USWNT in the Tournament of Nations 2017

Which is odd considering recent results. I would say the ranking process is flawed.[/QUOTE]
I think that most likely, the reason you say the ranking process is flawed, is that you are caught in the common error of making an evaluation on insufficient data. Ordinarily, this shows up as looking only at very recent data.

 

I think that most likely, the reason you say the ranking process is flawed, is that you are caught in the common error of making an evaluation on insufficient data. Ordinarily, this shows up as looking only at very recent data.[/QUOTE]

Quarterfinal at Olympics+last place at SheBelieves+ second at tournament of nations = #1.

Okay.

 

Was Lloyd the best player at the Olympics? Was she even top 10?

Where did I say the roster selection process is corrupt? It's outdated and has historically locked floaters out of major tournaments. That's design not corruption.

 

I don't really understand the need for the historical component. If ur trying to figure out who the top ranked college football team is for this fall u aren't giving Nebraska and Oklahoma points for being dominant in the 70's

 

We're really going to rehash (teaching you about) FIFA ratings again here?

Nothing about FIFA's woso rating method has changed recently, nor in ... years. Same algorithm, same data. (FIFA's mso rating system is slightly different.) Both systems do have sliding windows for a few years' worth of results, so they're roughly quadrennial moving averages, not this-week-hot-now soapboxen.

FIFA woso uses multipliers for importance of match: roughly 1x for friendlies, 2x for top-10 friendlies, 3x for continentals (Euros), 4x for majors (WWC, Olympics). SheBelieves and Tournament of Nations are six-sub friendlies, with some or most of the matches getting the top-10 bonus multiplier. Euros (and such frippery as CONCACAF WCQs and OCQs, go figure) earn the huge multiplier (so USA will win something like 0.02 x 3 points for an expected blowout win). Germany racked up points for winning Rio'16. All of these points decay slowly over time, as past results scroll out of the window.

Hence, recent results have very little to do with a team's amassed points total. Germany losing in Euro QFs will cost them some points (x3.0); same for France and England. Netherlands won big points for winning the whole thing, but with the home-field penalty boosting their effective rating (so their wins were less-of-an-upset than they'd be on a neutral field). Hence Netherlands made a soaring leap into the top 10, but their points gap was just too large for them to climb into the top 5.

USA is #1 because, over the ~4 year window, we've earned the most points. Germany didn't overtake us because they've only tied us in major wins this cycle (WWC'15 vs. Rio'16), we keep beating them in friendlies , and even Euro quals and Euro itself couldn't earn enough profit to cover the gap.

You're thinking only 3 tournaments ~= 1 year deep, which is exactly what cpthomas meant by "very recent". The ratings were deliberately designed to be insensitive over those evanescent timescales.

It would be more accurate to say that the FIFA ratings themselves are not a good measure of right-now "hotness" (however you want to measure that), rather than say that the measure is wrong. Like it or not, it's the way FIFA does things, it's official, it has worked quietly for the last ~10+ years without criticism from you, it hasn't changed, and it does eventually matter for things like seeding pots in a major. (There are those amongst us who have criticized FIFA's ratings for years because we don't like some of its edge cases -- and they haven't ceased doing that )

USA remaining #1 in the ratings is an objective fact, according to the published algorithm. An Excel spreadsheet could recompute that, with no human opinion. That's how it works. Now, we might all agree that Australia is "hotter" than USA right now ... and in a couple more years, they'll get a chance to prove it when it counts. Me, I figure USA has used this gap between majors in exactly the right way: by experimenting with our not-best teams and line-ups, to see who scales up to the pressure. So I shrug off your claim that we're not #1 by ... agreeing with it: the line-ups we've fielded are clearly not our "best" #1 line-up.

If you want weekly (or yearly) who's-hot-now ratings, try WWC_Movement's DIFA ratings in the Euro thread, which seems to be a combination of opinion, satire, hotness of play, hotness not of play , recent results, and players' performance. It's refreshingly not data-driven if you're more into that kind of thing. NWSL's weekly Power Ratings are similar: they very heavily weight recent results, plus player availability due to injuries or absence.

