[ChrisE]

Beineke mentioned in another thread that there's more scoring at the end of games than at the beginning, and so g/90 minutes would be biased towards substitutes. Seems likely, and I'd been wanting to look at this for a while, so I tried to draw up some numbers. Because substitution minutes and sub. goals aren't easily available, I have to say that my numbers are only approximate (and maybe occasionally wrong), but on the whole it's pretty interesting. I only used numbers from the 2003 season.

First of all, scoring rates (in g/90) by position (according to mlsnet's rosters), from the 2003 season. Unfortunately, PK goals have not been removed:

Pos	st g	st min	sub g	sub min	st. g/90		sub g/90

F+M/f	223	66843	39	8623	0.300		0.407
M	121	90270	8	5349	0.121		0.135
D+D/M	48	118602	0	3187	0.036		0.000

Some obvious trends emerge. First of all, although forwards and m/f's played almost half the number of starting minutes that defenders and defensive mids did, they played well over twice the sub minutes. Likewise, although they played only 75% of the starting minutes of midfielders, they played 60% more sub minutes. Coaches apparently have a pretty significant preference for subbing offense over defense; I'd wonder if this would be the same in, e.g. Serie A.

Another striking fact is that, although defenders and d/m's scored 48 goals as starters, they scored 0 as subs (regular season numbers only). This may, however, be simply because they didn't really have the opportunity; had defenders scored just one goal as a sub, their g/90 would have shot up to .028, quite close to their starting numbers.

However, the fact that defenders scored no goals as substitutes is a little surprising when we see that forwards (and m/f's) scored almost 25% more goals per minute as substitutes than as starters. Considering that the substitute forwards are presumably less talented than the starting forwards, this is a particularly striking number.

Midfielders didn't show a particular bias, but this may be because offensive midfielders and defensive midfielders and flank midfielders are all lumped together. I'll try to separate these out later and see if it makes a difference.

Probably the best way to combat the problem of comparing mediocre substitute forwards to good starting forwards is to compare players who played significant amounts of time as starters and subs. So, here's a list of players who played 500 or more minutes as starters and 200 or more minutes as subs:

Player Name	St. G	Min	Sub G	Min	St. G/90	Sub G/90

Noonan, Pat	10	1327	0	321	0.678	0.000
Arnaud, Davy	2	556	1	261	0.324	0.345
Cunningham, Jeff	3	1191	2	213	0.227	0.845
Harris, Wolde	2	1127	0	286	0.160	0.000
Johnson, Edward	3	1003	0	262	0.269	0.000
Magee, Mike	4	1471	3	241	0.245	1.120
Moreno, Alejandro	5	816	1	298	0.551	0.302
Rhine, Bobby	1	781	1	253	0.115	0.356
Stoitchkov, Hristo	3	612	2	290	0.441	0.621
Wolyniec, John	3	1114	2	269	0.242	0.669
Brown, Chris	5	1084	0	243	0.415	0.000
Lagos, Manny	2	1342	1	203	0.134	0.443
Total Numbers	43	12424	13	3140	0.311	0.373

The totals weren't as impressive as I was expecting, but I think a lot of that may be attributable to Pat Noonan, who played mid as an early season sub, and then blossomed as a forward. Without him in the picture, you get .268 G/90 for starters and .415 for substitutes.

[ChrisE]

Here's the scoring rate of forwards and m/f's, broken down into 10 and 15 minute blocks, based on when they were subbed in:

10 minutes:

	Goals	Min Pld	G/90
0	0	175	0.000
10	0	0	N/A
20	1	200	0.450
30	2	476	0.378
40	5	857	0.525
50	1	969	0.093
60	20	2600	0.692
70	7	1707	0.369
80	1	696	0.129
90	1	42	2.143

15 minutes:

0	0	175	0.000
15	1	200	0.450
30	3	651	0.415
45	5	1651	0.273
60	24	3572	0.605
75	4	1431	0.252
90	1	42	2.143

So, I suppose I would have expected a pretty much steady curve, with g/90 increasing as subs come on against players with less and less energy. And I would have been satisfied with something that looked somewhat like that curve, understanding that this isn't a very large sample that's being used. But that's not really what you see; instead, it seems that offensive productivity decreases with late late subs. Is this because players being brought in this late are going against bunkered defenses (or being asked to put pressure but not actually score when their team is leading), or is there some other explanation, or is it simply a statistical anomaly? I don't really know.

