MLS vs INT and newbie Nats

[NoSix]

In the table below I calculate career minutes, goals, and assists per 90 minutes in MLS and in USA international matches for select US MNT veteran players (greater than 1500 career international minutes):


Player Min MLS G MLS A G90 A90 GA90 Min. INT G INT A G90 A90 GA90 Pred
Donovan 9134 44 39 0.43 0.38 0.82 6161 25 22 0.37 0.32 0.69 0.73
McBride 13973 62 45 0.40 0.29 0.69 6019 29 10 0.43 0.15 0.58 0.61
Armas 18467 11 41 0.05 0.20 0.25 5430 2 4 0.03 0.07 0.10 0.23
Lewis 8867 9 35 0.09 0.36 0.45 4431 8 16 0.16 0.32 0.49 0.40
Beasley 7379 14 18 0.17 0.22 0.39 3897 12 6 0.28 0.14 0.42 0.35
Mastro 13981 3 15 0.02 0.10 0.12 3146 0 1 0.00 0.03 0.03 0.10
Wolff 10844 54 33 0.45 0.27 0.72 2273 9 5 0.36 0.20 0.55 0.64
Convey 6875 8 14 0.10 0.18 0.29 2195 1 4 0.04 0.16 0.21 0.26
Ralston 24865 51 102 0.18 0.37 0.55 1885 4 6 0.19 0.29 0.48 0.49
Olsen 12296 19 38 0.14 0.28 0.42 1518 5 1 0.30 0.06 0.36 0.37
Total 126681 275 380 0.20 0.27 0.47 36955 95 75 0.23 0.18 0.41 0.41


Comparing goals and assists per 90 minutes between MLS and international play, I find two interesting results:

1) The correlation coefficient between the MLS GA90 and INT GA90 columns is a whopping 0.95.
2) From the ratio of the total average MLS GA90 and INT GA90 values (and ignoring the existence of secondary assists in MLS) one can conclude that, on average, the INT defenses of USA opponents are only 15% tougher than MLS defenses.

Normalizing the MLS numbers by this average ratio, in the final “Pred” column I have calculated the predicted international goals and assists per 90 min for each player based on their corresponding MLS ratio. While there is obviously some player to player variability, agreement between predicted and actual performance is not bad at all.

Now let’s look at the same comparison for select US MNT newbies (less than 1500 career minutes):


Player Min MLS G MLS A G90 A90 GA90 Min. INT G INT A G90 A90 GA90 Pred
Zavagn 16610 5 20 0.03 0.11 0.14 1157 0 1 0.00 0.08 0.08 0.12
Dmpsy 4343 17 10 0.35 0.21 0.56 1041 3 2 0.26 0.17 0.43 0.50
Klein 16381 39 45 0.21 0.25 0.46 1016 5 3 0.44 0.27 0.71 0.41
Ching 4407 26 12 0.53 0.25 0.78 834 3 0 0.32 0.00 0.32 0.69
Twllmn 8587 64 18 0.67 0.19 0.86 758 5 3 0.59 0.36 0.95 0.76
Johnson 5303 24 6 0.41 0.10 0.51 696 9 1 1.16 0.13 1.29 0.45
Noonan 6034 29 22 0.43 0.33 0.76 625 1 1 0.14 0.14 0.29 0.68
Total 61665 204 133 0.30 0.19 0.49 6127 26 11 0.38 0.16 0.54 0.44


As a whole the new group, in their international performances to date, have bettered their MLS numbers, presumably due to weaker opposition, on average, than the veteran group have faced.

Comparing the international GA90 values with values predicted from the MLS data, it is clear that Eddie Johnson has far exceeded his MLS output to date. While no-one could reasonably ask him to permanently raise his MLS game to the astronomical levels he has displayed in his nascent international career, at a minimum I would hope he could at least begin to raise his MLS offensive output rate to Josh Wolff-like levels. While Taylor Twellman’s assist rate might be expected to dwindle over time, based on his MLS numbers he has a reasonable chance of maintaining his current goal-scoring rate. Were he to face stiffer international opposition, you would have to think Chris Klein would be hard pressed to continue scoring goals at twice his MLS rate. Note that, with the most minutes of the newbie group, Zavagnin’s and Dempsey’s international numbers are starting to approach the levels expected based on their MLS data. Ching and Noonan, on the other hand, have significantly underperformed relative to their MLS standards. Noonan at least has the excuse that the majority of his international minutes have come as a midfielder, although his 1 goal in 625 international minutes is perhaps eerily reminiscent of the 1 goal in 626 career minutes tallied by Jason Kreis. ;)

[numerista]

Looking at players who play the same role consistently in both the national team in MLS, it appears that the model's predictions are too high for dmids (Mastroeni, Armas, and Zavagnin all "underproduce") and too low for wide mids (Beasley, Lewis, and Klein all "overproduce," while Ralston produces as predicted).