Ratings lag is endemic in ratings; that's arguably why they exist. Look at every tennis major, where somebody must make it to the finals to earn #1 over somebody who lost in the 3rd round for the 3rd major in a row, or similar. FIVB (wvb/mvb) and FIDE (chess) have similar phenomena -- in particular, USA wvb is still billed as "#2" (we are), even though our 2017 spring/summer team had almost none of the Rio'16 starters, and our B-lineup full of roster newbies overachieved just to make it to the WGP final weekend, where we got squished. (Sound familiar? Very similar quadrennial-based ebb and flow.) USA MNT just had a huge drop (to #35?) as last year's Copa Americana Centenario results scrolled past a 1-year boundary (so its multiplier dropped?), and then rebounded (to mid-20s) after winning this year's Gold Cup.

Ultimately, being ranked #1 does carry a component of past reward for past achievements. You're justified in saying we're not playing like we're #1. That doesn't mean FIFA will change their formula tomorrow just to appease the millenial generation. Let some team earn more points over the next 4 years combined; then they'll leap over England, France, and Germany That's how we got here, and that's how some team can climb over us.

 

This *definitive* post should be copied/pasted everytime someone questions FIFA rankings in one of our threads here on BS (i.e. multiple times per year ).

It's worth noting that everytime anyone seems to take WWC_Movement's DIFA rankings seriously, he has the tendency to start including North Pole, just because.

 

I know how FIFA ranking work and don't take it as gospel. Actually, it's borderline useless.

This quote is where we disagree. I see us as a team with declining players ignoring a large player pool led by a tactical dolt. You are more optimistic. In 2019, I hope you're right.

 

"Incompetence is the worst form of corruption." (Electra Glide in Blue)

So you see-- you were saying it all along...

 

Is this a personal attack?

 

No, I was pointing out that now you have indicated a belief that the selection process is corrupt.

Not corrupt as in quid pro quo, but corrupt as in doltish.

 

And other than taking Coke's money...exactly what good is it? In their own WC, FIFA ignored this glorious piece Math and gave more seeding weight to which CONFED u were in. Of course it does make a nifty final tiebreaker for US hosted friendly tournaments.

 

It's a barometer that measures exactly what it does measure, namely total performance over ~4 years. It doesn't have to be "good for" any other purpose. It gives all nations an easy and impartial benchmark to look at, and maybe something to brag about when they achieve a new all-time high (and they do). That probably suffices to put it within FIFA's interest. Many other sports have similar kinds of rating systems, with equally arcane weights and sliding windows, for the same rough comparisons, so it's not like FIFA is inventing a fad.

FIFA's own "official" usage of their own rankings is their business. It might affect things like confederation slots per major. It usually affects seeding. That doesn't mean it's even the right way to determine seeds, but if it does get used, it's objective, and any two nations could run the same numbers and agree on the result.

It makes more sense than, say, FIFA's own PotY voting, or FIFA's WC bidding process, both of which are the opposite of objective, transparent, and algorithmic.

For most teams, most days of a quadrennial, the FIFA ranking is just brag data, and has no other meaning. I suppose a mid-major down in the 40s could track their meteoric rise into the 15-18 range over a few years (or even to #8 ), and count that as a good return on having invested more resources recently. If you chunk the rating into tiers of about 10 spots each, then cross-tier matches probably are correlated, so it's a quick-and-dirty way of estimating the favorite. (Exception: A nation that rarely plays, like Cameroon, could be grossly underrated compared to their actual strength in one tourney run. But that's by design, too: if they were underrated over 4 years, that's on them.)

Sponsors (and advertisers, etc.) might care. Teams might care that their sponsors care. Home fans who hear the next set of friendlies announced could care a bit about the caliber of opponent. For example, if USA WNT play Serbia in woso and they're ranked #44, we might collectively yawn. But if it's in wvb and Serbia is #3 (and USA is #2, behind China), then suddenly you appreciate that Serbia could just flat-out kick our ass, even knowing absolutely nothing about wvb or either team's roster. Is that useful to you as a fan? FIFA thinks so, and so do a great many fans and nations.

