Review: Weekly MLS Power Rankings.

I'm sold on neither SJ nor RSL, and I'm a fan of both.
As Vancouver should have made clear, RSL is vulnerable in the back this year for many reasons, especially against speed. SJ relies on smart, speedy wing play, smart forward play and good finishing. They have been lucky too. Luck always runs out. More importantly, they have an average #6 and a below average partner for Bermudez. They will get exposed by somebody come playoff time. My money is on Seattle to hurt them badly when it really counts.

 

They jumped KC in one ranking and they jumped LA and VAN in the official MLS rankings. Jumping a team losing at home (KC) makes some sense. Jumping LA (as hot as they are) or VAN (who lost on the road to what all consider the #2 team in the league) makes no sense to me.

 

Mabye extra credit for Ben Olsen due to the Allstar game?

 

+ 17 goal differential is not luck, and many calls still go against us ( look the game v Fire's three non-called handballs in the box). And his name is Bernardez and he should have been an all-star.

 

Except that San Jose has played one of the most difficult schedules in MLS so far, and has the easiest remaining and overall schedule, at least according to my ELO+ ratings (schedule strength here, current ratings here), which take home-field advantage into account.

 

A bit late this week, had a crazy travel schedule for work, but here we go. Not much change except for teams swapping places in the top third. Even with the loss Toronto creeps up on Portland a bit.

​

 

Meh, there's still a great deal of parity in MLS. At best one can possibly categorize a team as top 3rd, mid-pack, and bottom 3rd. And in the playoffs, where the teams meet once or twice, it's really a crap shoot. So really any team can "expose" any other team. You have to be good, but you need to get the bounces to fall your way as well to win it all.

The Quakes have a goal differential that is over twice that of the next best. But I'm not going to get carried away. The Quakes are good this year, but there are a bunch of good teams, and anything can happen in the playoffs.

 

Okay, here's the numbers again, with the cavet that the last set of numbers actually included the New York/Houston game by mistake, so that's reflected in this weeks numbers.

The big thing that shocked me and had me triple checking formulas was despite their thrasing of LA they actually DROPPED a ranking... This is more or less Chicago's fault, but the big take away I discovered was that Seattle's 9 game winless streak (now six game undefeated, spin it how you want) really put them in a hole since the system rewards form more than individual results and "remembers" past form from the season.... Now that they're rolling in the right direction, Seattle looks to make some big jumps, provided they keep the momentum going.

Outside of that, nothing too surprising this week I don't think...

​

 

Got some more numbers for ya. Nothing too drastic this week, but I think it looks pretty good, except that DC is probably high. I take a little solace in knowing they could still prove themselves with theit games in hand on the top teams though. They're also on a sharp downward trajectory.

I also think we've hit the point in the season where individual games will start to have less of an influence (though LAs recent 4 goal loss and then 4 goal win were both dramatic) on ranking as teams have established a lot of their season long identity. Which isn't to say the ratings are going to get set in stone, just that in the last push of the season teams are really going to need to maintain good form going into the playoffs. Of course that's hardly new knowledge or groundbreaking advice...

​

 

So if you're a guy like me, and I know I am, you like graphs.

I took a few minutes this morning to plot up my weekly rankings and put together a seasonal trend chart for the rankings my spreadsheet has spit out.

The only cavet I've got is that I didn't start doing these until Week 4 (they wouldn't have been meaningful anyway) and there have been some midweek ratings, so instead of going by "Week" on the x-axis, it goes by "rating point".

I've never tried directly uploading an image so let's see what happens... I imagine you can just click it to embiggen it?

 

Cool. I would actually love to see it from Week 1 because then you'd see more dramatic rises (San Jose, Vancouver maybe, etc.) and dives (LA, Portland maybe, etc.?) based on early expectations. As it is, it's surprisingly flat when you compare Week 4 to Week 24, especially Toronto :--).

 

Yeah, I thought about that. The formula is sort of setup based on five game form, so even the first couple iterations involved some fudging for me, but given how uncertain things can be at the start of the season I guess that's not really a bad thing. I might go retroactively calculate rankings for the first 3 sets and add em at the end of the season.

I really did try to make sure rankings don't swing too wildly week to week. I always think the way basically everyone does these rankings are too knee jerk... A 34 game season is a long time and every team is going to hit a rough patch or a good stretch at some point.

 

Could you graph the SHFR? I don't know if it would remain coherent, but there would be a lot more resolution and we might see trends that were not large enough to cause a change in rank(ie: DC's recent decent)

 

It does, actually, and it's probably the better graph. Good call. Here's what it looks like.

This is the whole thing.



No, there's no error with SKC, they were actually playing that well to start. The SHFR ratings would go from 1,000,000 to 47,619. So if a team won every game, they'd be at 1,000,000, and if they lost every game, they'd be... well, almost Toronto.

I thought about plotting it as a log plot but that didn't help much, so here it is linear with the axis adjusted for just the end of the season stuff.



Not sure I've got a lot more to add really, except that San Jose is playing really, really well. Edit - I do have one thing to add, normally I graph everything up in Excel 2007, but I'm on Excel 2003 right now, so these might not be quite as pretty as they could be...