Just keep rating... just keep rating. Rating rating rating rating.

So after some discussions on BigSoccer I wound up mocking up a new set of rankings, first as sort of a FINE, LETS SEE WHAT HAPPENS WHEN WE DO IT YOUR WAY sort of thing, but then after making some raw ratings (i.e. not corrected for playing time) I realized that I could do a similar weighted rating like I did with team averages to maybe come up with something better.

So, here's a list of the top some amount of players, based on both total playing time corrected rating and raw, uncorrected rating.

Full ratings available here

Rate ALL the players... AGAIN!

So I did this at the halfway point and decided to do it again. I went through and assigned ratings to every MLS player.

Of course, it's all very arbitrary, but the easiest way to do it was to grab their fantasy poitns from the MLS fantasy game and calculate based off those. There's plenty of room to argue that poitns are weighted poorly or that playing time is too much a factor in rating a players worth, all very valid and I would agree. This was just the easiest way to dump a set of numbers in and crank out some kind of result. The formula is basically a players points (modified by playing time to assure that guys who score 2 goals in 10 minutes don't jump to the top of the list) minus the average players points divided by the standard deviation of all player points + 5 (the average player would have a rating of 5.00).

Here's the top 25 players in MLS for the 2012 regular season

Full ratings are available here

They kind of break down for players who play little or not at all. I hesitate to call any one at the bottom of the list "poor", but I guess, if you wanted to, you could look at all players with 1,000 minutes or more of playing time (the average MLS player logged 991.3 minutes this season) and pick the worst player there. If we use that criteria the worst player in MLS this year was... Shavar Thomas of Montreal! Yay Shavar!

Given that the average player logged 991.3 minutes and notched 0.047 points per minute played (roughly 46.7 points) the most average player in MLS this year was... Bryan Meredith of Seattle!

No surprise that Wondo get's top honors, his was a season for the ages.

You can argue that keepers are overrated in this system, and I would tend to agree. Given that most of them play most of the full season they tend to accumulate more standing, so there's that.

I'd also like to hand out an Ironman award. I thought Jimmy Nielsen would be the only player to play every minute of every game this year, but Drew Moor actually did the same thing, and did more than just stand between the posts to do it. Gotta say that Drew definitely worked this year. Respect.

So I guess using these ratings as criteria, MLS awards for me would look like this

MLS MVP : Chris Wondolowski (San Jose)
 - runner up : Graham Zusi (Kansas City)
MLS Defender of the Year : Matt Besler  (Kansas City)
 - runner up : Aurelien Collin (Kansas City)
MLS Comeback Player of the Year : ummm... I guess Chris Pontius (DC)? I always thought this was a weird award...
 - runner up : oh god... ummm... umm... Eddie Johnson (Seattle)! Or do they have to come back from injury and not just wandering in a European wasteland?
Goalkeeper of the Year : Andy Gruenebaum (Columbus). Weird to give it to a keeper on a non-post season team, but Columbus could have been the new Chivas were it not for him.
 - runner up : Tally Hall
Newcomer of the Year : Patrice Bernier (Montreal). This is another silly award.
 - runner up : Lee Young-Pyo (Vancouver)
Rookie of the Year : Austin Berry (Chicago)
 - runner up : Nick de Leon (DC)

Added a few extra tabs to the google doc and took a few minutes to tabulate average team ratings. I did both all players and players >1000 minutes (to avoid "bench" players dragging the rating down), but what actually correlated best with results was taking the average of the two averages (i.e. (all_players_average + players>1000_minutes_average)/2). Correlated reasonably well with season long performance (R^2 = 0.74)... New York surprisingly low and colorado surprisingly high, otherwise seems reasonable. Kansas City probably too high compared to San Jose...

Utterly meaningless in the end, of course... but who doesn't like a big block of meaningless numbers.

Probably more to come later, but for now, that's that...

