Ed Küpfer
Joined: 30 Dec 2004 Posts: 130 Location: Toronto
|
Posted: Fri Feb 04, 2005 8:52 pm Post subject: Playoff Experience |
|
|
I posted this in the Raptors newsgroup where the question of the value ofplayoff experience came up. It's not the most rigourous study I've ever done, and among the sloppiest writing ever penned, but I thought it still might hold some interest.
I looked at every playoff series since 1974. I regressed the outcomes against the following variables using a logit model: Home court advantaged team Pythagorean record, away team Pythagorean record, series length (i.e. best-of-x), home team previous playoff experience in minutes, away team previous playoff experience. I also used weighted playoff experience instead of the last two, with minutes played in that playoff year as the coefficient.
Home team playoff experience turned out to be statistically significant at 5%, but away team experience was not significant at all. Very strange. [I've now given this quite a bit of thought, and I still have no answer.] Regression results below for posterity.
So how big an effect does playoff experience have on the probability one team beating another in a playoff series? To answer this, we'll assume two evenly matched .500 teams playing a 7-game series. If they both have 0 games playoff experience, the team with home court advantage (call them HOME) has a 61% chance of taking the series. Assume the AWAY team has 20 player-games playoff experience, which is the median number of games among AWAY teams (player-games defined here as 36 minutes played).
Code: | HOME experience Probability
in player-games of HOME win
0 58%
5 60%
10 61%
15 63%
20 64%
25 66%
30 67%
35 69%
40 70% |
So every 10 extra player games of playoff experience -- 7.5 full team games -- raises the probability of a team win by 3%. Seven and a half game is roughly a series and a half. That seems like a lot of playoff experience for such a little amount of gain.
[Note: I did the regression twice, once using straight playoff experience per team in minutes, and once using the average playoff experience per team weighted by each player's minutes. The results were very similar, so I've only shown straight playoff experience below.]
Code: |
[REGRESSION RESULTS]
Link Function: Logit
Response Information
Variable Value Count
HWin 1 313 (Event)
0 103
Total 416
Logistic Regression Table
Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant 0.537 1.311 0.41 0.682
BestOf -0.0033 0.1113 -0.03 0.976 1.00 0.80 1.24
H_Pyth 8.363 1.822 4.59 0.000 4284.83 120.57 1.52E+05
A_Pyth -8.553 1.893 -4.52 0.000 0.00 0.00 0.01
H_Exp 0.0017062 0.0007922 2.15 0.031 1.00 1.00 1.00
A_Exp -0.0007275 0.0007920 -0.92 0.358 1.00 1.00 1.00
Log-Likelihood = -200.462
Test that all slopes are zero: G = 64.729, DF = 5, P-Value = 0.000 |
_________________ ed |
|