Andreas Kyzis is a Cypriot football analyst currently at Sportradar, a global sports betting services company based in Gera, Germany. I came across Andreas’ profile through his work at Football Radar - a football data company I’ve been interested in for years (and even applied for a job there once myself!).
The other thing that immediately stood out was his career switch from chemical engineering and the food industry to football analysis! Since making this change, Andreas has worked in Bulgaria and Germany and with two English clubs.
I sat down with Andreas to find out more about how he got involved in football, advanced stat metrics, and the connections between betting and football.
The interview has been condensed and lightly edited for grammar and clarity.
[ Your background is pretty unique, how did you get started working in football? ]
I played football until I was 18 and finished school - I wasn’t a good player! (laughs) I enjoyed playing & watching football and I always had a passion to do work in football somehow but I studied chemical engineering in Greece and came back to Cyprus when I finished and worked as a quality control inspector in the food industry.
I was there for 4 years but I felt something was missing and I always wanted to get involved with football. Initially, I thought about coaching but it wasn’t right for me. Coaching requires very good communication skills, you have to start the lowest youth levels and well, I was looking for something different. At the time, analysis roles were very popular, many people wanted to do this and many teams were employing analysts, so I started doing a course in sports analytics online. It took 8 months, after I finished it, I quit my job.
I was unemployed for about 6 months looking for something related to football. I found an opportunity in Bulgaria as a match analyst, so I decided to go there!
[ How did your friends, family and colleagues react to your decision to change careers? ]
They told me that I was crazy! I could get many jobs in Cyprus, not just chemical engineering and get a good salary. I was making €600/month in Bulgaria, even if the expenses were not so high, I could get maybe €1800/month in Cyprus. But I went anyway.
"They told me I was crazy! But I went anyway."
[ What kind of qualifications are needed to work as an analyst? ]
A lot of people working in football clubs have a ‘serious’ degree - not only a course like I did but a post graduate diploma or even a masters degree (specializing in analysis) which you can do now in many universities, especially in England and around Europe.
So there’s two main paths, one you can get a proper education (e.g. 4 year degree) first or, you can do many courses like I do. This can be cheaper and save time.
Aside from this, some clubs will look for some coaching qualifications. It’s not always so important but some clubs will want you to have at least the lowest level of coaching licenses. It depends which department you might work in.
Currently I’m finishing a course where I’m learning some fundamental skills in programming. We're learning python, SQL and Machine Learning, mostly for data visualization, extracting data for analysis and even creating your own analysis models. It’s difficult but I think it’s very interesting and I would like to do more of this.
[ What were you responsible for as a match analyst? ]
I was watching live matches mostly and we were assessing both teams performances in terms of how many dangerous situations occur during the game.
For example, when a team is attacking, we are finding the most dangerous situation at all times. Usually this is a shot, but sometimes it’s actions before the shot. I had to assess the most dangerous point (i.e. the cross, pass or shot) and then the ‘danger level’ and enter that into the software.
We were (usually) giving the data we collect and analyze to betting companies.
[ How does this relate/differ to expected goals (xG) and other ‘attack’ metrics? ]
Expected threat (xT) is the probability to score at any time.
If there’s a great through ball and the striker is one-on-one with the goalkeeper, but takes a bad touch and doesn't actually end up shooting, there's no expected goal metric because there was no shot in the end but obviously that's still a great chance. So that's where this expected threat comes.
—^interviewer’s note: interested in learning more about xT? Check out this useful overview/guide from Ishdeep Chadha
We were also assessing the finishes in the same way. Similar to expected goals on target (xGOT).
[ How did your job change after switching to player analysis? ]
We were assessing the impact of every player during a game. So after a game, we can say which actions were good and bad. How good or bad they were affects the probability of their team to win.
In match analysis we care only about one thing, the highest level of danger in each location, but now we’re recording everything!
We assess every action of that player, not only in possession, but also out of possession. Maybe the most important and the hardest to assess was the players off ball & defensive abilities.
I think, even in the top companies, it's very hard to evaluate how good or how bad a player is without the ball because we need to address their positioning in relation with their teammates, opponents, ball location, tactics, etc.
[ Who was ‘using’ this player analysis? ]
Also (like the match analysis data), many betting companies will use these stats and ratings. Maybe they need to know, if a player is not going to play, how we rate them and the impact on the team. If Real Madrid plays Man City and Vinicius Jr. is not going to play, it will have some effect on the prices (odds) - and this depends on each player’s ratings.
But we give (sell) the stats we collect to clubs to help give them an advantage in the transfer market. We might find a player who is very underrated and nobody knows him so his price is very cheap, but we assess him as very good. We would give that information to the team and maybe they will have a look at him.
[ How many matches did you watch on average? ]
It was too much! (laughs)
On a weekend, we were watching at least 6 live matches, 3 on Saturday and 3 on Sunday. Maybe then on Monday or Friday where many leagues play, or a cup game during the week.
It was probably at most 10 games a week.
[ How do you see the field of football analysis changing in the future? ]
It will only grow and grow, maybe it will never stop growing. Video analysis, scouting analysis, data analysis, I think all of them will grow and there will always be something new to add.
Every season, every year the connection with analysis and the teams will become stronger and stronger. The funds are there.
Many people don't trust the numbers now or they don’t think it will be beneficial for them, you know they say, “Oh, they’re just numbers! Who cares about numbers?” Their opinion will change, I think. It’s already changing.
—^interviewer’s note: there’s ample evidence for slow adoption of stats (advanced or otherwise) assists weren’t officially recorded in the Bundesliga until 2004 (!)
Analysis will continue to go further and further and discover new metrics and more accurate metrics. I think many people want this and so much money is invested that it will only go higher.
But, the other problem is to analyze it correctly. Just taking the numbers doesn’t mean much. I think the most important thing I see on social media and the sites I follow is that every serious data analyst is focusing on the context of the numbers. A number can show you anything so the difficult part is to have a better view about what the numbers are showing and communicating this.