Introducing YTA Radars - Analysing Max Clark 29/03/2019

One of our new services is YTA Data Science. You may have seen some statistics being displayed on our Twitter account (for example, see this tweet), and our main focus has been visualizing the progression of our players using objective statistics.

Recently, we have been focusing on comparing players with their peers, using radar charts.

You might know radar charts from websites like StatsBomb or even from video games as FIFA. It’s a relatively simple tool that allows you to see a lot of information about a player at once. StatsBomb, and Ted Knutson in particular, have been using and tweaking their radars for years, so for more technical information, check out this article. We used those radars as inspiration and tried to build further on that, using our own tactical methodology.

For now, we identify 5 positions that have distinctly different in-game demands: Center Backs, Full Backs, Midfielders, Attacking Midfielders/Wingers, and Strikers. For each of these positions there are different things that you expect from them in terms of output, so you’ll need different statistics to judge them on. Some stats will be different, but with other stats we feel that they say something meaningful for every playing position, so a few stats you will see on each of our radars. For every position, the nine most relevant stats are displayed on the radar.

Basic Radar Characteristics

Before we dive in, let’s discuss some of the basics.
  • Peer Comparison: We compare players to their peers. That means we compare a Center Back playing in the Eredivisie only with other Center Backs playing in the Eredivisie.
  • Stat ranges: To eliminate the influence of outliers, the lower and upper bounds of the range for each stat are set at the 5th and 95th percentile respectively. That means if you’re at the very outside edge of the radar, you’re in the top 5% of all players of the same position in your competition. If you’re on the very inside of the range, you’re in the bottom 5%.
  • Stat clustering: We decide the order of the stats based on how they relate to each other. Attacking and defensive stats are clustered, and even within attacking stats there are groups: for example shots and expected goals are next to each other, as are passes to box and expected assists. Some stats that don’t show any correlation are kept from each other on the radar.
  • Repeatability analysis: Because we use these stats to evaluate players, we want to make sure that these stats actually say something about how the players are performing now, and how they might perform in the future. To that end we conducted a repeatability analysis and only included stats that have a sufficiently high R2-value, meaning that these stats can, with some degree of confidence, predict future behavior on the field.
  • Stat definitions: All stats are adjusted for playing time and calculated as per 90 minutes. Most of the stats speak for themselves, but there are some that might require some explanation. Defensive duels are 1v1 duels that involve physical contact. PAdj T+I are Possession-adjusted tackles and interceptions. These numbers are adjusted for possession because players of high-possession teams have less opportunity to tackle and intercept, adjusting for this makes it a fair comparison. xG is expected goals and xA is expected assists. xG/Shot is the average shot quality.
  • Is scoring highly on a stat necessarily good? Some stats are more stylistic in nature than they are used to evaluate a player’s quality. For example, having a high amount of defensive duels is not necessarily good or bad, but it does say something about a player’s style of play.

Let's look into a Radar - Max Clark

Max Clark, Vitesse’s left-back. He also shows high amounts of tackles and interceptions, relative to the amount of possession Vitesse has, and is not shy of a physical duel.

More impressively, he is in the top 5% of completed crosses and has high amounts of passes to the penalty area. He’s been valuable at both ends of the pitch.

His passing accuracy has been relatively low, probably due to the higher difficulty of his passes. Passes into the box are harder to complete than passes back to your Center Back.
Voor meer positie analyses en het bron artikel:

Bron: YourTacticalAnalyst / Foto's SV Deel online: