Tuesday, October 6, 2009

CONCACAF Qualifying - Matchday 9 Preview

It is that time again. Much should be decided this Saturday in the CONCACAF.

Here is the table:

In chronological order, here are the matches Saturday:
Mexico - El Salvador (6 PM Eastern)
Costa Rica - Trinidad and Tobago (10 PM Eastern)
Honduras - United States (10 PM Eastern)

The United States, Mexico and Honduras could clinch a spot in South Africa Saturday. For the US and Mexico it's a win and they are automatically in. For Honduras, a win plus a loss for Costa Rica would do. In practice they would also be in with a win and a draw for Costa Rica, though not mathematically. They are 10 goals ahead in goal differential, so finishing even on points with Costa Rica will put them through. Costa Rica can't clinch a spot Saturday but a win against Trinidad and Tobago or Mexico getting any result against El Salvador would guarantee the Ticos at least a playoff spot.

Since the last update, I've changed to an improved model, which I call the PLM. Its predictions have been shown to be more accurate than the Poisson model. Here are the qualification percentages based on the PLM.

Mexico's chances of winning in each of their matches dropped a bit according to the PLM compared to the Poisson model, so they fell from the team most likely to qualify to third best. None the less, they are overwhelming favorites. Keep in mind that the model uses only results and not injuries, see below for more on that, and it does not take into account that Trinidad and Tobago, El Salvador and possibly the United States will have nothing to play for the last matchday.

Match Previews

Mexico - El Salvador

Home Field Advantage

You can always expect it to be tough playing away in the hex. That is especially true in Azteca. In my view, Mexico playing at home has one of the biggest home-field advantages in all of sports. That seems true this campaign as they've struggled away from home but remained as dominant as usual at home: Mexico are 8-0-0 at home with 21 goals for and only 4 against. In the hex they are 5-0-0 with 7 goals for and 2 against. El Salvador have struggled on the road with just 2 wins, 1 draw and 6 losses in all of qualifying. Their wins are over the footballing powerhouses Anguilla and Suriname. Their draw is against Haiti. In the hex the El Salvadorians have lost all 5 matches away from home.

Form and Injuries:
It's a bit odd to talk about form for internationals, but it seems relevant because these teams have gone in completely different directions recently. On June 6th, El Salvador beat Mexico 2-1. Since then Mexico have done nothing but win. They defeated Trinidad and Tobago as well as the United States 2-1, they won 0-3 against Costa Rica and their last match was a 1-0 win over Honduras. El Salvador have struggled, but did pick up a win last month. After beating Mexico they lost 1-0 to both Honduras and Trinidad and Tobago. They then lost 2-1 to the US in Utah. Their last match they turned it around with a 1-0 win at home against Costa Rica.

I don't believe either team has any injury problems.

PLM Prediction

The PLM has Mexico winning 69.6% of the time, El Salvador 6.7% and a draw the other 23.6%. Due to the home-field discussion above and that El Salvador are practically eliminated, I think Mexico have a better chance than that - as high as 90% to win I'd say if I had to guess. The PLM says that 1-0 and 2-0 are the most likely scorelines. I'll say Mexico win comfortably, 3-0.

Costa Rica - Trinidad and Tobago

Home Field

Before the Mexico match last month, I could have written something pretty similar to what I did for Mexico above. After the 0-3 loss, Costa Rica now have a home record of 7-0-1 in all qualifying matches with 19 goals for and 4 against. In the final round their record is 3-0-1 with 6 goals for and 4 against. For the whole campaign T&T have a 2-2-4 record with 9 goals for and 15 against. In the hex they are 0-1-3 with 4 goals for and 11 against.


Trinidad and Tobago's results have been pretty consistently bad, Costa Rica appear to be in free fall. Since winning at T&T in June, Costa Rica have lost three in a row. What's worse, twice they were crushed and the other loss was against a relatively weak El Salvador side. They went down 4-0 at Honduras and got beat 0-3 versus Mexico before losing 1-0 in El Salvador.

To make matters worse for the Ticos, they look to be without midfielder Celso Borges due to injury and defender Gilberto Martinez due to suspension for skipping out on practice to go the beach! My source for that is the oft-cited Soccer By Ives.

