Thursday, March 16, 2006

MLS: Starting Lineup Stability (2005)

I mentioned previously that I have wanted to study the effects of starting lineup stability on a team's performance. Basically, my goal is to figure out the number of starters returning from the last match, and check out the points per game depending on that number.

MLS is obviously very different from other leagues because of the games players miss because of national team duty. That could mean that it would be a better indicator of the impact of those absences, but all teams miss players at the same time. And of course, the much debated playoffs. I'll have to try and do the Premiership some time.

I have put together the numbers for this past season, and not only that, but I'm also figuring out how many days there are since the team's previous game, and whether or not a cup game was played in between. So when I eventually do this for every season (probably a while from now), I will have a bunch of great stuff at my disposal.

Let's start out by looking at some basic team stats (using games 2-32 of the MLS schedule):

Returning Starters Per Game

1 SJ 9.77
2 KC 9.48
3 NE 9.32
4 CHV 9.29
5 CLB 9.13
MET 9.13
7 DC 9.10
8 COL 9.00
9 LA 8.81
10 RSL 8.71
11 DAL 8.58
12 CHI 8.55

League Average: 9.07

The top team has the fewest changes in their starting lineup. The Revs are right up there as well. I think that you also can go without a lot of changes because you really don't have anybody else to turn to, which explains the high ranking of Chivas. The two teams who changed their lineups the most were Dallas and Chicago, not too surprising. Both teams missed several players through international callups and struggled with consistency.

I'm not sure how much you can really tell from that stat. But I think it's pretty clear that the best teams rarely change their lineups. Just look at Reading in England this year. Again, with MLS it's hard to draw conclusions. Let's take a look and see how teams' rankings in this stat compare to their 2005 rankings in points per game (PPG) and subs per game (SPG). RSPG stands for "returning starters per game."

SJ 1 1 12
KC 2 7 (tie) 8 (tie)
NE 3 2 11
CHV 4 12 8 (tie)
CLB 5 (tie) 10 1
MET 5 (tie) 6 10
DC 7 3 5 (tie)
COL 8 7 (tie) 3 (tie)
LA 9 7 (tie) 2
RSL 10 11 7
DAL 11 5 5 (tie)
CHI 12 4 3 (tie)

Of the top 6 teams in RSPG, 5 are in the bottom half in SPG, and vice versa. So it would seem like the more starters you have returning per game, the fewer subs you use. Which again, either means that what you have is working, or you have little to work with. Or it could also be a matter of who's doing the coaching.

PPG by # of Returning Starters

RS PPG Games
4 1.33 3
5 0.67 3
6 1.50 10
7 1.14 22
8 1.30 77
9 1.39 107
10 1.55 95
11 1.35 55

This does not take into account anything but the basic stats, so there's nothing in here to account for the strength of the team, or whatever else. And, I don't know how teams matched up in terms of the number of returning starters. Since international absences usually affect everyone at the same time, it might be more useful to do a direct comparison of head to head matchups in this stat. That's for another day.

There were very few instances of 4-7 returning starters (a combined 9.9% of total games). If you were to combine those categories together, it gives you 1.21 ppg, and a more obvious pattern:

4-7 1.21
8 1.30
9 1.39
10 1.55
11 1.35

Though this is very preliminary, there could be somewhat of a relationship between the two statistics. Again, I believe teams that don't change their lineup are either very strong or very weak, which averages out. In this case, keeping all starters led to 1.35 ppg, compared to a 1.39 average for every league game (including the first six games which weren't included above, obviously).

Although, a quick check says the numbers don't back that up. The league average is 4.58 games per team with all 11 starters returning. SJ and NE did that 9 and 7 times respectively, while CHV and RSL did it only 4 and 2 times. I guess teams don't perform as well keeping all their starters than they do with ten. Every team returned all their starters at least twice, and only three had better PPG with 11 than 10. DAL and DC posted modest improvements in the comparison (0.11 and 0.12 ppg), while only LA showed remarkable improvement:

10 0.75 4 3
11 2.40 5 12

The last two columns are games and points. Of course, with the small sample size there could be other reasons. The 5 games with 11 returning starters were games 2-6 of the season for LA, including 4 straight at home. Conversely, 3 of the 4 games with 10 returning starters were on the road.

While this is very interesting stuff, we can't really conclude anything just yet. Gathering more data from previous seasons will help to determine any true trends. Such as, do better teams perform better when returning more starters?

Once I have that data (and/or data from the Premiership, which again I believe would be a better test), we'll know for sure. Or at least a little bit better. Then I can have a big enough sample size to really analyze the data to determine significance, variation, and all that other stuff I don't remember how to do. I actually got a C in my stats class, can you believe it?

If anyone wants this data to help me with this project, or to analyze yourself, don't hesitate to contact me.

Comments on "MLS: Starting Lineup Stability (2005)"


Anonymous jamesey said ... (9:50 AM, March 16, 2006) : 

I think it would be good to figure out what the W/L/D records were for instances when a team started the same 11 players as the previous week.


Blogger scaryice said ... (3:33 PM, March 16, 2006) : 

1.35 ppg, so I would guess that's a .500 record.


post a comment