07 Jun England Won’t Win World Cup 2018. Or will they?
According to Gracenote, England have only a 4% chance of winning the football World Cup this summer.
That doesn’t sound very good.
But what does a 4% chance of winning really mean anyway?
Should England even bother?
Or is there more going on here? (hint: there is)
OK, but how can we get any practical value from stats like these back in the real-world?
So what does 4% really mean?
It means if they ran this World Cup competition 100 times, we’d win it 4 times.
We’d win it 4 times!
Actually that doesn’t sound so bad.
It gets better.
Put another way, England are basically guaranteed to win this World Cup every 25 times it is played.
It gets better still:
(OK, I think I know what you’re thinking, but just go with me for a minute …)
This World Cup is the 21st …
so not that far off 25 …
And England haven’t won for a while …
so that means …
OK OK. I’ll stop.
But sometimes we can get our stats mixed up like this, especially when we have a desired outcome we’re looking for.
What the stats are actually saying is that, over the long-run and after multiple running of this competition, 1 win in 25 is the likely outcome for England.
I’m sure you’ll know, that there are a couple if big issues with my hopeful analysis.
The easiest issue to see is that the 4% chance of winning isn’t consistent across the 88 years of the competitions.
It only applies for THIS World Cup – the 2018 version.
For other World Cups we’d have had a greater or lesser chance of winning, depending on the other nations form, etc etc.
Scrub that then.
The bigger issue is that we can’t run THIS World Cup 100 times or even 25 times.
So where does that leave us with these stats then?
How can we possibly get any value from them then?
Actually there is quite a lot we can learn.
But we need to dig a bit deeper.
With stats like these we’re looking for things like:
1- What is a sure thing (or not)?
2- What can we infer?
3- Can we work out how the stats were generated and use that knowledge to our advantage?
Firstly. What’s a sure thing? Well …
1- Gracenote really has no idea what is going to happen (Sorry guys!)
According to Gracenote, Brazil has a 21% chance of winning.
Firstly, as an aside, 21% also doesn’t sound that good (sorry Brazil fans.)
But more importantly, 21% is actually a rubbish “confidence interval.”
Simply, this is how confident the prediction is – higher the better.
21% chance of something happening is simply not a strong enough predictor.
If you find yourself faced with some stats and the confidence interval is at this level, be very cautious.
Despite my profession as a “data guy”, as a sports fan, I think the fact Gracenote don’t know is actually a lovely situation.
Sport is meant to be unpredictable, right?
So are these stats useless then?
Lets look a bit deeper.
What can we start to infer?
2- What if we look at the stats compared to each other instead?
There are a couple of things here.
Lets start by comparing England with the chances of all the other teams.
Luckily for England fans, it doesn’t look that great for other teams either!
In fact, of the 32 teams
England is 8th favourite to win!
8th out of 32 sounds a lot better than 4%, well, I think so anyway!
I feel better already.
Stepping it up a level:
What about the relative overall chances head-to-head.
So Brazil are 21% and England 4%.
Comparing them shows Brazil is five times more likely to win than England.
5x more is a lot.
What about next best Spain on 10% with England?
Comparing them shows Spain is two and a half times as likely to win than England.
Not great (for England) but a lot better.
What about Brazil and Spain?
Again Brazil is twice as likely to win than Spain.
It still doesn’t look good for England but by comparing the stats of the other teams, it starts to give a bit of comfort many are in the same boat, as we try to forget that 4% feeling.
Stepping it up another level:
What if we used those relative overall chances, as a direct head-to-head measure?
So this would mean:
Brazil are 5 times more likely win a match against England –> 84% chance Brazil wins.
Spain are 2 and a half times more likely to beat England –> 71% chance Spain wins.
This is possibly quite a leap.
Maybe not …
Go with me.
So., if Brazil played Spain it would translate to a 68% chance Brazil would win over Spain.
In other words, if Brazil and Spain played three times, Brazil would win twice and Spain once.