East vs West Part II : The Bloodening

So a while ago I did a post on comparing the Eastern Conference vs the West and almost instantly regretted it. The two problems were that I used all the data, not just intra conference data, so it included stuff like west vs west and east vs east and... okay, well, just those two things. But that's more than enough to "smooth out" the apparent differences. The second was I used points per game, which is probably not correct, since who's to say a win should be worth 3 points instead of 2, or 4, or 10, or whatever. So, I redid it a slightly different way, using a T-Test to look at whether there's a difference in the number of wins, losses, and draws between the two conferences during inter-conference play. Here's the summary table for each conference.

Team	Wins	Draws	Losses
SKC	6	2	1
HOU	4	1	2
NER	4	1	3
CLB	3	2	2
NYR	2	3	2
DCU	2	2	2
PHI	2	2	4
CHI	2	1	4
MON	2	1	5
TFC	0	1	4
Average	2.70	1.60	2.90
Standard Deviation	1.64	0.70	1.29
Team	Wins	Draws	Losses
SEA	4	3	2
SJE	4	2	2
RSL	4	1	2
COR	4	1	3
VAN	3	3	2
POR	3	2	2
FCD	3	2	5
CHV	2	1	4
LAG	2	1	5
Average	3.22	1.78	3.00
Standard Deviation	0.83	0.83	1.32

So I fed each set of data (# of wins for each team, # of draws for each team, and number of losses for each team) and came up with the following P values from the T-Test.

Wins
T Test P Value	0.39
Draws
T Test P Value	0.62
Losses
T Test P Value	0.87

Obviously the big one is the Wins, as that would be where you'd see a real difference if one conference was better than the other. Though it looks like your average Western team has done a bit better than your average eastern team, the P value of 0.39 suggests that there's really no statistical significance at any serious level. It's not as non-existent as I had posited earlier, there is some tendancy for the West to win a little more than the East, but it's not clear that the West is definitely the better conference either.

If we ever get stable conferences and can track the results from year to year that would be a much better way to look at it, but on a Year 2012 basis only I'm still not quite convinced.

Also, here's a sort of nifty chart that really doesn't explain much. It's the win distribution for Eastern Teams, Western Teams, and all MLS Teams if the distribution was normal.

This is an abomination

Well, I'd apologise for not updating this more often since it's been an excellent week of soccer, but I'm stuck off in New Jersey doing some work and have been putting in extra long days, and I'm not sure there's any one out there to apologize to anyway.

I've been able to catch most of the Euro matches though, and the US qualifier match, and I've gotta say this is my favorite month of soccer outside of the World Cup. Top it all off with going to another away match (I'll be in Philly for the Union-SKC game) and it's gonna be a great great month.

Given all that, I thought I'd try something new. Since I've got a spreadsheet formula mocked up to rate MLS teams over the season, I thought I'd try to apply the same formula to the UEFA teams. Now there's one major flaw in using this system, and that is that in the most recent competetive matchces for these teams they played in different groups against opponents of very different ability, and that often the friendly matches are used for tweaks and trials and not results, so I'm not going to say I feel these ratings and rankings are fair and representative. I will be interested to see how they evolve over the course of the tournement, though.

I've looked at the last 10 matche for each side, most recent five friendly and most recent five competetive (except for Ukraine and Poland) and come up with a rating and ranking in the same way I do for MLS sides. It actually seems like it sort of worked out not terribly, but I'll let you be the judge of that.

I'm going to try to update the results after the end of each set of matches and we'll see how it plays out. Ranking a knock out tournement is probably a silly thing to do anyway, but now that it's the weekend and I'm away from family and friends I've got some time to kill.

Team	Rank	Rating
Spain	1	4.975
Russia	2	4.875
Sweden	3	3.825
Ukraine	4	3.75
Portugal	5	3.65
Czech Republic	6	3.45
Germany	7	3.2
France	8	2.7
Ireland	9	2.65
Denmark	10	2.575
Italy	11	2.425
Netherlands	12	2.2
Greece	13	2.175
England	14	1.875
Poland	15	1.55
Croatia	16	1

The big surprise of course is how far down the Netherlands are, but then again they just lost to the Danes, so maybe if I had posted this early I would have been seen as prophetic?

