So, until I figure out what the current problem on my Box Score Analysis data sheet is (don’t worry though, it’s a problem that only affects the next stage of the analysis), the team-by-team analysis is on hold.
To summarize, Christopher Reina pointed out that the Lakers this regular season were 26-18 when Kobe shot more than 20 shots, and 31-7 when he shot 19 shots or less. Ball Don’t Lie, in turn, calls that study way too short-sighted, limited and confounded to be considered a viable conclusion.
The truth is, the study is certainly not conclusive enough to make the statement ‘the Lakers are better when Kobe doesn’t shoot as much’. That’d just be ridiculous. Not only is the criteria used here too limited (as Ball Don’t Lie points out, the very nature of a ‘cut-off’ point between completely opposite categories is insufficient), attributing causality based on it is ludicrous.
Here, we want to make it not-so-ludicrous. We want to answer two questions:
Is it true that the Lakers are more likely to win when Kobe’s shot attempts are lower?
If so, does Kobe’s “high-volume shooting” cause the Lakers to under-perform, or does another variable cause both the Lakers’ under-performance and Kobe’s high-volume shooting?
So to start out, we want to answer that first question better than the original study did. We’ll consider two different ways to do this: Overall Averages and Overall Correlation.
I’m really not sure why the original study opted for using a cut-off point to judge whether Bryant’s shooting matched up with the Lakers’ win/loss record when there’s a much more logical (and equally simple) way to do it – but hey, to each his own.
Let’s consider something a bit more informative: Kobe Bryant’s averages in wins and losses.
FGA in Wins: 19.32
FGA in Losses: 23.56
Well, that decides it, right? Kobe averages over 4 more shots in losses than in wins. So it has to be true, right?
Not quite. Why not? We’ll get to that a bit later. This just shows how misleading a very cursory glance at the stats can be.
But as long as we’re here, let’s look at other differences between Kobe’s stats in wins and losses – just for curiosity sake, right?
- Kobe averaged .9 more assists in wins than in losses – which, depending on your point of view, means either he actually gives his teammates opportunities in wins, or they actually hit on the opportunities he gives them in wins.
- Kobe averaged 1.1 more 3-point attempts in losses, but made about the same number regardless of the game’s outcome.
- Kobe shoots 1.7 more free throws in losses than in wins.
- And, though it’s pretty irrelevant, Kobe plays 4 more minutes/game in losses than in wins.
Hopefully you picked up on the sarcasm on the ‘irrelevant’ part of that last bullet. You mean the more minutes Kobe plays, the more shots he takes? Extraordinary!
Let’s not get too excited though (or depressed, if you’re hoping I show Kobe really does over-shoot in losses) – that difference only accounts for about half of that four-shot discrepancy.
So why does Kobe average less minutes/game in wins? Consider that same ‘special case’ we considered in the last analysis: blowouts. 18 times this year, the Lakers blew out (>12 point win) their opposition as Kobe played less than his average number of minutes. Those are games when it’s only natural for Kobe to shoot less – not because it actually affects the team’s performance, but because he doesn’t play as much.
Throwing out those blowouts from consideration (it’d be more statistically sound to look at Kobe’s stats through three quarters of each game, but I haven’t found a database that has that easily parseable), we arrive at some different statistics: 41 minutes/game in both wins and losses, 21.44 FGA in wins, 23.56 FGA in losses.
But, it would only be fair to throw out those blowout wins if we also threw out blowout losses when Kobe also played less than his average number of minutes. But, that happened only once – Lakers vs. Utah, November 30th. Throwing this game out barely impacts the stats (+.19 FGA in losses, bringing the difference to 2.31).
Not considering blowout games, Kobe averages 21.44 FGA in wins and 23.75 in losses. That’s a huge difference from the 19.32 and 23.56 we considered initially. Right away we blow away that silly 20-shot cut-off used in the original study, since the only time Kobe shoots less than 20 shots (on average) is when the Lakers are in a blowout win.
But that 2.31 difference in FGA is still notable – not nearly as notable, but notable. So, considering what we know now, let’s revisit the method used by the original study of cut-off points at different shot attempts – but, let’s do so far more thoroughly.
Unfortunately, this data isn’t exactly conducive to a real correlation analysis – the sample sizes for each number of shots are far too small to really establish a correlation between FGA and win percentage.
So instead, let’s go back to the original approach (the cut-off point), but instead of arbitrarily choosing one (or worse, choosing one specifically to prove a certain point), let’s look at every number of FGA and how the more/less correlates to wins/losses – and, like above, we aren’t considering blowout wins/losses, given how they confound his statistics.
Doing so reveals an interesting division, far more informative than the original 19-or-less/20-or-more split. The Lakers were 9 and 1 when Kobe shot 17 or less shots. Above that, they were 28-19 with him taking 18 to 28 shots; and finally, 2 and 4 with him taking 29 or more. Note these divisions aren’t arbitrary, they’re chosen from what appear to be the critical points where the Lakers’ fortunes change – naturally there is likely a more smooth progression, but the small sample size does not allow us to establish one.
These observations, unlike the earlier portion of the analysis, do not debunk the idea that Kobe’s increased shot attempts correlate to the Lakers’ winning percentage dropping – rather, they seem to suggest it. The Lakers are best-off when Kobe takes relatively few shots (winning percentage of 90% when he takes 17 or less, 100% for 15 or less), substantially less-well-off when he takes a medium number of shots, and downright bad when he takes a large number of shots.
LITTLE WHITE TAKEAWAYS
This study is trying to find if there’s any truth to the idea that Kobe Bryant handicaps the Lakers sometimes by taking too many shots. In order to analyze this idea, we want to examine two questions: is there actually a connection between Kobe’s shot attempts and the Lakers’ winning percentage, and does Kobe’s overshooting cause the losses or does something else cause both his overshooting and the Lakers’ losing?
This first portion addresses that first question. One study revealed that the Lakers were substantially more likely to win when Kobe shot less than 20 shots than when he shot more than 20; but this study had several issues, so we wanted to more thoroughly analyze it.
What we found is notable: while it is true that the Lakers were more likely to win when Kobe shot less, it wasn’t nearly as severe as the original study suggested. The initial study suggested that the Lakers were 38% better when Kobe took 20 shots or less, but it would be more accurate to say the Lakers are 11% better under those conditions. A cursory glance at the statistics would show that Kobe averaged four more shot attempts per game in losses than wins, but it would be more accurate to say that he only averages two more shot attempts in losses.
The reason for this discrepancy is that the Lakers had 18 blowout wins and 1 blowout loss in which Kobe played less than his average number of minutes, suggesting that his playtime was altered by the nature of the game. It’s obviously more likely for Kobe to take less shots in games where he plays less minutes, so these games should not be considered on the same level.
So, in short, there’s some truth to the idea that Kobe shoots more in losses, but it certainly isn’t as severe a problem as the statistics would immediately suggest – in fact, it’s only a minor alteration when blowouts aren’t considered.
In order to more fully address these questions, though, it would be better to instead utilize a shots-per-minute statistic than to simply throw out games under certain conditions. I’ll look into this to see if it represents any major alterations to the conclusions reached here – if it does, I’ll write about that, but I don’t anticipate that it will.
Given that there’s still a discrepancy between Kobe’s shots in wins and losses (though it is less notable than originally anticipated), it would still be useful to see where the causality lies: does Kobe over-shooting cause his team to under-perform, or does something else – for example, teammates under-performing – cause both Kobe’s extra shooting and the Lakers’ eventual loss?
This is difficult to analyze numerically, so instead, we’ll likely resort to a few case studies of the games in question. But, that’s a job for next time.