WCQ: Third Round - Window 3 (November 12-16, 2021)

[Scottie]

3 hours ago, jonovision said:

 

Canada 2 - Mexico 2 in Edmonton (October 2008) also seems like a result where the weather surely must have helped. How else does a starting midfield of Stalteri, Bernier, Harmse(!) and de Jong (replaced by Pozniak(!!) for most the match) hang with Sven Goran Eriksson's Mexico. Not sure of the state of the group at that point.

I was there for that one. The weather ended up being pretty mild for mid october. I'm pretty sure I got by with just a sweatshirt, toque and voyageurs' scarf. Contrast that with Grey Cups that I've been to at that same stadium where it was much colder, where frequent bathroom breaks were necessary to keep warm even with a full winter coat on. Of course it was still quite cold for the Mexicans, I recall Gio dos Santos in particular running down the wing in front of us with long sleeves and gloves. That was our best result of the whole round and it happened after we were eliminated a few days prior.

[WestHamCanadianinOxford]

14 minutes ago, Scottie said:

I was there for that one.

My wife happened to just mention that game. It was her first Canada match; apparently she posted about it and it showed up in her memories on Facebook today.

As mentioned the weather was pretty forgettable so probably mild enough for someone living to that point in central/northern Alberta.

I shamefully know I have lost my winterization, so I am bringing lots of layers for my first game in Canada forever.

[jtpc]

More sanctions from FIFA, this time for El Salvador's next home match on November 12th.

Just putting it out there: Should anything similar happen in the El Salvador/Jamaica match next month, the disciplinary commission would have quite a bit of time between November and the January break to render a decision for El Salvador's first home match in the new year (against us in February). 😂

 

Edited October 16, 2021 by jtpc

[CanadaFan123]

So Mexico definitely has to play their January home match behind closed doors. Or do they hope everyone forgets in a few months?

[ardsa]

47 minutes ago, jtpc said:

More sanctions from FIFA, this time for El Salvador's next home match on November 12th.

Something that could be interesting: If anything similar happens in the El Salvador/Jamaica match next month, and the disciplinary commission has rendered a decision by the January window, El Salvador's next home match is against us in February.

I'm not sure we have to wait that long for poor fan conduct.
The just finished October 13th match vs. Mexico had laser pointers, water bottles and other projectiles raining down. Perhaps FIFA will implement a fan ban (full or partial) for the Feb 2nd match based on the conduct of the last match.

[RJB]

15 hours ago, theaub said:

I don't see how needing to beat Costa Rica would result in looking past Costa Rica?

 

There's a lot of talk about rotation and about needing all six points. I think that taking it one match at a time and simply ignoring the Mexico game until we're past Costa Rica is the best approach. 

[footballfreak]

Rather than saying we need 3/4/6 points from this window, if we look at things holistically the most realistic pathway to qualification is:

3 points from the Costa Rica home game

3 points from the Jamaica home game

3 points from either USA or Mexico 

If we do those things, we only need 1 point from the 4 away games to hit 20 points. Even that may be enough to finish 3rd. If that draw becomes a win, we’re at 22 points which will almost certainly be enough.

Obviously, it would be preferable to win all our home games and avoid road shenanigans, but chances are something goes wrong in at least one of those games. But if we can get 9 from 12 at home, we’re still in very good shape.

Edited October 16, 2021 by footballfreak

[scullion44]

Just saw for Panama- El Salvador that Panama wont be allowed to have fans. Big chance for ES to do us a favour.

[Cblake]

What are people thinking regarding the kickoff times for the games in Edmonton, will they be positioned for east tv with a say 5-6 pm local time kickoff? That would be nice considered the possible weather, the earlier the better in my eyes, will be interesting to see what considerations prevail in picking it.

[rkomar]

6 hours ago, maplebanana said:

Cool.

Considering your original message disparaging the simulations, I would argue you have no clue how these simulations work to begin with. They do not just say 'team A is in position x, therefore will finish position x'. But go ahead and think whatever you want. 

They do say that 'team A is in position x, therefore is most likely to finish position x'.  The simulations are based on the teams' current record, and it is expected that the teams will perform the same for the rest of the tournament.  A week ago, Jamaica had 1 point after four games and had about 4% chance of qualifying based on the simulations.  Two games later, they had 5 points and an 8% chance of qualifying.  Clearly, the probability distributions changed greatly in those two games.  I think that's Unnamed Trialist's point, the probabilities are not static, but change from game to game.  The basic idea of the simulation is fairly bogus.  Rerunning the simulations after each round of real results is also "wiping the shit off one's face".  In Jamaica's case, the 4% prediction is wiped away and replaced with 8%.