[NoSix]

Looking at players who play the same role consistently in both the national team in MLS, it appears that the model's predictions are too high for dmids (Mastroeni, Armas, and Zavagnin all "underproduce") and too low for wide mids (Beasley, Lewis, and Klein all "overproduce," while Ralston produces as predicted).

Yes, I noticed that too. Maybe that has to do with Arena's expectations for those roles compared to the MLS norm?

[numerista]

Yes, I noticed that too. Maybe that has to do with Arena's expectations for those roles compared to the MLS norm?

Could be, although it may also be related to the fact that MLS awards secondary assists, while the national team does not.

[NoSix]

Could be, although it may also be related to the fact that MLS awards secondary assists, while the national team does not.

While the secondary assist effect is there, I don't think it explains the data. The wing midfielders are scoring goals at a higher rate for the USA than in MLS for some reason. I don't find it surprising that the defensive mids international offensive production is lower - anecdotally I've noticed Armas and Mastroeni playing box to box for their club teams, while rarely venturing into the offensive third for the Nats. The wing midfielder result, on the other hand, is not something I would have expected prior to seeing the data.

[numerista]

The wing midfielders are scoring goals at a higher rate for the USA than in MLS for some reason.

It isn't just wing mids. Across the board, goals are scored at a higher rate for the US.

Looking at things in a bit more detail, Beasley and Lewis both became better goalscorers as they matured, so their MLS numbers are skewed by their pre-NT level. Ralston's NT goalscoring is below the expected level. It's pretty inconclusive.

[Soccernethost]

Am I right that if the correlation between MLS and Int'l is so high that you would be correct to make assumptions about how a player might score for the NATS using his MLS #s?

[NoSix]

Am I right that if the correlation between MLS and Int'l is so high that you would be correct to make assumptions about how a player might score for the NATS using his MLS #s?

Almost. I would say that the data indicate that for MLS players who play well enough to earn significant playing time (>15 full matches) with the Nats, their offensive production in MLS is a good predictor for their offensive production with the Nats. However, this data doesn't tell you whether a player like Jason Kreis or Brian Ching, who didn't do/hasn't done as well as his MLS performance would suggest in limited minutes with the Nats, would or would not achieve his MLS performance levels if he nevertheless had been/were to be given significant Nats playing time anyways.

[numerista]

Am I right that if the correlation between MLS and Int'l is so high that you would be correct to make assumptions about how a player might score for the NATS using his MLS #s?

If you had done that, you might've assumed that Eddie Johnson wasn't worth capping, since his career MLS stats were not very good. You might also have assumed that it wasn't worth sending Carlos Bocanegra forward on set-pieces, since he didn't get involved in many goals in MLS. That is to say, you risk making some very bad assumptions.

Certainly, correlations exist. A lot of soccer skills do carry over from one level to another, and it would be downright shocking if a forward was suddenly getting outscored by a d-mid. However, it's risky to generalize a correlation without validating it on new data.

[NoSix]

blah blah blah

Pay him no mind, soccernethost, he thinks the only valid stat is one he posts himself.

[numerista]

Pay him no mind, soccernethost, he thinks the only valid stat is one he posts himself.

Come on. In another thread you accused me of "wacking off" to someone's stats that I respect. Do you really need to attack me personally just because I don't always agree with you?

[scaryice]

NoSix, you seem to have a lot of enthusiasm for stats, but your attitude sucks. I have to take Numerista's side.

[NoSix]

NoSix, you seem to have a lot of enthusiasm for stats, but your attitude sucks.

My attitude sucks? I'm not the one following him around and attacking his every post.

[numerista]

My attitude sucks? I'm not the one following him around and attacking his every post.

Following you around? Apart from this thread, I'm aware of one other thread where I've interacted with you at all, and I'd been posting in that thread for weeks before I had any contact with you. What's more, if you read my responses in this thread as "attacking every post," you have an awfully narrow view of discussion.

[scaryice]

My attitude sucks? I'm not the one following him around and attacking his every post.

I think you need to learn how to take criticism.

[the stranger]

I know somebody talked about correlation between MLS playoff numbers and national team success. On a similar tangent, I wonder, what do your numbers look like when broken down between Home and Road games. MLS teams play pretty conservatively on the road, and attacking players get fewer touches, chances, etc. So who can still get it done?