my favorite part is his jump from 1.6ast36 to 2.3ast36. Here’s a guy who averaged an impressive 2.3apg in 25.9mpg his junior year at Florida, and (thanks to Brad Miller?) now it looks like he’s finally brought that ability to the NBA. That rate of 2.3 assists per 36 would place Noah 6th in the NBA among centers (Miller, Horford, Hawes, Bogut, Biedrins) and it makes me think he can soon develop into one of the league’s top offensive centers, too.

It all shows up in the Statistical APM. Offensively, Noah went from -3.41 on average the first few months to +1.05 since. Fundamentallysound listed Noah’s defensive statistical APM at +3.12, meaning we can estimate that Noah’s been around +4 overall since January 25th. That +4 would’ve put Noah around the 35th best player in the NBA according to statistical APM last year.

by YaoPau on Apr 15, 2009 10:01 AM CDT up reply actions 1 recs

interesting work as always, YaoPau.

Well done.

Fire Vinny.

by fundamentallysound on Apr 15, 2009 11:59 AM CDT up reply actions 0 recs

Cool, thanks.

Joakim Noah: Better than you.

by Prevenge on Apr 15, 2009 2:41 PM CDT up reply actions 0 recs

Now to whom would you give credit?

Vinny???

by hlac on Apr 16, 2009 10:55 AM CDT up reply actions 0 recs

Joakim

look at his numbers from March and April last season, he was great, so it’s not like he’s playing above his head. Vinny deserves part of the credit for giving him minutes.

by YaoPau on Apr 16, 2009 4:45 PM CDT up reply actions 0 recs

another interesting thing that you can see is that there is a rough correlation

between low standard deviations in GameScore and rank by Adjusted Plus Minus. Brad is the lowest s.d. guy on the team and he’s number 1 on the team in APM. Joakim is second lowest s.d. and he’s number 2 on the team in APM. Kirk is third on the team in APM and he’s third lowest in s.d. It seems that if you’re consistent that’s a pretty good thing for your APM numbers which seems to make some sense. Good players are good regardless of matchups, whereas less good players will be more prone to wild swings depending on matchups.

Fire Vinny.

by fundamentallysound on Apr 14, 2009 9:51 PM CDT reply actions 0 recs