Thursday, November 22, 2007

Coaches Better the Second Time Around?

The recent firings of Bruce Arena and Frank Yallop made me think about an old Bigsoccer thread that I once referenced on this blog, which referred to how coaches performed with the second team they coached compared with the first. I thought I'd revisit that here today.

In MLS history, there have been 13 men to take the reigns of two or more teams. Counting PPG in all competitions (with shootouts counted as draws), only twice has a coach actually done better with the second team. And even that is not as impressive as it sounds; one of the two men is Fernando Clavijo, who's had a losing record with both teams. With the Rapids, he's only been 0.02 PPG better than with the Revs and could possibly still end up worse now that he's somehow kept his job.

The other man is Dominic Kinnear, which really shouldn't count. Even he isn't too much better thanks to San Jose's 2005 season.

Two other current coaches are in their second run. Sigi Schmid is almost a certainly to end up worse with the Crew than with the Galaxy, while Mo Johnston could possibly end up better. However, that would take a few good years.

I don't think this info is too meaningful, but it's fun to look at.

The year represents the coach's first season with that team.


First team vs second team

(PPG, all competitions)

DC 1996 Arena, Bruce 1.883
NY 2006 Arena, Bruce 1.250




CHI 1998 Bradley, Bob 1.736
NY 2003 Bradley, Bob 1.350




NE 2000 Clavijo, Fernando 1.253
COL 2005 Clavijo, Fernando 1.274




TB 1998 Hankinson, Tim 1.400
COL 2001 Hankinson, Tim 1.248




MIA 2000 Hudson, Ray 1.730
DC 2002 Hudson, Ray 1.258




NY 2005 Johnston, Mo 1.353
TOR 2007 Johnston, Mo 0.833




SJ 2004 Kinnear, Dominic 1.589
HOU 2006 Kinnear, Dominic 1.634




NY 1998 Mondelo, Alfonso 1.294
TB 2001 Mondelo, Alfonso 0.588




LA 1996 Osiander, Lothar 1.480
SJ 1999 Osiander, Lothar 1.105




TB 1996 Rongen, Thomas 1.763
NE 1997 Rongen, Thomas 1.066




LA 1999 Schmid, Sigi 1.710
CLB 2006 Schmid, Sigi 1.123




SJ 2001 Yallop, Frank 1.713
LA 2006 Yallop, Frank 1.444




LA 1997 Zambrano, Octavio 1.910
NY 2000 Zambrano, Octavio 1.523

Comments on "Coaches Better the Second Time Around?"

 

Blogger UtahGamer said ... (11:58 AM, November 22, 2007) : 

I'm not sure I'd say that it doesn't mean much. A paired t-test of the differences shows the differences statistically significant (t(12)=5.70, p <.001). Applicability of a t-test when each score is composed of varying number of entries might be questioned, but I don't doubt that this is substantial evidence that recycling an MLS coach is a bad idea. We can simplify further by doing a binomial test on 11 out of 13 coaches doing worse on second stint if the expected proportion is 50%. The probability for that binomial is .022 (2 tailed). So, I think you have more evidence than you were willing to admit.

 

Blogger henryo said ... (6:01 AM, November 23, 2007) : 

How about the stats for coaching a 3rd/4th team?

 

Blogger ERic said ... (11:36 AM, November 24, 2007) : 

I was wondering the same thing, henryo. But we're probably a long way away from having sufficient statistics for that. I only remember Rongen (DC) and now Yallop has having coached a 3rd team.

 

Anonymous Anonymous said ... (5:23 PM, November 25, 2007) : 

utahgamer ... nice analysis, but I think it has a big hole.

By giving evidence that coaches who get hired again do worse, you're really just proving regression to the mean. The coaches who get hired again are generally ones who've been reasonably lucky in their previous jobs. In the future, it only stands to reason that they won't get so lucky again.

 

Blogger Kinney said ... (1:30 PM, November 26, 2007) : 

Bob Bradley at Chivas is another coach with a 3rd team.

 

Blogger scaryice said ... (9:24 PM, November 26, 2007) : 

Bradley @ CHV - 1.343, worse than both the first and second teams.

Rongen @ DC - 1.412 and @ CHV - 0.400.

Those are the only guys with more than two.

 

Blogger UtahGamer said ... (8:25 PM, November 29, 2007) : 

Regression to the mean is certainly a reasonable hypothesis, for the reasons you say. Well noted!

 

post a comment