Obviously this isn't intended to be used for predictions or anything, but, if we just go ahead and guess that the higher ranked team will will, it gets Russia over Czech Republic right, Poland and Greece drew, though they're very close in ratings, Denmark over Netherlands got picked correct, and Germany over Portugal was incorrect, though they're also very close in ratings. Today's picks would be Spain over Italy (seems reasonable but we're tied in the 15th minute as of my typing this) and then Ireland over Croatia.

Only time will tell...

Oh, and go Poland!



Hypergeometric statistics are fun

I got curious the other day to see just what the odds of the recent 6-0-0 run were. More specifically, is it possible that SKC is just an average team who have gotten lucky? In the end, of course, it depends on how you define luck, but the math is pretty easy to work out. Assuming an "average" team plays like a 3 sided coin and wins 1/3rd of the time, draws 1/3rd of the time, and loses 1/3rd of a time, it's pretty easy to see that the odds of six in a row is (1/3)^6, 0.0014, or 0.14%. That's really small. It would be expected to happen only once every 714 runs of six games, assuming they're truly an average team. Of course you could have a run of six games every time. Assuming a 36 game season, this would happen about once every 20 years. Since the league has been around 16 years, it would stand to reason that over the course of the league's history, this would happen about once. That it's happened twice now (LA in '96 did it first) isn't really all that surprising when looking at the league as a whole. But really, for any team to do it is pretty remarkable, and I think it's much more likely to be the case that SKC isn't just lucky, but good.

To go a little further, he's a breakdown of how the league would look with true parity (i.e. each team wins 1/3rd of the time, draws 1/3rd of the time and loses 1/3rd of the time).

Points	%	Expected # of teams	Full Teams
0	0.14%	0.03	0
1	0.82%	0.16	0
2	2.06%	0.39	0
3	3.57%	0.68	1
4	6.17%	1.17	1
5	9.05%	1.72	2
6	10.42%	1.98	2
7	12.35%	2.35	2
8	13.17%	2.50	3
9	10.97%	2.09	2
10	10.29%	1.95	2
11	8.23%	1.56	2
12	4.80%	0.91	1
13	4.12%	0.78	1
14	2.06%	0.39	0
15	0.82%	0.16	0
16	0.82%	0.16	0
17	0.00%	0.00	0
18	0.14%	0.03	0
sum	100.00%	19.00	19

 

See that blip on the far right? That's SKC. Now, obviously we don't have fractional teams, so here's how we'd look with full teams only.

Pretty boring, huh?

Well, that's all fine and good, but I decided to try to play with the percentages to see what win-draw-loss rates would give SKC a reasonable chance to go 6-0-0. The percentrages all feed in to the hypergeometric distribution as such.

 Initially I used 1/3rd for each probability, but adjusting the wins % lets you play with the odds. To go from 0.14% to a reasonable probability, in this case I chose 5% (i.e. there's a 20-1 chance that SKC goes 6-0-0), I needed to adjust the win proability from 33% to 60.7%. I left draw and loss probabilities equal to each other, so they came out as 19.65% each.

Basically, what it looks like from six games so far is that Kansas City should win about 60% of their matches, and not win the remaining 40%. If we break it up equal as draws and losses we can then come up with some predicted points. Let's look at the conservative case first, in which they win 60% and lose 40%.

Based on where they are now, with 18 points and 28 matches to go, assuming they win 60% of the remaining matches (28 *0.6) would give them 16.8 wins. Let's round down to 16 and say they earn 48 more points, putting them at 66 points.

In 2011, LA won the shield (and the MLS Cup) on 67 points, and some took to asking if they were the best MLS team ever.

Now let's add draws in. 16 wins out of 28 means 12 not wins. Split those evening and we get 6 draws, or 6 more points. So, 6 + 67 would put SKC on 73 points at the end of the season.

To be clear, I don't neccessarily think SKC is a 60%-20%-20% over the course of the season, but I think they're definitely playing like it right now. With the road trip to Vancouver and Portland on short rest this week we might to start seeing some early injuries, and summer months in the US can really beat on a team physically. Their depth is going to be tested, for sure, but right now, SKC looks to be on pace to flirt with a 70 points season.

In short, I don't think this is some kind of statistical fluke. After beating RSL last week, I really think this team is the real deal.