[Unnamed Trialist]

9 hours ago, Floortom said:

The homophobic chant just happened in our last match with Mexico numerous times…are they going to be punished too?

I have a suspicion that since Mexico plays its next round away, they'll just sit on any decisions regarding sanctions. Which favours Mexico, again (eg longer appeals process). 

I mean, the ref stopped the match. FIFA can't pretend it didn't happen, a ref stopped a match following FIFA protocols and they have to investigate. If they can decide on such specific sanctions vs Panama and ES so quickly they could do the same for repeat faults at Azteca. They have to for some as it affects ticketing for November. But since no match is coming up at home for Mexico lindo until late January, they can fiddle with it. 

Edited October 16, 2021 by Unnamed Trialist

[narduch]

3 hours ago, Cblake said:

What are people thinking regarding the kickoff times for the games in Edmonton, will they be positioned for east tv with a say 5-6 pm local time kickoff? That would be nice considered the possible weather, the earlier the better in my eyes, will be interesting to see what considerations prevail in picking it.

It's 705pm local time both matches 

[footballfreak]

3 hours ago, rkomar said:

They do say that 'team A is in position x, therefore is most likely to finish position x'.  The simulations are based on the teams' current record, and it is expected that the teams will perform the same for the rest of the tournament.  A week ago, Jamaica had 1 point after four games and had about 4% chance of qualifying based on the simulations.  Two games later, they had 5 points and an 8% chance of qualifying.  Clearly, the probability distributions changed greatly in those two games.  I think that's Unnamed Trialist's point, the probabilities are not static, but change from game to game.  The basic idea of the simulation is fairly bogus.  Rerunning the simulations after each round of real results is also "wiping the shit off one's face".  In Jamaica's case, the 4% prediction is wiped away and replaced with 8%.

There’s a whole lot wrong to unpack here, and I don’t know if I have the energy to do it after two years of arguing with anti-vaxxers, but here goes.

First and foremost, the simulation of remaining games is not based on current record, but on elo rating. For example, Canada’s most likely finish after match day 5 was still 3rd, despite the fact we were in 4th in the live standings. If we’re getting even more pedantic, I don’t think the guy who produces the model even uses updated elo rankings each round, but rather a static set from before the ocho began. So no, the odds of each individual game don’t change afik. Even if they did update elo rating each round, tweaking the odds based on newer, better data is a feature not a bug.

Second, the fact that the probability of qualifying fluctuates as the model is updated with results is not some gotcha. That is what is actually happening irl. Canada winning against Panama altered both teams’ respective chances of qualifying. Every team will ultimately end up in one of two bins - in or out. You will either be at 0% or 100% qualified. The results teams achieve throughout qualifying, and the results of other games, will begin to shift them closer to one bin or the other. All this model does is try to accurately reflect how close each team is to each bin. You can argue it doesn’t do an adequate job, or factor in enough variables, but you’re not gonna have better accuracy with some static crystal ball.

If Mexico has a 65% chance of winning their next game, the probability of them qualifying afterward will have changed irrespective of any other result - because you can’t 65% win a soccer game. The pendulum is always swinging. The plinko balls are always falling left or right. Pick whatever metaphor you want, but if you’re not factoring in current data into your predictive model, then it has no more value than a magic 8 ball.

Third, Jamaica was a particularly poor choice of example, because their changing fortunes in reality are quite well reflected in the model. Before the start of game 1, they were given a 10.3% chance to finish in the top 3, and a 23.8% chance to finish in the top 4 - with ~60% of outcomes putting them in the 5th to 7th range. This more or less passes the sniff test. You can argue they should have been slightly higher or lower, but most assessed them as being something of a wildcard that could fall anywhere in the 3-7 range depending on how well their new players bedded in. 

Anywho, they then had a bad run of games which significantly lowered their chances (go figure), then rebounded slightly when they outperformed the average number of points they were predicted to win in Honduras. None of that is surprising. Ultimately, their odds are currently better than they were before game 6, but worse than at the start of qualifying. Again, that passes the sniff test - they went from being on zeros with everyone else to being in a point deficit, with fewer games remaining to overcome it. 

I’m gonna go to bed and trust someone else will be able to explain the finer points of Monte Carlo models and Poisson distributions way better than I ever could, but in closing...