[beineke]

Originally posted by ChrisE
Another striking fact is that, although defenders and d/m's scored 48 goals as starters, they scored 0 as subs (regular season numbers only). This may, however, be simply because they didn't really have the opportunity; had defenders scored just one goal as a sub, their g/90 would have shot up to .028, quite close to their starting numbers.

Wow ... more great stuff, Chris.

I wonder if this is partly a matter of tactics ... when you sub in a defender, you're usually protecting a lead. That player isn't expected to go forward at all. In that respect, the main exceptions would be guys who score goals with their head, like Alexi Lalas and Chris Roner. (In fact, Roner scored one as a sub in the 2003 playoffs, two if you count the own goal against Chicago. )

On a mildly related note, the charts in your second post might also be related to tactics. When a forward enters in minutes 61-75, it's usually because his team needs a goal. Have you thought about breaking down subs' scoring rates when ahead/tied/trailing?

[ChrisE]

Originally posted by beineke

I wonder if this is partly a matter of tactics ... when you sub in a defender, you're usually protecting a lead. That player isn't expected to go forward at all. In that respect, the main exceptions would be guys who score goals with their head, like Alexi Lalas and Chris Roner. (In fact, Roner scored one as a sub in the 2003 playoffs, two if you count the own goal against Chicago. )

This was exactly what I was thinking; if you're looking to score goals, you're not going to be subbing in defenders. Nevertheless, I don't think you can draw a conclusion like 'defensive subs are less offensively-minded than defensive starters' simply because the sample is so small (and had I counted playoff goals, the numbers would have been about the same).

On a mildly related note, the charts in your second post might also be related to tactics. When a forward enters in minutes 61-75, it's usually because his team needs a goal. Have you thought about breaking down subs' scoring rates when ahead/tied/trailing?

Obviously, that's often going to be why forwards are being brought on. But I'd suspect that it's just as much the case with forwards being brought on in the 80th or 85th minute, so I think it's strange that they score at a rate that's no better than that of starters. Your suggestion about goal differential is definitely a good one, but I'm not sure I'll be able to implement it.

[ChrisE]

Originally posted by beineke

On a mildly related note, the charts in your second post might also be related to tactics. When a forward enters in minutes 61-75, it's usually because his team needs a goal. Have you thought about breaking down subs' scoring rates when ahead/tied/trailing?

Well, I did my best.

First of all, the general substitution patterns:

	Total	Leading	Tied	Trailing
F,M/F	347	86	128	133
M	229	65	81	83
D,D/M	268	50	49	48
Total	844	201	258	264

Nothing particularly surprising. Generally, a club is more likely to substitute when tied or when trailing than when leading; attacking players are always preferred over defensive players, but moreso when tied or when trailing. Interestingly, there was hardly any difference between subs when tied and those when trailing; apparently, MLS coaches are going for the win regardless.

As for the ahead/tied/trailing statistic, your suspicions proved correct. The F,M/F numbers are:

Ahead	1506	5	0.299
Tied	2918	12	0.370
Trailing	3293	21	0.574

though the midfielder's numbers were a little bit discouraging

Ahead	1136	2	0.158
Tied	1856	2	0.097
Trailing	2072	4	0.174

Nevertheless, the forwards were the part that I was really interested in here; the substitute forward's job is clear, God knows for what reason a coach is bringing in a midfielder.

[beineke]

More great stuff ... I'm not too concerned about the midfielder numbers, since the number of goals is so small.

BTW We really need to make sure that this forum has some good archiving in place ...

[mpruitt]

I was thinking about that today actually. I know that all the lists and numbers would probably be saved on people's hard drives, but what would be a good way to make sure these threads get saved in some permanent way? I know we're still a very very small group here, but maybe we should start thinking about taking some of this stuff off Biggsoccer and onto Soccermetrics.com

In the meantime, hell I'd settle for getting us on Bigsoccer Live.