vorosmccracken.com

Whoops, I took a quick nap there. I’m up now so I’m back. Let’s see what time it… HOLY CRAP! SIX YEARS???!!!

Well yeah, that’s a little too long a layoff. Anyhow, curiosity got the best of me and I decided to fire up the old soccer neurons and use a bunch of my new found skills and technology to update my soccer ratings and projection system.

To recap, I generate ratings for all the National Soccer teams based on past results, and then use these to predict future results. I then simulate the remainder of qualifying 10,000 times to estimate the chances of various teams qualifying for the World Cup. Same thing I did back in the day, but I’ve updated the technology and my own skills a bit.

The new rating system uses something called a Cumulative Link Model which essentially works as a custom distribution allowing increases in units/levels that are not equidistant. For our purposes, the reasoning is that in soccer goals the difference between zero goals scored and one goal scored is effectively not the same as the difference between one goals and two goals and so on. Yes it’s ‘one goal’ either way, but the ways and reasons teams score those goals differ in the two scenarios.

There actually are other ways to get at this (like poisson distributions with an additional parameter), but for now I’m using this way. In the future I’m going to experiment with others.

I won’t go too much into further detail now, especially since the model will undergo significant adjustments from here. But I wanted to get this out since the crucial games happen in just a few days.

Three of the most interesting major teams are the USA, Argentina and the Netherlands. Each has seen headlines that qualification may surprisingly be in danger. The data shows that the Dutch are in far worse shape than the other two and are major longshots to qualify at this point. While the Americans and Argentines still find their chances near 90% despite struggles.

In CONMEBOL, while 5th place only gets you a playoff, most CONMEBOL teams should handle New Zealand fairly easily, and that goes double for Argentina. Despite their struggles, Argentina still looks a very good bet to be there in Russia in 2018.

In CONCACAF the US situation is a little better than it looks on first blush. While wins in the final two qualifiers doesn’t technically guarantee qualification, effectively the tiebreakers are such that it would. Given the matchups, the most likely point total for them is 6, followed by 4, followed by 3. All of which give them good to decent qualifying chances. Furthermore, they would be either slightly favored over Australia in a potential playoff, or heavily favored over a Syrian team with no home leg in would would be a political nightmare of a playoff. And so their qualification chances are still a very good 89%.

In addition to the teams who have already officially qualified, Germany and Uruguay qualified in all 10,000 simulations. Uruguay finished 5th in 14 sims, but won all 14 matchups with New Zealand. Germany won their group every time except two, and won both playoffs. Would take a major miracle for either not to qualify now.

Finally some assumptions that aren’t necessarily safe, but are hard to otherwise model. Panama and Honduras will be facing Mexican and Costa Rican teams that don’t have much to play for. Under normal circumstances these would be challenging matchups, and they might still be under these. Players will still be playing for a spot on next year’s roster, so just laying down seems unlikely. So the assumption is that the ratings won’t be affected any more than normal variations affect them. Not a safe one, but hard to model.

Another assumption is that the South Africa v Senegal match that ended in a 2-1 win for South Africa will be replayed (as is currently scheduled by CAF). In case you don’t know the referee in that game has been banned for life for match manipulation of that match. A ball that struck a Senegalese player’s knee, was ruled a handball and the decisive goal in the match was scored by the resulting penalty.

If either of these assumptions prove inaccurate, qualification chances can change quite a bit, and drastically in the second case. If the match stands, Burkina Faso, the Central African Republic and South Africa would all see their chances increase at the expense of the group favorites, Senegal.

Anyway, sorry for being away so long. More to come, including an explanation of what I’ve been up to.

— Voros