Remember that: if they played 3 times.
Ok it’s a bit tenuous, I’ll give you that.
But short of any other data, and as long as you know I know, lets just see where it takes us.
So if you calculated all the head-to-head percentages, this is what it would look like:
Brazil still seem to be out right favourites, compared to the chances of everyone else.
But, that low 21% is nagging.
If Brazil are so good, why isn’t the prediction much higher?
I averaged Brazil’s chances against all the teams.
It shows they have on average a 90% chance of winning!
21% just seems stupid.
Or is it?
3- So how were these stats created and why is Brazil not more of a favourite?
By looking at the rules or the “context” of the competition, we can start to understand.
The World Cup has Group stages and then 4x Knockout stages.
The Group stages have multiple games.
Given my newly inferred 90% winning average against the other teams, you can (more than) reasonably expect Brazil win most games, and easily make it through to the knock out stages.
In fact, so much so that, it would be a major shock if Brazil did not make it through.
The Gracenote stats for the Brazil in Group stages, further down the page (linked again here), bears this out.
Gracenote thinks Brazil have a 90% chance of progressing through the Group stages …
But we could be on to something …
Aside: This only started out trying to understand that 4% and look what happened.
Ok that 90% is a great level of confidence.
At that level stats can be really empowering for your decision-making.
We should now have LOTS of confidence in Brazil winning their group – and sadly the bookmakers are on to us!
In fact, if I was a Brazilian supporter I’d be lobbying FIFA to change the World Cup format to be one big group competition and scrub the knockout stages!
They won’t of course … (thankfully – me, a Sports fan!)
We’ve therefore great confidence in Brazil sailing through the group stages.
What about the knockout stages?
This is where things start to become tricky (for everyone actually).
The issue is that ANY loss and the team is out.
Furthermore, it is also not clear who Brazil will play, as they progress through each stage.
I can feel things about to get complex.
Yes, a little.
Gracenote say they have run millions of combinations, for each team.
But put simply:
– In a group competition your odds are AVERAGED.
– In a knockout your odds are COMPOUNDED.
It is the compounding that causes the issue.
So if you’ve a 50% chance in one match.
Then a 50% chance of winning the next match.
To get your chances of winning BOTH matches you need to times those chances together.
50% x 50% = 25%
So compounding means you’d only have a 1 in 4 chance of winning both games.
Put another way,
You’d win those two games in a row, once every 4 goes.
Not great hey.
Anyhow, Maths lesson over!
Remember back to those three games between Spain and Brazil?
Brazil would win two and Spain one.
Well there are 4 knockout rounds.
4 games in a row.
The odds say that if Brazil played Spain in each of the knockout rounds Brazil will lose one or more those games.
Eek (Brazil fans)
Brazil not looking so good now!
Brazil are not going to face Spain four times.
But, they are going to face other strong competition.
Competition where the odds of winning, whilst really good (see table) are still not 100%.
It is the compounding that gives us low overall odds for all the teams.
We explore this further in this next post.
I’ve even made a little spreadsheet which I’ll give you to try with your team.
It even updates as you move though the stages …
But for now, let’s conclude.
4% sounded like really duff odds for England.
Yes it’s still not looking great, however, we’ve started to dive into the stats a bit more.
We’ve discovered that it really doesn’t look good for anyone – even Brazil.
We’ve been able to infer some likely head-to-head stats, which will be interesting to explore.
We’ve also discovered the reason why the predictions are so low is to do with the nature of the competition – Group stage competitions are much easier to predict than knockout.
Check out the next post where we’ll start to look at:
– The sensitivity in the predictions, and what this means to you.
– More on how Gracenote have got their stats, including the assumptions they’ve made.
– How you can translate all this into something actionable, say if you were Gareth Southgate.
– Plus, we reveal who we believe WILL win the World Cup.
P.s. Like this article? Then sign up below to be the first to hear more insights:
Pace Insights Ltd. ©2018 All rights reserved.