Rankings

[ArdeeBhoy]

With you on the dream group....

So that won't happen.

[pineapple stu]

Stu- I've just tried unsuccessfully to find my previous post about the Elo system. IIRC, it was from about December 2009, ie just after all the European qualifiers for the 2010 finals were known. A number of them were ranked lower than other countries who had failed to qualify, with fewer points. That suggests to me that the Elo system, for all its complexity, has a pretty basic flaw. By not simply presenting the ranking based on points at the end of qualifying, or a whole tournament, it ignores reality.

Without seeing exact examples, it's hard to properly refute your point, but differences in group strength would easily explain this. You wouldn't expect the same points tally from a group containing Montenegro as bottom seed versus one with San Marino as bottom seed. Consequently, fewer points could still indicate a better performance, and a better team. Similarily, coming third - and missing qualification - could easily come down to a freak result, like the group winner slipping up at home to second with the group already won. Does that make the third placed team any weaker?

As a chess player - which uses the same system to rank every player from beginner to World Champion - I can say the Elo system is statistically very fair. You may have to adjust it for football a bit - reduce the importance of friendlies is the obvious one, but that's easily done - but it doesn't have the flaws you suggest.

[Gather round]

Without seeing exact examples, it's hard to properly refute your point

Fair enough. Alas the Elo rankings site doesn't easily enable me to compare the whole table from one year to the next, as FIFA's does. Unless I've musunderstood its structure. Anyway, as far as I remember three of the teams who qualified for 2010 had a lower ranking in the first half of that year (ie after qualifying finished) than three who didn't, and who earned less points.

The current table ranks Wales at 35th in Europe, NI 37th. Respective comparable qualifying record to date? We've got five points from four games, they've chalked up a big fat zero.

but differences in group strength would easily explain this

They don't explain my admittedly less wide-ranging stat above (we're lower because we normally lose friendlies). Wales's group has only two 2010 finalists, ours has three. Differences in group strength are a red herring; even if the seedings were based on only the most recently completed qualifying, as I'd prefer, you'll get some anomalies. Which is why I say the groups are of roughly equivalent standard, not exactly. And the standard varies across even a single tournament. In August 2010, Euro 2012 Group C looked stong; after NI couldn't beat Faeroes, Estonia lost there, Slovenia only managed one point of six against us and Serbia have been a bit of a train wreck off the field as well as on, it doesn't so much.

You wouldn't expect the same points tally from a group containing Montenegro as bottom seed versus one with San Marino as bottom seed

Only if it considered only those two games out of ten in isolation. Remember that in that group's case, it also included the less than World-beating Georgia and Cyprus- they won three matches between them in 30 attempts. So, see above: the overall standard of the groups tends to even out. Even Montenegro's special status (as a 'new' country) is a bit misleading. In WC 2014 qualifying, you might not expect as many points against Wales as someone else could notionally hope for versus San Marino- but you'll still need them to progress from the group, and they'll still be equally discounted when comparing second-place teams. Wales are bottom feeders this time because they are at their weakest for years. They are a bit unlucky (I reckon) because if the draw had been delayed until November I imagine they wouldn't still be in the bottom pool. But they'll just have to tough it out.

As a chess player - which uses the same system to rank every player from beginner to World Champion - I can say the Elo system is statistically very fair. You may have to adjust it for football a bit - reduce the importance of friendlies is the obvious one, but that's easily done - but it doesn't have the flaws you suggest

I don't play chess, but asked a friend (decent standard club player) who agreed with you. Then again, you don't need to compare every football team globally; just the 53 taking part in Euro qualifying, fully seeded into nine groups to provide roughly equivalent standard.

However...saying that you may have to adjust it is a bit odd, when the Elo system's supporters have been publicisng their rankings since 1997. Isn't 14 years long enough for the maths professors to finesse out the crap friendlies?

I've outlined its basic fault- at the end of qualifying and tournament finals (ie the only time when the rankings really count for much; you don't read that much into the English PL table in November or February) it has a number of results that contradict reality. The complexity and the reference to long-past tournament results don't help (the latter shared with FIFA's ranking system).

ELO's own website says, "Ratings tend to converge on a team's true strength relative to its competitors after about 30 matches. Ratings for teams with fewer than 30 matches should be considered provisional". Which rather misses the point that for most European teams, each competition only lasts 10 matches over two seasons.

[pineapple stu]

No offence, but there's a difference between not understanding why things are and that thing being wrong. You haven't outlined any basic fault at all; I just don't think you're looking at the results in enough detail. I don't know the ins and outs of the ratings, but for example, NI's draw against the Faroes possibly cost them as many points as any of Wales' defeats.

saying that you may have to adjust it is a bit odd

You're missing my point - you may have to tweak the system used in chess when comparing football teams. This could easily have been done from the start; just needs a bit of thinking. So you just weight friendlies less than competitive games (there's no comparative weighting in chess, which is why I'm saying you may have to adjust it). So for a game between two teams of the same rating, a competitive game might see the winner jump ten points and the loser drop ten points, while the same result in a friendly might see a jump of +/- 2.5. And the same result in a finals game might see a 15 point swing. (And indeed, just looking on wiki, that is taken into account)

I don't see what the problem is that a team's rating is really over 30 games. What this means is that say had Montenegro won their first game, they'd have been rated infinity. That's clearly nonsense. Had they won 2 of their first 3, they could have been 1900 (125 points above the average of their opponents). Say they then lost their next game, they might drop to 1800 (the same as the average of their opponents). That's a big drop, and that's why you need a set number of games before these big variations die out (30's a bit high; I've seen 20 or 9 used). And what it also means is that results more than 30 games ago don't really affect the rating - which is what you want. One year - as you suggest - is way too short. Six years - as I think FIFA use - is arguably too long.

