Predicting the finalists and semi-finalists of the 2020 Euros - using the ELO rating system and random forest machine learning algorithm
After nearly a year of delay, the UEFA Euro 2020 tournament is finally upon us – and the excitement at Idiro is real. Twenty-four teams will battle it out to see who will be the best in Europe. Things are finally heating up not only for the competitors but for the analysts, bookmakers, and average punters making their picks. Don’t worry we’ve got you covered. We’ve predicted the finalists & semi-finalists for this year’s competition.
Everybody has their own techniques for formulating a prediction. Whether it be extensive research of a team and player’s performance over a multitude of games or consulting their local marine life (see Paul the Octopus!). We haven’t consulted any Pundits for this prediction instead, we let AI & machine learning (prescriptive analytics) do the heavy lifting for us.We created a model based on the historical ELO rating performance of teams in the Euros.
The ELO rating system was originally devised as a method of measuring chess players’ ratings relative to each other.
The basic principle is that every player (training set) is given a score and whenever one player beats another they take a certain amount of points from the loser. The bigger the gap in the ELO rating, the bigger the point gain/loss, the smaller the gap, the smaller the gain/loss. Eventually, over a large enough sample of games, the future events should be obvious.
This rating system now applies to numerous games and sports, including soccer. Using the ELO rating, random forest learning model and several other factors such as rating changes over the past year, the number of home and away matches, and the number of goals scored, we’ve created a predictive model to find our picks for who’ll make it to the finals.
Our random forest model has predicted Spain and Portugal. Both teams are sitting at around 5th and 6th for odds of reaching the finals across bet making sites – which is not bad.
The ELO system does have limitations when it comes to team sports predictions, as it can’t account for individual performances or changes to the line-up, but it still provides a solid statistical background for making value bets.
So, tell us what you think in the comments below? Who are you backing for the finals and would you put the house on it?
If you like this post, here is a link to some more of our work.
Number portability has been important for the development of competition in telecoms. The conference addressed a wide variety of topics around the subject.
Here are some of the points that resonated:
According to one speaker, 75 countries have implemented number portability (NP) on their fixed (FNP) or mobile (MNP) networks.
Many others, including Jamaica, Trinidad, Afghanistan, Armenia, Togo and Tunisia are likely to implement number portability by the end of 2014.
Some countries, e.g. Russia, are struggling against technical and political barriers to implementing number portability
User experiences of MNP vary widely. In Portugal, callers to ported numbers are greeted with a message warning them that the call may cost more. In countries like Ireland, Ghana and Israel, mobile numbers can be ported in under an hour, whereas in some other countries it can take weeks.
In some countries (e.g. UK) the customer approaches her current network and requests porting (this is known as donor-led porting). Best practice, followed by many countries, is that the customer requests porting from the network to which they wish to port (recipient-led porting).
The technical platforms and processes underpinning porting continue to evolve, in response to customer needs (or rather operators’ new product opportunities), technical advances and the pursuit of efficiencies.
My talk to the conference covered three areas:
1. The evolution of in the importance of number porting
Mobile numbers will continue to be an important way to be reached by almost all mobile users, but callers can now find and contact at least some of their targets on social media.
Against that, the cost and difficulty of porting is now very low in most markets, so porting will continue to be popular for the foreseeable future.When truly portable mobile phones arrived (first for businesses, then with the advent of prepaid, for the mass market), the mobile phone number filled a need left unfulfilled: a simple reliable means of reaching someone anywhere, anytime. Porting was introduced to improve the free functioning of telecoms markets. In 2003, the value of porting to the Irish economy was estimated at £IR 129M.
More recently, social media has emerged as a far superior way to find and contact people. Although it has limitations, it removes many of the costs of changing the mobile number. However, in parallel the costs (monetary and service interruption) to users of porting continue to decline, and many operators incentivise port-in. Number porting is here to stay.
2. Insights based on analysing data around porters.
Idiro has analysed data relating to porting customers in a variety of markets. I presented a number of insights (anonymised, of course) on the characteristics of porters based on multiple markets. I also described in detail the phenomenon of porting contagion. The power of word-of-mouth results in many consumers following their friends when they switch networks. This accounts for a high proportion of porting overall. Big thanks to my colleague Lorcan Treanor for the analysis behind these insights. Please contact Idiro to learn more about these insights.
3. How Idiro SNA helps meet the challenges of porting churn
Idiro SNA is a perfect fit for the marketing problems around mobile number porting. Idiro scores can be used:
In Member-get-member acquisition campaigns. Idiro identifies the customers on competitor networks who share communities with phone users on the operator’s own network. The likelihood of these to port in is measured. For the most promising targets, the on-net friends are identified for targeting with a member-get-member campaign. This can provide very strong results.
In retention campaigns to reduce porting churn. This Idiro score is particularly popular with Idiro’s customers. Idiro runs weekly or monthly models to predict porting churn, and Idiro’s customers use these scores in automated weekly or monthly retention campaigns, as well as in other areas such as the call centre.
I was conference chairman on the second day of the conference, which focussed on Service Portability. There is great interest in the topic – where the customer can port not only their fixed and mobile numbers but other elements of their package as well, up to the entire quad-play bundle.
Though the concept is an appealing one, in practice the challenges are large. Imagine being a customer with a home phone, mobile phone, TV and broadband bundle, and moving it to a competitor. Every provider’s service bundle is different, and porting the entire bundle will require the customer (or the recipient operator) to make careful choices. In addition, speakers pointed out that the delay in porting different services will vary, so during a transition period the customer will have some services from the door operator and some from the recipient operator.
There are challenges aplenty there and it is clear that there is no consensus over the best way forward. One might (at the risk of overestimating the similarities) say that the discussion on service portability is where the number portability was 25 years ago.
Overall, the conference was well-organised and the speakers well chosen. However, like with many other telecoms conferences, the voice of the customer was hardly heard at all. Quality was mostly described in narrow telecoms terms, rather than the quality as measured by the user. Almost no primary or secondary research on customer experience was presented by regulators, operators or vendors. At the end of the conference (I missed one talk) I had not learned anything about consumers’ expectations for porting and how well they were being met.
If the voice of the consumer is not heard, how will their needs be met? It was ever thus in the telecoms industry – or at least, it has been for the last 25 years – and it is reason that OTT services like Whatsapp are eating SMS and MMS’s lunch. Despite being excellent in what it did cover, by its omissions this conference reminded me again of why the telecoms industry needs to cop itself on and develop a passion for the customer, or risk its share of customer communications being progressively eroded.