World Cup 2014 Predictions and Results | Hacker News Reader

World Cup 2014 Predictions and Results

bloomberg.com

46 pointsdanso40 comments

Now I wonder if an octopus would be more reliable this time.

That's a whole lot of 1-0s in their "projection" of the knockout stages, and not a single game in the entire tournament where a team scores more than 2. Not at all realistic.

Yes, goal average is about 2.5, very likely one of the weaker teams gets 3 or more against.

You're missing the point. They are showing the most likely scores. Take a look at the betting for any of the matches and you won't find the most likely correct score being 3+ for any side.

Of course, taking the entire set of games into account, there's likely to be some games with higher scores, but when you consider each game individually, as this web site does, the likely result will not have many goals.

Then what is the point really? You could figure all this out by looking at the odds, which are probably more useful then this info-graphic.

Well, yes. But the point of these pages is to show opposing views and how different people and different systems come up with the chances.

Personally I love the mathematical complexities involved in trying to come up with odds/chances for events. It's applied maths with the potential for monetary reward :)

It's much easier to look at than digging through pages of odds. I enjoyed the color and convenience.

This is the single most likely outcome, but this exact outcome is still very unlikely because of the sheer number of possible outcomes.

Nahhh.

This would be my prediction if the WC was a long series of randomly defined games. The specific games calendar has effects, though.

My own model goes from the Elo ratings at http://eloratings.net (with Elo, a pair of scores gives you an expected result) and simulates 10^7 World Cups, updating scores for each simulation as it goes and so on.

There aren't any clear favorites, although of course there's an elite group (IIRC 5 or 6 teams together have 25% of likelihood of being champions) and Germany is slightly ahead. But in the modal scenario Brazil is eliminated by Spain in the eighth-finals.

Not sure updating the Elo ratings throughout the simulation is a good idea. The rating should reflect the 'optimal' estimate given the current information. A simulated result is not actually new information.

Nice visualisation but assumptions seem to be based on past performance in world cups. The Netherlands is supposed to have the 6th best defense in the competition, which is ludicrous.

> assumptions seem to be based on past performance in world cups

Not even that: England has a 35% chance of winning their match against Italy, which only has 33%. Overall stats in the competition are twice as good for Italy, as you can see at http://www.bbc.co.uk/news/world-25233859

Wonder if the host advantage is taken into account (not that Brazil need any kind of advantage).

As last time France hosted the world cup they won, yet they were not the best team in the tournament.

I have a similar model for league matches, and the home advantage usually works out to something like 0.1 goals per match. It's important in estimating the odds, but rarely tips the balance with respect to who the most likely winner is.

"Soccernomics" claims iirc that the home stadium (not exactly the same as home country) advantage is a full 1 goal

Why they were not the best team as they won the Euro just after that ?

Hey, one of the Bloomberg developers here. Thanks for the comments.

The prominent most likely score is a bit confusing, if you click through to the match detail then the histogram shows a range of goal differences. So for the first match we predict a goal difference of >2 to Brazil, although the highest probability for an actual score result is 2-0 (16%). Our model makes projections for up to 8-0 in both directions, but we aggregate them for display in this interactive.

This is awesome but the predictiveness is quite unpredictive. I guess that's what happens when you quant everything

Or it's what happens when you look at data in the simplest way possible. We know there will be upsets, we know some players play better together than others, etc. This could have been enriched to show something quite interesting.

On a side note, personally, I think England will surprise everyone by making it to the finals.

I wish I could share your optimism about England.

You mean out of the group stage? That wouldn't be too much of a surprise; oddschecker.com has England and Uruguay about equally likely to qualify, with Italy a bit ahead. And if Luis Suarez isn't fully fit, England have every chance.

Current odds for England reaching the final are about 14/1, i.e. about 6.7%. So it would be a surprise!

Odds to reach the quarter finals are roughly 7/4 or about 36%.

We shouldn't forget about Costa Rica too. They finished above Mexico in qualifying, they have a decent team and they will probably have a slight climate advantage in the games at Forteleza and Belo Horizonte (vs. England).

I wish I could also share the optimism about England. If we do get out of the group stages we may have to play Brazil or Spain in the quarter finals.

I'm Uruguayan, and I think it will be very hard for us to get out of the group stage, and if Suarez isn't fully fit, it will be even harder.

We're fielding a slow team, with several veterans who are at a speed disadvantage (I wish Lugano would retire), and a midfield more suited for counterattacking than for ball dominance. Our strikers, if fit, are the best duo in the tournament though :) .

I think Costa Rica will be the kingmaker, whomever falters against Costa Rica is out.

and be beaten by Portugal, on penalties, no less.

England will make it just like every other World Cup except 1966.

Favorites win every game! Try doing your March Madness basketball bracket like that and you will not get very far. Saying the 1st and 2nd seed teams will be in the final is not a reasonable prediction.

But that's how predictions go, same as the betting odds.

It doesn't mean a lesser team can't win or a bigger team not loosing all their group games (as France did one time)

It's possible you're thinking of a previous event I've forgotten, but presuming you're referring to South Africa four years ago, France actually lost 2 and drew 1 - sorry for pedantry :)

that's alright, was actually thinking of 2002 when they did the same (lost 2 and drew 1). Anyway, they did badly

No upsets is much more likely than any other prediction. Just because you know there will be upsets doesn't mean you know which specific upsets will occur. In fact, it's probably more likely that no upsets occur than one specific set of upsets occur. It's only because there so many ways for upsets to occur (think about combinatorics/entropy), that you're virtually guaranteed to get an upset.

If you're driving down the street, do you think you're more likely to see an ordered license plate (123456) or an unordered license plate (163542)? Obviously unordered plates are more common. But if I ask: are you more likely to see the specific plate 123456 or the specific plate 163542, they are equally likely (even though the second plate has more 'apparent' randomness).

To go an entire World Cup without conceding a goal is unprecedented, and I would expect the probabilities to reflect this. I seems that the probabilities are calculated independently.

They aren't claiming that. The page is just showing the most likely outcome for each individual game.

Brazilian here. The brazilian politics already bought the World Cup. It's 100% chance for Brazil to win. Mark my words and make some bets.

This would be more interesting if people made and posted their own predictions, which this graph lets you do!

I wonder if they analyzed the Messi factor correctly against Spain...

It's funny to see all scores being 0-0, 1-0, or 2-0... ;)

Reply on news.ycombinator.com