[An Observer]

7 hours ago, rkomar said:

They do say that 'team A is in position x, therefore is most likely to finish position x'.  The simulations are based on the teams' current record, and it is expected that the teams will perform the same for the rest of the tournament.  A week ago, Jamaica had 1 point after four games and had about 4% chance of qualifying based on the simulations.  Two games later, they had 5 points and an 8% chance of qualifying.  Clearly, the probability distributions changed greatly in those two games.  I think that's Unnamed Trialist's point, the probabilities are not static, but change from game to game.  The basic idea of the simulation is fairly bogus.  Rerunning the simulations after each round of real results is also "wiping the shit off one's face".  In Jamaica's case, the 4% prediction is wiped away and replaced with 8%.

Clearly, none of you play cards or bet at the casino (at least successfully)

[CanadaFan123]

10 hours ago, jtpc said:

More sanctions from FIFA, this time for El Salvador's next home match on November 12th.

Just putting it out there: Should anything similar happen in the El Salvador/Jamaica match next month, the disciplinary commission would have quite a bit of time between November and the January break to render a decision for El Salvador's first home match in the new year (against us in February). 😂

 

Questions:

Was the Panama match stopped at any point for homophobic chants? If not why is there an outright ban? If anything fans running on the pitch should be punished. 
 

Apparently homophobic chants by a select few is punishable with no fans. But a select few throwing projectiles at players just means no fans at either end. 
 

All this to say it’s clear CONCACAF prioritizes feelings over the physical safety of players. And all this to play in a World Cup in a country where being gay is illegal.

You can’t make this stuff up. 

[Floortom]

Utterly bizarre objections to these ELO based simulations - "The probabilities change with actual events!!"

[sstackho]

3 hours ago, footballfreak said:

If we’re getting even more pedantic, I don’t think the guy who produces the model even uses updated elo rankings each round, but rather a static set from before the ocho began. So no, the odds of each individual game don’t change afik. Even if they did update elo rating each round, tweaking the odds based on newer, better data is a feature not a bug.

FYI - the model runs "hot" and does update the Elo ratings on the fly based on the results of every simulated outcome.  This adds some additional volatility, but probably not all that much in a relatively short tournament. 

[Unnamed Trialist]

Guys trying to give us lessons about Monte Carlo Simulations, the level sometimes.

All I know is they were originally conceived to help with the particle physics involved in developing the atomic bomb. If you just scratch at the surface you realise these folk are taking something of great complexity and trying to drag it down into their mud-wrestling event. I mean, maybe one of the easiest applications involves programming sensors, and even then. 

I believe there are people on the board who are able to handle the math behind Bayesian inference, that I can believe. And likely even some more able to handle the differential equations you may need for its application. I can't.

Then I read that Enrico Fermi was called in to sort out the math for them back in Los Alamos or wherever they were. But no problem, the MC simulation party boys getting on our case because, as I understand it, they are saying us naysayers aren't up to the level of Fermi.

Oh, fuck me, what dunces we are.

[B_Leaf]

40 minutes ago, CanadaFan123 said:

Questions:

Was the Panama match stopped at any point for homophobic chants? If not why is there an outright ban? If anything fans running on the pitch should be punished. 
 

Apparently homophobic chants by a select few is punishable with no fans. But a select few throwing projectiles at players just means no fans at either end. 
 

All this to say it’s clear CONCACAF prioritizes feelings over the physical safety of players. And all this to play in a World Cup in a country where being gay is illegal.

You can’t make this stuff up. 

How dare you trying to make sense or connect dots.

[Redpunkfiddle]

15 hours ago, jonovision said:

I don't think anyone is claiming that the climatic edge will be decisive, just as playing at midday in San Pedro Sula isn't the only reason we've lost there in the past. And more than most, I'm skeptical of how cold-hardy the Brampton National Team really is when faced with proper prairie cold (which we might not get; +15 would be just as likely as -15 on either day). 

But you'd be foolish to deny that it doesn't give Canada at least a little edge, especially for the Mexico match where Canada will already have been in Edmonton getting acclimated for more than a week.

I can accept that it might be a small advantage, just not comparable to the significant disadvantage we have playing in sub-tropical extreme heat.

 

[rkomar]

I actually was a nuclear physicist in school and afterwards, and spent a lot of time writing and running Monte Carlo simulations.  I don't need any lessons in what they are, or what a Poisson distribution is (although I could use an explanation as to why Poisson distributions might be used here when they are only meant to apply to cases where the probability of something happening is very small).  The problem isn't that the predictions of these can change with new results, it's that they can change so much.  If Jamaica's chance of qualifying can go from 4% to 8% in two games, what did that 4% really mean?  Were we really supposed to believe that Jamaica only had a 1 in 25 chance of qualifying, which is pretty hopeless?

