NBA, Winning Ugly, and 1984

[Latrinsorm]

I touched on this in this season's NBA thread, but I finished the analysis out to 1984 (the first year of 16 playoff teams and, conveniently, a 30 year sample) and there's kind of a lot to get through, so I thought I'd give it its own thread back.

.

Do close games tell us anything? Does ignoring close games tell us more?

.

We can take every playoff series (16 * 30 = 480) and predict the outcome based on three things: each team's record in close (decided by 5 or less points) games, each team's record in all but close games, and each team's overall record. If there is a tie, we count that as 0 correct for 0 attempted. When we do this, we get:

258 of 439: 58.8% ± 4.7%
351 of 449: 78.2% ± 3.9%
328 of 431: 76.1% ± 4.1%

The first bin is clearly the worst, although it is better than guessing at random. The second and third bins are too close to call. Of course, from 1984 to 2002 the first round was a best of 5, does that change anything? Here are its values:

86 of 149: 57.7% ± 8.1%
119 of 152: 78.3% ± 6.7%
110 of 145: 75.9% ± 7.1%

Same story.

.

Alright, what if we look just at the champion of each game type going into each postseason, and see how far they get? The grit and grind king, the flashy frontrunners, and the overall #1 seed? For championships, we get:

4 of 29: 13.79% ± 12.81%
14 of 29: 48.28% ± 18.56%
11 of 26: 42.31% ± 19.38%

Just like before, the grit and grind king model is statistically indistinguishable from picking any of the 16 teams at random (6.25%). But even more interesting is to consider cases where the three models disagree:

{Grit and grind} and not {frontrunner}: 0 of 25.
{Grit and grind} and not {overall #1}: 0 of 25.

Looking at the best record in close games tells us literally nothing.

{Frontrunner} and {overall #1}: 11 of 19: 57.89% ± 22.65% - pretty good!
On the 6 occasions those two models disagreed, the {frontrunner} was the champion twice, and the other four times they were both wrong: 85 Lakers, 02 Lakers, 07 Spurs, 11 Mavs. The only common thread I can think of there is aging stars take the regular season easy, then turn it on in the playoffs...? Regardless.

.

Another way we can look at each model's performance is to take their prediction and see how far it got on a continuum, rather than second place = first loser. The most straightforward way to do this is to assign the highest point reached a number as follows: first round = 1, second round = 2, conference finals = 3, finals = 4, champion = 5. When we do this, we get...

{Grit} = 2.48 (halfway to the conference finals)
{Flash} = 3.76 (damn near the Finals)
{Overall} = 3.69 (pretty much the same)

.

The last thing we can look at is okay, maybe champions don't do any better at winning close games, but do they do any better at not playing them in the first place? Here are the %s for the entire NBA for >15, 11-15, 6-10, 1-5:

23.6% ± 0.4%
18.9% ± 0.4%
29.2% ± 0.5%
28.3% ± 0.5%

And here they are for NBA champions of the past 30 years:

26.9% ± 1.8%
19.6% ± 1.6%
27.9% ± 1.8%
25.6% ± 1.8%

Just barely, yes! The floor for the NBA is 27.8%, the ceiling for champions is 27.4%. The average champion has more blowouts (and wins way more of them at 86% ± 3%) and less nailbiters (and wins slightly more of them at 61% ± 4%).

.

Tomorrow, we see how Jordan did. :)

[Latrinsorm]

First of all, I meant to say this in the first post but shout out to Stef for mentioning vlookup in passing once. That has made all the difference.

.

Jordan! Famously(/infamously) competitive, widely acclaimed as the best player of all time, almost universally acclaimed as the most clutch player of all time. Indeed, the google string "more clutch than jordan" returns a mere 37 results, "more clutch than michael jordan" only 47 (and I hadn't heard this before, but apparently someone in the Obama presidential campaign remarked that of the then-Senator). He even famously won the NCAA Tournament with a last-second shot, giving him invaluable seasoning over today's preps-to-pros pups... right?

Well, let's start with an exercise for the reader. Here is a list of Jordan's 11 full Chicago seasons (so 1986 and 1995 are out) in randomized order. Can you identify the 6 championship years?


A 15 of 30 50.0%
B 15 of 37 40.5%
C 11 of 17 64.7%
D 19 of 26 73.1%
E 19 of 32 59.4%
F 15 of 28 53.6%
G 12 of 16 75.0%
H 12 of 22 54.5%
I 16 of 26 61.5%
J 16 of 23 69.6%
K 10 of 20 50.0%
...and, time! The correct answers are K, H, A, G, J, C. Merely picking the 6 highest close winning %s would give you 3 out of 6 correct answers, which is of course not terribly impressive when 6 of 11 are correct.

Now, let us look at the numbers for all margins of Jordan's Chicago career (again, only his 11 full seasons). The top entry is >15 points, then 11-15, 6-10, and finally 1-5.


193 of 229 84.3% ± 4.8%
121 of 166 72.9% ± 6.9%
144 of 230 62.6% ± 6.4%
160 of 277 57.8% ± 5.9%Surprise! Jordan was at his worst when it came to winning close games, barely distinguishable from league average (50% by definition). If we look at only the six championship years, we get:


143 of 155 92.3% ± 4.3%
83 of 94 88.3% ± 6.6%
86 of 115 74.8% ± 8.1%
76 of 128 59.4% ± 8.7%Surprise! Still barely distinguishable from league average at close games.

.

Kobe.


226 of 316 71.5% ± 5.1%
174 of 264 65.9% ± 5.8%
261 of 404 64.6% ± 4.8%
217 of 362 59.9% ± 5.2%Kobe without Shaq.


108 of 155 69.7% ± 7.4%
88 of 142 62.0% ± 8.1%
129 of 220 58.6% ± 6.6%
118 of 205 57.6% ± 6.9%

.

LeBron.


140 of 196 71.4% ± 6.5%
108 of 155 69.7% ± 7.4%
147 of 239 61.5% ± 6.3%
124 of 214 57.9% ± 6.7%Lebron without Wade.


85 of 131 64.9% ± 8.3%
68 of 105 64.8% ± 9.3%
102 of 177 57.6% ± 7.4%
94 of 161 58.4% ± 7.8%

Duncan.


283 of 341 83.0% ± 4.1%
187 of 238 78.6% ± 5.3%
237 of 358 66.2% ± 5.0%
181 of 327 55.4% ± 5.5%
.

Nobody dominates close games. Nobody. Three of the most lauded players of the last 20 years managed "distinguishably better from league average, barely", a fourth didn't even manage that. These are the facts.