Scoring patterns in #WorldCup Knockout Phases 2010-2018

The Quarter Finals are about to take place at the 2018 FIFA World Cup.

I have been using a naive Bayes approach to anticipate when goals might be scored in the knockout phase of the 2018 tournament.

I chose some priors from the outcomes of the 2010 and 2014 tournament knockout phases. The posteriors for these were:

2010

2014

My priors for 2018 were:

The posteriors for the Round of 16 were:

A comparison of Priors and Posteriors after Round of 16:

It will be interesting to see if this relationship changes in the forthcoming games … particularly in regard of extra time.

Temperature, humidity and ball in play time at 2018 FIFA #WorldCup after 20 games

There is a rich variety of data available on the 2018 FIFA World Cup website.

The FIFA live blog for each game records temperature and humidity.

After 20 games played in the tournament, I thought I would explore these data with regard to ball in play time in each game.

The data and the RCode I used are available on GitHub. This post is another learning out loud approach to my use of R and RStudio.

Temperature and Humidity for each of the 20 games:

Humidity and Ball in Play Time:

Temperature and Ball in Play Time:

These ggplots are created with secondary data. As with all my World Cup posts, I am mindful that I have not investigated the validity and reliability of these data. I do make some basic face validity assumptions about these data.

Photo Credit

_IGP5474 (Victor,  CC BY-SA 2.0)

Using quantiles to explore AFL performance

I have been exploring the quantile function in R as a way to explore AFL performance.

The generic function quantile produces sample quantiles corresponding to the given probabilities. The smallest observation corresponds to a probability of 0 and the largest to a probability of 1.

I have data for 12 rounds of the 2018 AFL competition.

I wondered if the quantile measures might give me a game signature.

My data are:

For the teams that have won games

For teams that have lost games

For game outcomes

In Round 12, I looked at Geelong and Sydney in terms of winning quantile profiles (how good a winner were they?) and North Melbourne and St Kilda in terms of losing quantile profiles (what kind of losing team were they?).

My quantile signatures for these games are:

(Note quantiles are expressed in terms of winning performance for Geelong and losing performance for North Melbourne. Geelong’s final quarter is rated highly in winning terms (top 20%). North Melbourne were in the top 20% of losing performances in the third quarter.)

(Sydney’s first quarter performance was in the top 5% of winning performances. St Kilda elevated their losing performance in the second and fourth quarters to top 25% of losing performances.)

Discussion

Exploring quantiles in R has helped me think about how to visualise games and express the competitive nature of these games.

I am mindful that I am using a macro indicator for this visualisation but the ability to specify quantiles in a growing database of AFL performance seems promising.

They might also help with early prediction of game outcome.

Photo Credit

Footy (Beau Lebens, CC BY-NC 2.0)