PLM Predictions

The model predicts a winning probability of 75.8% for Costa Rica, 5.9% for Trinidad and Tobago with an 18.3% chance of a draw. To be honest, I think it's tough to say how far it is off based on the factors above. Costa Rica have been struggling and have injuries, but Trinidad and Tobago have nothing to play for. If I had to guess I'd say those things wash out and the predictions are pretty close. According to the PLM, the most likely outcome is 2-0 for Costa Rica. That seems like a good prediction to me.

Honduras - United States of America

Home Field

I'm repeating myself a fair bit here, but the advantage home sides have in CONCACAF is a point difficult to overemphasize. In all qualifying matches, Honduras have a perfect 8-0-0 record with 22 goals for and 3 against. In the four hex matches they have 12 goals for and just two against. The US have a 4-1-3 record in all qualifying matches, but 3 of those wins are against Barbados, Cuba and Guatemala in earlier rounds. In the final round they have 1 win, 1 draw and 2 losses with 5 goals for and 7 against. Despite that record not looking stellar, it actually is good compared to the other teams. Getting a point per match is a big reason that the US find themselves top of the table and on the verge of clinching a trip to South Africa.


There's not much to write about here. Since the two teams played in June, Honduras have won 3 out of 4 matches with the only blemish a 1-0 loss in Mexico City. The US have done the same, winning their other three matches and losing by a goal in Mexico.


The big injury story is that the US will be without Clint Dempsey who has a shoulder sprain. It is up in the air who will replace him. For updates on that situation, I again recommend Ives's blog.

I'm aware of no injuries to Honduras.

(Edit: I missed two suspensions for Honduras - midfielders Amado Guevara and Danilo Turcios. Both are starters. Guevara should be familiar to American fans as he is a regular for Toronto FC in the MLS and played the whole game in the Honduras 2-1 loss in Chicago. Turcios came on as a second-half sub in that match. I should also have mentioned that star forward David Suazo will be returning from injury and though he will almost certainly play, may not be 100%. Thanks to 2+2er mmbt0ne for the corrections.)

PLM Predictions

The model predicts a 53.8% chance of victory for Honduras, 14.8% for the United States and 31.5% chance of a draw. The most likely outcome is a 1-0 win for Honduras. I personally think the loss of Dempsey is a pretty big blow for the US. Because the model does not take it into account, I think the Hondurans' chances should probably be higher and the Americans' lower. I'll predict a hard-fought 2-1 win for Honduras.

Overall Prediction and a Look Ahead

My picks, and those of the objective PLM, point to Mexico clinching Saturday while the US, Honduras and Costa Rica are forced to battle for the last two automatic spots next Wednesday. Saturday night check back here for an update and a preview of Wednesday's matches. In the meantime, I'll be posting similar previews for the other confederations.


  1. Do you know why exactly the PLM has Mexico as the third best? It doesn't make any sense to me given that: 1) USA, Mexico, and Honduras all have one home and one road game left, but 2) Mexico has, by far, the easiest remaining schedule of these three teams.

    Intuitively, it would seem like Mexico is the top-three team with higher odds of qualifying directly. The final part of your post seems to suggest that you share that impression. So I just don't get why the PLM disagrees...

  2. Thanks for the comment.

    This confused me as well, to be honest. I made a post after the last matchday with Poisson predictions, which have Mexico as the most likely at 94%, 8% higher. I actually expected the PLM to rate Mexico even higher because they are significant favorites today and the Poisson model tends to underrate big favorites.

    I will study this more later because I think it is weird and a possible sign that something is wrong. When I was writing the article I compared the predictions for each Mexico match and they were just a bit less likely to win each one than the Poisson predictions. I think the PLM does significantly better at individual match predictions, which in turn helps the simulations. However, the data I used to make that assertion, and the coefficients for the PLM are from league data. Maybe these coefficients should be different for national teams. It's something I'll look into tonight and tomorrow using data from the last World Cup qualifying campaign. It's also possible that different models will be more effective in one type of competition compared to another or even for different leagues (the PLM uses data from various leagues)

    Here's a link to the Poisson predicitons:

  3. I'm ashamed to admit that it was a programming issue. I suppose that's good as it means the model is fine.

    After fixing the program and running it again pretending like Saturday's matches hadn't happened, it gave this for finishing in the top 4:
    Mexico - 93.22%
    USA - 92.18%
    Honduras - 88.77%
    Costa Rica - 25.51%

    These numbers are in line with what you thought.

    Thanks again for the comment.