On the groups thing, don't forget friendlies are included in the ratings as well, so winning a group isn't the be all and end all. And yes, while group strength will roughly balance out - though it's still a factor to consider - you didn't address my point of a freak result unconnected to your team being the difference between second and third.

Ultimately, the system works, more or less literally by definition. I think arguing against it on vague grounds while not exploring the reasons behind the ratings changes is a bit like arguing against science because it sounds wrong, even though it's been proven.

[Gather round]

[

No offence, but there's a difference between not understanding why things are and that thing being wrong. You haven't outlined any basic fault at all

None taken. Er, I have explained a basic fault- that the rankings don't match performance in competition- and explained why.

but for example, NI's draw against the Faroes possibly cost them as many points as any of Wales' defeats

Indeed, so one point is worth less than none. It's pointless (pardon the pun) complexity. Note that I compared results excluding the Faeroes, as there are only five teams in the Wales group. On 1 July last year, we started level on zilch, now we've five more points than they do. There's no need (other than pandering to its sponsors) to have a ranking system beyond that revealed in qualifying and tournament finals; that can be used to set seedings. Similarly, there's no need to track detailed rankings back to the dawn of FIFA time- tournament performance summarises it quite well.

You're missing my point - you may have to tweak the system used in chess when comparing football teams. This could easily have been done from the start; just needs a bit of thinking. So you just weight friendlies less than competitive games

No, I get your point. Including friendlies in the rankings is genuinely pointless (again, sorry) because they're separate and irrelevant to competition results. I'm assuming they're in the calculations largely to bul k the data and make it more sellable?

I don't see what the problem is that a team's rating is really over 30 games

I explained. For most teams, who don't qualify, a tournament lasts ten games. Then, on 1 July after the finals end, you turn the clock back to zero.

One year - as you suggest - is way too short. Six years - as I think FIFA use - is arguably too long

Strictly, it's about 18 months (July to the following December). Without laboring the point, each tournament could easily be self-contained for ranking purposes. At the end of each qualifying, you have a ready made table to seed the next one.

you didn't address my point of a freak result unconnected to your team being the difference between second and third

I don't think it's significant. Your position depends on aggregated results over 30 games in the group, of which you play in 10. That's more than enough to overcome freak results.

Ultimately, the system works, more or less literally by definition

It wokrs in chess; it's unnecessary in football and contradicts more straightforward, widely accepted results (you know, where one point outranks none).

I think arguing against it on vague grounds while not exploring the reasons behind the ratings changes is a bit like arguing against science because it sounds wrong, even though it's been proven

Straw man alert. I backed my claims with evidence

[pineapple stu]

None taken. Er, I have explained a basic fault- that the rankings don't match performance in competition- and explained why.

And I've said that if you can provide a solid example, I'll have a look at it and refute it.

No, I get your point. Including friendlies in the rankings is genuinely pointless (again, sorry) because they're separate and irrelevant to competition results. I'm assuming they're in the calculations largely to bul k the data and make it more sellable?

Friendlies are in at half the weight of qualifiers; I think they're still a generally good indication of team quality. Remember that the more info you get about the relative quality of teams, the more meaningful the ratings will be. Why leave friendlies out just because they're not qualifiers?

I explained. For most teams, who don't qualify, a tournament lasts ten games. Then, on 1 July after the finals end, you turn the clock back to zero.

I don't what setting the clock back to zero has to do with anything?

Strictly, it's about 18 months (July to the following December).

Nope; FIFA rankings use results going back over the last four years.

And again, why restrict info on seeding teams to one campaign? Why not use more info to get better results?

Your position depends on aggregated results over 30 games in the group, of which you play in 10. That's more than enough to overcome freak results.

Again, you've missed my point, which is to say that a third-place finish in a group could be an equal or better achievement than a second-place finish in a similarly strength group if one result, which you have no influence over, goes against you.

It wokrs in chess; it's unnecessary in football and contradicts more straightforward, widely accepted results (you know, where one point outranks none).

There's absolutely no reason why one sport would be different to another. This part of your post - and in particularly, the bit in brackets - shows you don't actually understand the basic premise of Elo ratings. NI outrank the Faroes by 250 points, so they should score 75% over time. So if they draw, they've scored 0.25 less than they should have. Wales are 300 points lower than England, so should score 17% over time. So if they lose, they've scored 0.17 less than they should have. So we can see that the North drawing against the Faroes is a worse result than Wales losing to England (relative to what was expected of them). The point is that the Elo system is far more precise than just "win/lose/draw".

