Special Edition - World Cup Draw Scenarios

Now reproduced with permission at the '2014 World Cup Report' blog. Follow @MVFCLR22 or  @2014WCReport on twitter for ongoingcommentary of the 2014 World Cup.
----------------------------------------------------------------------------------------------------------

While we focus here mainly on Australian Football (of the VFL/AFL variety), we do take an interest in other codes as well.
And, tonight (Melbourne time) there will be the allocation of teams into groups for the biggest sporting event in the world - the 2014 FIFA Football World Cup.

Folks we follow on twitter have been pre-empting the draw using this fabulous simulator, and coming up with the odd grouping, and forecasts of joy or misery for Australia. We have seen these, and thought that one or two random run-throughs using the draw simulator is one thing, but what about building a bigger picture and more data, and try to predict what could really be the draw.

So, we have used that same simulator and conducted 100 simulations to determine the most likely opponents, and the most easy and difficult draws for Australia.


COMMON OPPONENTS FROM 100 DRAWS
With Australia placed in Pot 3 with other AFC and CONCACAF teams, we cannot be drawn against them. Instead, we are to be drawn only against teams from Pots 1, 2 and 4... possibly as below.

Australia's Likely Opponents From Pot 1.
Pot 1 contains the seeded teams for the world cup draw... with seedings based on the FIFA World Rankings.
The rankings are as determined by the governing body, to calculations similar to what we do here for AFL teams.

Being a random draw for this pot (other than Brazil automatically drawn into Group A), the 100 simulations undertaken performed very much to theory.  There is very little to separate the opponents, with an almost universal likelihood of meeting anyone from this pot.

The most drawn candidates though were Germany (16%), Switzerland and Colombia (15%)


Australia's Likely Opponents From Pot 2.
Pot 2 contains the teams from the CAF, and the last two CONMEBOL teams (Chile and Ecuador).
This pot is also affected by the FIFA decision to place all the unseeded UEFA teams together into Pot 4. This has meant that Pot 4 has 11 teams, and Pot 2 only 9 teams.

In our simulations, we only drew teams from Pot 2 90 times, as there were cases where a team from Pot 4 crosses over and is drawn in the stead of a Pot 2 team to face Australia in a group.

In this pot, there are other rules at play that prevent Chile and Ecuador from being allocated into the same group as a Pot 1 CONMEBOL team. Accordingly, there is skewing of the numbers here, with just under 1/3rd of the selections favouring Chile and Ecuador combined (based on the 57% of Pot 1 teams coming from UEFA).

There was no restriction on drawing CAF teams, so it is a little surprising to see a wider variety of selections (from 7 to 22%) given that it should be purely random.

Again, the three most drawn potential opponents were Ecuador (23% - 21 selections from 90), Nigeria (22% - 20/90) and further back, Cameroon.


Australia's Likely Opponents From Pot 4.
As before, there is some cross-over of teams from Pot 4 into selections from Pot 2 as there are 11 UEFA teams in this pot.
In our simulations, we found only 10% of the time a team came from Pot 4 as a 'Pot 2' selection drawn against Australia. This created a pool of 110 selections from Pot 4.

From this pot, other than the one team being drawn as a 'Pot 2' team, the draw should be completely random. And generally, our selections  in simulation found an even spread of opponents, with three outliers.
For some (random) reason, Australia

- attracted Portugal very often (12 times from 110 selection - 19%), and
- did not draw often Russia (6 times - 5%) and the Netherlands (9 times - 8%).

There is no logical answer to these outliers


GROUPS FROM 100 DRAWS
With the 100 draws completed, we also could examine the toughest and softest draws dealt to Australia.

The Path of Least Resistance. 
Using both the rankings points from the FIFA website (as of 6 Dec), and the ranking position, the easiest of the 100 draws, we could determine a rating tool for splitting the best and worst draws for Australia.
Adding the opponents ratings points* and also tallying the opponents ranking position**, the ease of draw (or group of death) could be calculated.

*    lower the number the better
** higher the number the better


The easiest draws we found using the total FIFA rankings points were...  

Working from the positional ranking tally total, the best Australian outcomes were

Three, The Hard Way.

Using the same system of totaled points, the groups of death for Australia were found as follows (remembering a higher rankings points total or low positional tally is unfavorable for Australia in this scenario).

And the 'Group Of Deaths' our 100 iterations found were (based on rankings points)...

Note... we drew Spain / Ecuador / Portugal twice!

The three other worst outcomes for Australia - sorted by total opponent ranking points were...

with a three way split for 4th and 5th places.

And it would be fair to say (as the realists that we are at the Institute) that whoever Australia draws at the World Cup, we will have a very difficult time against.

The 59th ranked Aussies closest opponent is Cameroon in 51st, followed by Nigeria at 36th and Algeria at 26th (excluding the CONCACAF and AFC teams). 

It is a steep mountain for the Aussies to climb in Brazil. Best of luck lads!

Competition is a Good Thing

Business and economic theory suggests that competition is a good thing... both for business, innovation and consumers.
We embrace the competitive spirit in football, where we have written many times about the need for fair and even competition, and even proposed a fair and balanced approach to the draw.

We also embrace the competition in ranking teams outside of the traditional AFL ladder framework. Our blog is not the only effort out in the market, with others such as Footy Forecaster, MAFL Online, the WayBeyondRedemption blog, Roby at BigFooty and Troy Wheatley's Power Rankings also attempting to rate teams realistically at any point in time.

