Thursday, December 22, 2005

Formatting-World Cup Style

A couple of Fridays ago, I spent much of the afternoon tuning into the live webcast of the World Cup Final Draw. My main reason was to check out seeding and formatting for the biggest tournament in the world. In the end, I was quite surprised by the acceptance of pure randomness in generating the pools.

The tournament starts with 8 pools of 4. After pool play, the top 2 teams from each pool qualify for a round of 16, then qtrs, semis, and finals. Pretty good so far.

While the above makes sense from a formatting point of view, putting the teams into the pools is done by a random draw. There is a committee that selects the top 8 teams. For this year, Germany, the host team, was given the A1 slot. Brazil, as the defending champ, was given F1. Coming from my point of view, I wondered "why these slots?". Germany is probably a 6-7 seed at best, so why the A1 slot? And then, if Brazil is the actual #1, why F1? I asked a friend of mine who played semi-pro soccer in Russia about this. He is a fanatic of the sport. He had no idea why the teams were put where they were, and, furthermore, seemed to have no interest or critical thought about either the seeding or format. I was a bit surprised as I have had heated discussions about seeding and formatting BUDA Summer League tournaments.

Once Brazil and Germany were in place, the other top 6 teams were drawn at random. From there, the other 24 teams were also literally drawn out of a hat. There are geographical rules about where teams can be placed. For example, I believe that no more than 2 teams from Europe could have been in the same pool.

The bracket basically just places a first place team against another pool's second place team, but as there is no attempt and seeding reflecting ranking, there is no sense that, for bracket play, the #1 team is playing the #16 team.

I spent much of the time pondering the contrast between seeding and formatting the World Cup and what we do in Ultimate.
I came away with a great amount of respect for the care and thought that goes into the UPA formats manual. I think ranking teams in order of strength is positive, and designing brackets that set up "correct" match-ups also works. I also think that the formats go out of their way to correct for possible errors. There is no double elim in soccer, or even pre-quarters. I find it hard to believe that the World Cup format selects only two teams out of each randomly-drawn pool. I feel quite coddled now.

As I have said before, there is work to do around developing consistent seeding guidelines, but I think the inevitable quibbles and bickering about both seeding and formatting is something to be both appreciated and expected. At least we seem to care about what is going on.


llimllib said...

Those crazy soccer guys also barely keep statistics - I've gotten the opinion that they just don't care about anything other than perceived on-field performance (and they're absolutely *crazy* about that).

Marshall said...

Germany is in the spot it's in because, as the host team, they get to open the tournament (which game is pre-determined in the schedule) and play in the marquee venue for group games.

Kevin said...

One of the main reasons for the randomness is to prevent collusion or cheating. Germany is the 6th or 7th best team, but if they get the 4 spot then every other nation comes up with conspiracy theories as to why they got that high of a seed. Or if China gets the 25th seed instead of the 20th seed they say that because they're "communist" the other nations put them at such a low seed. Sure this happens in Ultimate, but I think that most people believe that the people running the office are genuinely trying their best to get accurate seeding. Even if the same is true for the FIFA office, it wouldn't stop billions of people from being unhappy about the seedings. This way FIFA makes sure there is no power pool of Italy, Brazil, Germany and England, knocking out 2 of the best teams early on, puts a few geographic restrictions on to make the pools interesting, and lets the chips fall where they may.

gcooke said...


Do you know where I could go to look up that reasoning? It makes sense to me. I think another way to frame it is: We don't know the relative strength of the teams, so why bother?

Still, though, it seems to leave it all to chance.....


Kevin said...

This article touches on some of the points, including how the draw this year is unprecedented in that it has two groups of death. Basically it says that at a certain point (doesn't say when, I'm still looking around) FIFA decided that the best idea was to switch from a seeding based system to a geographic based system. The old system ensured equal groups, but maybe 4 European teams in one group. This system ensures max 2 Euros and no Mexico and US in the same group or Argentina and Paraguay, but at the expense of equal groups.