If you try to shoehorn FIFA's ranking into your own notion of whatever, you'll probably find that it fits your idea very poorly. That doesn't mean it's "bad", just the wrong tool for whatever you're doing.

 

The way I think about it is that it isn't about the strength of teams, it is about the strength of programs.

 

Well uve failed to move me. If an effort were made to produce actual rankings that reflected the current status of the teams it will still accomplish all the things u considered positives with the added benefit of being a truly useful tool. When the "rankings" come out this month The Netherlands will likely climb into the top ten. For me if they aren't placed around #7 then reality has fallen victim to Math.

 

Your citing one Euro tournament as a sufficient sample size on which make a determination of reality is a fairly common erroneous human way of thinking. See Kahneman and Tversky,s work on how human ways of making judgments in unclear cases have common errors.

Statistical models such as FIFA's rating system are designed to tell the likelihood of the outcome of the next match based on events that occurred in some time period over the past. The time period used, and other aspects of the model, are tested against outcomes to find the time period and other aspects that will produce the best correlations between the model and actual results. Good models will outperform most humans who simply use tier personal views of "reality." A few humans will outperform the models, possibly because they simply fall on the lucky end of the bell curve or because they have special knowledge that a model can't take into account.

 

Sounds idyllic! Post the details when it's ready

... for example, I would like to know the current status of Cameroon, so that your system's prediction of their next match comes out correct.

 

And let's remember, it is widely agreed among experts in rating systems that where you have a large number of teams, it takes close to 30 games played by a team being rated to come up with a reasonable rating.

But better yet, KT, make a list of the top 60 teams in the FIFA rankings and, over the next year, before they play games, let's see your predictions of the outcomes -- who is going to win and who is going to lose. Let's see if you can beat being correct 76.9% of the time, being incorrect 9.8% of the time because the team you predict to win ties (treat games that go to KFTM as ties), and being incorrect 13.3% of the time because the team you predict to win loses. If you can do better than that, then you will have beaten how a very simplified version of the system FIFA does when tested against US Division 1 women's college soccer results.

My studies of a variety of ratings systems, as applied to the most recent year's games, suggests that those numbers are about as good as you can get for soccer. And believe me, plenty of experts have tried to do better. So, if you can do better than that, copyright your system and make a lot of money!

 

And that's a flaw of the system. U don't need to rank every country with a women's team who has been active in the last two years. A greater effort has been put into a system that produces a ranking for Spotsilvania than one that accurate ranks the top 24 teams. Even if Natasha were still young enough to be effective, Spotsilvania isn't going to France 2019. Ur arguing like a champ for reasons my local NAPA should stock Edsel parts

 

So based on the ranking I should have expected the following: Germany, France, England & Sweden face off in the Euro semi, US wins SBC, ToN & Olympics. If u were a betting man u would have had a tough 12 months. If I were FIFA Id get out of the rankings all together. Let the CONFEDs rank their teams and let the WC sort out the rest.

 

No, again you're simply mis-using the FIFA rankings based on what you want them to mean. FIFA rankings are not predictive, and nobody trusts them to predict winners. They're known to be only about 70% accurate as a predictor, which is astonishingly good on the one hand, but pretty bad in practice, i.e. you'll go bankrupt if you blindly follow them. So nobody does. And evidently, nobody has gotten rich over decades by plug-and-chug betting based on rankings. (From recent events, it's actually easier and more profitable for a cartel to hire/blackmail members of a team to shave points and throw matches, than it is for the cartel to simply predict the results outright. That's how hard predictions are, even when there's incentive to do it.)

Nonetheless, we still use data-driven rankings based on past results + simple formulas because it's just about the most acccurate methodology we know of. In decades of CS, math, and economics research, we've found nothing that guarantees better than that. (You could say that it's only the least worst of many blah ideas, and we'd happily agree with that.)