To add to this we also know of others who do their own private systems, and we join in the MAFL online call for others to get creative and try their own method, as well as to go public with it too.

Create your own AFL team rating system http://t.co/EugBqOVgjw #AFL #stats #teamratings #football
— Tony Corke (@MAFLOnline) November 27, 2013

Competition and different players in the market drive innovation. Join us and have a go.


In a later post, we will further benchmark our 2013 data against others in the field, but for this post we found a recent update at WayBeyondRedemption that we thought we would run a comparison against now.

That post, "Model version 9 over the years", shows Micheal's model (version 9) applied to all seasons of VFL/AFL history.

The blue dots represent the absolute number of correct tips per season, and the red dots are the accuracy percent of those tips.
Note: For 2013, accuracy is ~72% based on ~150 correct tips from a 207 game season.

What you can see here is a very steady and consistent tipping regime of 60-70% accuracy year on year.


We were inspired by this post to produce our own chart based on our version 2 algorithms, and we could produce the chart as below.

Again, the blue dots represent the absolute number of correct tips per season, and the red dots are the accuracy percent of those tips.
Note: For our 2013 season, accuracy is ~71% based on ~146 correct tips from a 207 game season.

We are also striking in a similar zone, between 60 and 70% accuracy over the recent 40 or so seasons (since the late 70's). Prior to that there was a higher frequency of higher 70% to +80% accuracy on tips.
We suspect that the more recent lower tipping accuracy is due to the more equal nature of the AFL competition, particularly the 2000-2009 season period, where equalization was a key plank in the AFL's growth.

In this period, 7 separate teams won the flag and the divergence in percentage was also lowest over all teams in the competition. This is the most diverse and balanced decade in VFL /AFL history, and we found it the hardest to tip accurately, tipping right only 64% in that decade (compared to the overall number of 70%).
We have also highlighted the 2000's as a great ear in football on this post, as the renaissance after the preceding 2 decades that almost killed more clubs than just Fitzroy.


Histogramming the above tipping data produces the chart at right.

The blue bars represent the number of times we have tipped to a percentage accuracy.
The red bar is the median (middle point) for the 117 data points
Including season 2013, the average (mean) is 70.2%.
The yellow bar is the 2013 result... happily on the right side of the median and mean!

We have to also be concerned that our tipping performance is showing a declining trend over the 117 season collected data above... how to correct for that is another problem.

As we have mentioned before, our tipping is not only concerned with getting the right tip, but also tipping the margin well... with the aim of bettering all comers at the Monash University football tipping competition. This will be covered in a later blog, but suffice to say, the results are extremely variable.

Cheap Fills - Supplemental

* No, not those sort of supplements.

Another blog post on rankings points, this time a bit of system-theory, and fun with Richmond.

System Theory
Last week we posted a few notes on the VFL/AFL's best and worst teams, as we rank them, and how they aligned with the season premiers.
From this, we were asked on twitter about the gradual increase in the ranking of the seasons best team.

@troywheatley @Footy_Maths An interesting conundrum is the ever increasing ranking. Do teams get better over time?
— Lucas Garth (@lucasgarth) November 19, 2013

The data presented in that first blog post (and we have also done similar for the worst ranked team as well) can be graphically represented as per the right.

Here the trend to ever increasing rank points total is seen (with some waxing and waning) as teams move up and down game on game, year on year.

The blue line shows an increasing trend for the best teams to have a higher rating over time.

At the other end of the scale though, there is a less than clear trend visible. The chart is much more erratic as the worst teams drop to lows and the troughs are quite deep.
Nonetheless, there is, over the 117 seasons, a slight upward trend to that chart.


So plotting the best and worst teams shows increasing rankings points, but as outliers do they reflect the completion as a whole.

The below chart is the average of the teams rankings points for all teams and all seasons.

Here, the trend of increasing rankings points is seen clearly. We have also identified some of the wild changes and change points on the chart.
There are two key points on the chart that can be noted when teams enter, or leave the competition

1) New Addition Depression

As new teams enter, they start on 1000 points but usually their rank points allocation falls after the first season and beyond. As a result, a bigger pool of teams with new teams sub 1000 points depresses the average rankings points of the whole collective

2) Exit Surge

As an existing team drops out of the competition, there is a corresponding leap of the average rankings points the following season. This is evidenced by the three instances of teams dropping out of the VFL/AFL (University, after the 1914 season, and Fitzroy and Brisbane Bears*, after the 1996 season). The following season had seen an almost 50 point increase.
Obviously, a now smaller pool of teams, with lower ranked teams gone provides a boost to the average overall.

* We have considered in our system that the Brisbane Lions are a new team (entering at 1000 points) and that the Fitzroy and Brisbane Bears rankings entities ceased as of 1997.


Richmond
Over at the 'wooden spoon' blog post in this series, regular blog commenter and Tigers fan asked "how many times has Richmond been the 9th ranked team? ;)"

Well... I'm glad you asked!

Since the introduction of the AFL Final Eight system in 1994, Richmond (or as some call them... Ninthmond) have finished 9th on six occasions (in 1994, 96, 98, 2000, 06 and 08).

In those years we ranked them as 10th, 5th, 7th, 10th, 11th and 11th, respectively.
We have we ever ranked them as ninth once! Back in... 1999!