Semar said...

If you go here, you'll see lots of the speculation/intense interest/conspiracy theories that surrounded the draw this year.

The hew and cry in ultimate would be deafening if seeding was done in the same fashion. For instance USA beat Mexico in their last match, but Mexico is seeded ahead of the Americans because the formula takes into consideration results from 8 years ago.

Net outcome? US in group of death, Mexico facing Iran, Angola and Portugal.

gcooke said...

Kevin: Thanks for that article.

Semar: If the draw is random, how can Mexico be seeded higher? In other words, it seems to me that placement within a group is not indicative of a seeding position.


gcooke said...

I think the only way in which a random draw would work in Ultimate is in situations where there is no data for seeding.

In terms of outcry, if you go to my "Crap Formats" post, I would have been happier with a random draw than with seeding that was so clearly biased.

This past June, when I was working on seeding Mixed Easterns, I talked with Dave Lionetti and Jim P about running the scramble format. I decided against it due to implementation issues, but it does begin with a random draw, I believe.


Kevin said...

Mexico was seeded higher because they were one of the eight seeded teams, while the U.S. went into the draw. The U.S. was pretty close to being one of the seeded teams, and in fact some publications speculated that they would be, but bowing out of the 1998 World Cup 0-3 and never beating Mexico outside of the U.S. Northeast (until last World Cup) didn't really help. What I think is interesting is that Ultimate attempts to seed teams based on half-reported tournament results where a full team might or might not have been there and could've been half drunk sunday morning, wheras soccer DOESN'T seed teams based on knowing every little thing about every game played between two international teams.

gcooke said...

My bad. I forgot that Mexico was seeded.

I think if you put your name on the team, and it enters a tournament, you have to accept the results, even if you were half-drunk. Another way to say it: if you want to play drunk on Sun, call your team "drunk on sunday".

As Ultimate gets better at developing a season, the data will be easier to come by.


Kevin said...

I completely agree. What is it that Mike G always says, you play with the team you got?

I think also if more tournaments move to being upa sanctioned it makes it easier to track results and such.

parinella said...

For the record, although the scramble format _can_ assume a random draw, it doesn't have to. You can choose a schedule so taht each team has about an equal strength of schedule. But more important is that it doesn't use pools after the random draw, instead effectively creating one pool with a random selection of games, then using the results of each game to create a power rating. I think that one year a team that went 1-4 made the top 8 (out of 24) because each of its games was against top 8 teams and each was close.

I could see ultimate doing a very limited random draw just to avoid conflicts of interest and seeding manipulation to avoid certain draws. Maybe seed 1-4, randomly assing 5-6 to pools C and D, randomly assign 7-8 to A and B, randomly assign 9-12, and randomly assign 13-16. Since there is no formula, usually you can argue about almost any two adjacent seeds (excepting the requirement that #1 from a Region is ahead of #2).

Tarr said...

I agree with most of what Jim said. The extreme care that is taken to fairly handling seeds in the UPA formats does create the potential for seed manipulation. You rarely see problems anymore thanks to all the oversight that goes into seeding UPA events, although sectional and regional upsets do often lead to some contentious seeding arguments. These concerns probably have a lot to do with the randomness of the WC format.

(That said, it still does not excuse some of the other problems with the WC format, like teams being eliminated on point differential. It's still a better format that some of the Olympic tournaments, which were just horrible. Women's soccer was ten teams in three pools, with the last quarterfinalist determined by point differential. This is inferior in all meaningful ways to two pools of five followed by semifinals.)

On the other hand, if you CAN trust the seeders, then any piece of information you have may as well be used, so random draws are inferior. It's the inverse of what what George said - if we DO know something avout the relative strength of the teams, we may as well make the seeds reflect it. With the advances on the score reporter, it's getting easier and easier to make a somewhat meaningful comparison.