(I actually think that rating systems gravitate to a sweet spot between accuracy vs. cost. A public rating system should be fast and cheap, i.e. cost ~$0 to gather all needed data, and ~$0 in time, compute power, and electricity to do the computation. For that, we can indeed make rating systems that are ~70% accurate as predictors for ~$0 cost per update. Then you almost surely could improve on this a little bit by throwing more $$, hardware, data, time, etc. at it ... which is what we all suspect CIA/NSA, Fed/Goldman Sachs, Google, NASA, etc. are doing with their time. Look at FiveThirtyEight.com's vastly geeky big-data-crunched analyses of everything related to sports and elections , or xG (expected goals per shot), and you might agree that they paid more resources to get slightly better data. But even that doesn't help them predict winners often enough to make a profit by betting, and so they don't. So it's not strictly true that humans are limited to ~70% accuracy ... but it costs more to improve on it than you'd earn back by using it to bet or invest.)

You might actually be correct about a mythical system that acquires more data about a team's current state, and then extrapolates using that. But there's the cost of getting that data, which could put it out of reach for all of us. (DoD might do exactly this to wargame threat levels, e.g. I hope they are doing this right now, 24/7 with farms of supercomputers, to crunch every new sat photo of North Korea in nigh-realtime, and correlate the movements of individually recognized trucks to the probable readiness of a launch. Or something like that.)

Well, that's a philosophical difference. You suggest it's desirable to discard 110+ nations, and cater to only the top 24. (Maybe that's OK, like an NCAA Top 25 poll.) Most FI** organizations feel otherwise: their rankings acknowledge all of their members equally. It might be OK to do compromise and do both, i.e. use the FIFA rankings to get the top 24, then do the GIGA rankings for only them. I wouldn't mind that. Probably FIFA won't ever bother to do it for us, though.

I'm not the one arguing for change. I'm just explaining to clear up some misconceptions that I see written in this thread Remember, the default action is that we all move on from this, nothing changes, FIFA ratings continue unchanged for a few decades, and those of us who grok what they're for are content to keep using them as intended. I had no complaints about it; I'm neither pro nor con. It's just a mathematical tool, and it shall not be all things to all people.

In fact, I'm surprised you're so doggedly arguing this side now (in 2017). Are you just playing devil's advocate, or do you really have a better system in mind? Why now, and not earlier? We have threads that focus solely on the FIFA rating system (generally it gets bumped after every official update). That would seem to have been a much better time and place to raise any substantial objections.

I don't particularly care for the FIFA ratings either way (nor for FIDE's, FIVB's, FIS's, Pac-12's, nor any others). I'm neutral to them all, and I'm content to use them for what they are. They're not predictive, they're not current or instantaneous, they all make the weak assumption that past results are correlated to future performance, and they're all surprisingly good at actually predicting. And yet upsets have always happened, and always will. That's not a flaw in the ranking systems; it's a reminder to us that reality is complex.

 

Gilmoy, KT thinks KT does have a better system. It's the one that comes out of KT's head.

I gave KT a good challenge. Let's see if KT meets it or doesn't want to go to the effort. And, if KT doesn't want to go to the effort, which I think would be a reasonable response, whether KT is willing to say so.

But, for anyone appreciating this somewhat ridiculous discussion among KT, Gilmoy, and me, please understand that Gilmoy and I are total nerds (sorry, Gilmoy, if I've insulted you, it's really a compliment) who believe in the scientific method more than in our own subjective opinions.

 

...whereas this board, and maybe some places on the larger internet, is a place where subjective opinion holds sway to the exclusion, or at least diminution, of scientific argument.

and, if we bear that in mind, it's probably ok, because it keeps the discussion going. so, for example, i too would like to see

- kt's new ranking system

- puttputtfc's entirely new 23 woman roster of post 295's query

and, (not really, but)

- the uswnt lose 2 but play well in 3 games of a challenging tournament, just so that kt can be happier on this board (i'm really doubting that promise will be kept).