Straw man alert. I backed my claims with evidence

Where?

[ArdeeBhoy]

Stu wins on points I reckon.
And by 'points', mean a unanimous decision....

Actually, make that a country mile.

[Gather round]

Friendlies are in at half the weight of qualifiers; I think they're still a generally good indication of team quality

I disagree, as I've repeated; they're basically irrelevant to the quality shown, and obvious, in competition results.

Remember that the more info you get about the relative quality of teams, the more meaningful the ratings will be. Why leave friendlies out just because they're not qualifiers?

Baloney. You leave them out because they're irrelevant. You might as well include results from U-21, women's matches or Eurovision.

I don't see what setting the clock back to zero has to do with anything?

Er, it would reflect what happens in real life; England and Montenegro start every competition equally, ie on nil points: if the latter team win the group, it makes them automatically better (and thus higher ranked, you'd think) than the former.

Nope; FIFA rankings use results going back over the last four years

I was talking about what I think the rankings for European should be, not what FIFA uses; a simple table of results from one July to the following November.

And again, why restrict info on seeding teams to one campaign? Why not use more info to get better results?

I've explained repeatedly. It disadvantages teams (generally from smaller, weaker countries) who have one good competition amidst others where they do less well. An example between Slovenia in WC 2010- they finished five points ahead of NI in the same group, yet were seeded lower in the next qualifiers.

Again, you've missed my point, which is to say that a third-place finish in a group could be an equal or better achievement than a second-place finish in a similarly strength group if one result, which you have no influence over, goes against you

I haven't. Third place in group X with say, 19 poiints obviously means earlier elimination than runner up with 18; but it still takes credit into the next seedings, in my ideal system. Otherwise, it's just a quirk of the impossibility of every team in the competition playing every other twice.

The point is that the Elo system is far more precise than just "win/lose/draw"

I understood Elo's premise quite well, thanks. Aggregating points in competition over 10 matches is 100% precise, simple to understand and uncluttered by irrelevance like friendly results. Of course I appreciate Elo's worth elsewhere, it simply isn't needed for Euro 2012,

[zero]

apologies if this has been mentioned, but if the info on wikipedia is correct:

http://en.wikipedia.org/wiki/UEFA_Eu...g#cite_note-11

' Each nation's coefficient and ranking position for the draw and final Euro 2012 draw will be calculated per results up to and including 11 October 2011'

then should we finish second, we will (almost?) certainly be seeded in the playoffs. finishing as best second place team is now very unlikely given sweden's form, but of the clutch of second place teams our ranking

http://www.world-results.net/uefa/ranking.html#2011

is better than:

in group A: Belgium and Turkey, who are gunning it out for 2nd
in group C: Slovenia and Serbia (ditto)
in group D: Belarus / Bosnia
in group G: Montenegro / Switzerland
and group H: Norway currently second (we're above them in ranking), but could be caught by Denmark (above us).

that leaves 4 or 5 teams below us, meaning we would be seeded, and therefore playing one of the above mentioned teams. serbia and switzerland look the 2 to avoid to me. actually turkey or bosnia would be another grim prospect.

[Stuttgart88]

Sweden still ahve to go to Hungary and Finland, so may well drop more points. Don't they have Holland at home too, or have they played twice?

[ifk101]

Yes Sweden have a fixture against Holland remaining. Don't see them dropping points against Hungary or Finland.

[the bear]

http://www.uefa.com/uefaeuro2012/sta...961/index.html

still have hollnd at home on final day, and hungary away in next match who need a win to retain any hope of qualifying. i reckon they will defo drop points. bear in our results against the bottom teams in the group won't be counted in

[ifk101]

Sweden have played Hungary a number of times in recent qualifiers and won all those games. They certainly have the metal edge over the Hungarians who are a pretty average team tbh.

Sweden beat Finland 5-0 in their most recent qualifier. Think our match against NI to get an idea of how one-sided that game was.

Playing Holland at home last suits the Swedes. Holland will be qualified by then and are likely to rest players accordingly. Holland hammered Sweden in the first game so I'd imagine the Swedes will be out to prove they are better than the result in Amsterdam.

I think Sweden have the best second placed team in the bag.

[the bear]

if sweden beat finland, san marino & hungary but lose to the dutch, they will end up on 24 points (-6 from games v san marino = 18)

we would require to win all our games to beat that with 25 points (19), but sure if we win all our games we would win the group anyway so i agree sweden seem to have it in the bag. unless hungary or finland can pull one over them, they could prob afford to draw one of those matches and still get it

[geysir]

Swedes are far from certainties. Hungary are no pushovers, Finland would be up for their game and Holland play a mean dead rubber.

[ArdeeBhoy]

Let's worry about coming second ourselves before we worry about anyone else...

[the bear]

lets worry about coming first before we worry about coming second

[AlaskaFox]

31st: http://greenscene.me/2011/06/ireland...orld-rankings/

England 4th? Italy 6th? Mexico 9th?

[Crosby87]

So England are better than Brazil now?

[cornflakes]

The North went up 3 places? Here is the full rankings