We as fans know what we want the numbers to mean, I just question whether they are really what we want them to be.

[shorty]

1 hour ago, CanadaFan123 said:

And all this to play in a World Cup in a country where being gay is illegal.

Second one in a row, isn’t it? Or does Russia just claim that no one is gay in Russia?  Also, the Qatari stadiums are built by Nepali and Bengali slaves essentially.  I must admit as jazzed as I am for our team and its progress, the Qatar World Cup is objectively a poisonous reminder of all things wrong with FIFA. (sigh)

[SpursFlu]

I found it odd that our game against Mexico was stopped once again because of some apparent puto chant but I watched the Mexico player take a pk in El Salvador as objects were flying past his head and no stoppage or anything. I guess names like puto will break my bones but aluminum cans lobbed from 100ft bouncing off my skull will never hurt me

[Craig11rm]

All I know is at this level the margins are thin and the games are played for a reason so I take any simulation or prediction with a grain of salt. Way too many variables to take in to account. 

[frattinator]

5 hours ago, footballfreak said:

I’m gonna go to bed and trust someone else will be able to explain the finer points of Monte Carlo models and Poisson distributions way better than I ever could, but in closing...

Going to try to take a stab here. 

So without looking under the hood, what the model appears to do is to use ELO ratings as an estimator for each team's probability of win when they face each other. Now, you can argue that ELO ratings are not the best representative of the teams' relative strengths (and better if you can suggest a better alternative), and I would have time for that argument. But if you can assume (and I think plenty of people smarter than me have done the due diligence on the use of ELO ratings in many different sports) ELO ratings represent a decent estimator, than you can use that as the basis of some informative models. 

What Monte Carlo simulation does is taking advantage of the Law of Large Numbers and Central Limit Theorem and treating each match up as an individual trial with three outcomes. More simply, if you can assume ELO ratings do a good job of estimating the relative strengths of each squad, then you can generate a probability distribution of the results between every single pair of teams. However, since these are probabilistic, each single sampling might not represent the most likely case. Imagine El Salvador played Canada at BMO field, a random sample might generate an outcome where El Salvador came out with a win, but that would not be representative of the expected result between Canada and El Salvador. Monte Carlo simulation gets around this by making tens of thousands of samples, and using Law of Large Numbers (which basically states that the average result of a large number of trials converges to the expected value (in this case, ELO ratings), so that we can see based on the ELO rating what the most likely result is. What Monte Carlo simulations do even better however, is using the thousands of simulations that it ran and generate the probabilistic outcome of what result was. It's as if you can watch Canada play El Salvador 1000 times, Canada might win 500 times, draw 300 times, and lose 200 times. You can now use this one result (50% of 3 points and  30% of 1 point) and aggregate it to other similar matchups between other teams to understand how the overall table looks. Once again, without having access to the code, I don't know how they generated the priors for each matchup (my assumption is they used the relative proportion of the ELO ratings to generate the win and draw probabilities of the multinomial distribution, or even just a generalized Bernoulli's distribution and sample from that for each trial), but assuming they are not donkeys I can't see how they can screw it up that badly. 

As to the criticism that ELO ratings are being updated between each trial, I think there is a misunderstanding of what the model is designed for and the conclusions you are trying to draw. If the goal is to create a predictive model PRIOR to the start of the Octo and see which teams are likely to qualify, then yes, using updated ELO ratings would be inappropriate. But in that case, what you are really evaluating is how good of an estimation are ELO ratings PRIOR to the Octo final standing. However, the model here doesn't try to do this, and instead it's trying to take all available information to predict who, after match day 14, will be occupying the top three spots in CONCACAF. In addition to using the updated ELO ratings, it also takes in account of the fact that there are fewer matches for teams to play, so the uncertainty bound for the final result also shrinks each after each match. Complaining about this model's approach with that goal is akin to complaining about election models that incorporate polling information during the campaign, which is unreasonable in my mind. Now, you can make an argument that we shouldn't treat each match as completely independent draws out of a probability distribution, especially since things like yellow card accumulation, fatigue, or injuries could have an impact on future matches. But no model can account for everything, and that's the world we live in.

One final thing, it's important to remember that all models are wrong, but some are useful. Probabilistic models provide likelihood of events, they are not deterministic and shouldn't be used as soothsayer on your future Qatar travel plans. But you can use that to start planning on when  you want to setup your flight ticket sale notifications (for example, when Canada hit 95% 😉).