We will automate the data analysis of PES! Part5

■ Introduction

Thank you Yajun. This article is an article ** that creates "analysis software for PES" with the slogan "e-sports" x "engineering". If you are interested, please also check out Back Number ~

The software in this article is written in Python. The referenced sites are summarized in ■ Reference URL.

■ 2020 e-sports national polity

In PES recently, a team has come out to practice for ** Kagoshima National Athletic Meet in 2020 **.

The national polity adopted ** 2on2 ** format! Two players control 11 players. In PES, it is also possible to operate 11 players one by one with 11 players ♪

An important skill in multiplayer is ** "Movement when you don't have the ball" ** In soccer, it is called ** off the ball **, and it is being studied as an important element in actual soccer. The increasing importance of off-the-ball in games may mean that the game is closer to real football! !!

▼ I want to study off-the-ball in games

I am also one of the players practicing for the National Athletic Meet. Looking back at the practice video, I suddenly thought while watching the player's off-the-ball.
"That's a hassle !?" *
It's hard to see 22 players moving fluidly. .. .. That's why I wanted to make off-the-ball research easier with software.

■ Purpose

To develop software that makes it easier to look back on the ball.

■ Design

The purpose is too rough, but what I want to know is "location information of each player". If you can extract only the location information, you will be able to achieve your goal!

▼ Data

The input data is ** "match video" **. The output data is ** "trajectory of player and ball" **.

How to track a player's trajectory from the above data! ?? Use ** "Radar Map" ** on the match screen. ** [■ Results> ▼ Input data] in the chapter below (https://qiita.com/junya0001000/items/8c5479b9fa084bcec307#%E5%85%A5%E5%8A%9B%E3%83%87%E3% From 83% BC% E3% 82% BF) **, the radar map is smaller in the lower center.

▼ Processing flow

- ① "Data noise removal" *
- ② "Recognition of target object" *
- ③ "Integrate location information in the time direction" *
- ④ "Data output" *

■ Results

Yes! Let's actually analyze with the completed software!

▼ Input data

It is a video that the left side is easily broken from the opening and the goal is conceded.

▼ Output data

◆ Video with only markers and balls extracted
Even with this alone, it is easy to analyze because only the player's position information is extracted from the original video!
◆ Image that tracks the trajectory of the marker Oh beautiful! Like fireworks!

Only the ball is blurred because the marker of the player who became the ball holder changes. It looks like Title of Kimetsu no Yaiba, and I personally like it! Lol

◆ Overlaid video Yup! You can track it properly!

▼ Consideration from analysis results

Now let's consider why we lost points in this scene.
"No, it's because the right SB has disturbed the line! Lol" "It's too easy to grow grass! Lol" *

However, it is a phenomenon that occurs just before the goal, and is not a "factor of the goal". There seem to be various factors, large and small, but there are two factors that are noticeable.

** ① "There is no marker on the opponent's left OMF" ② "The left CB is not calling out the timing of raising and lowering the line" **

** The former ** is the task of this right WG or right DMF, both of which are unmarked as ball watchers. As a result, it is difficult to take a position on the right SB.

** The latter ** has the control of the line in the CB near the ball holder side, The intention to raise the line is not conveyed to the right CB and right SB. The act of raising the line increases the defensive strength, but it increases the space behind it, so be careful.

▼ Narrow down the necessary information

By extracting only the information you want to know, "the trajectory of the player", the analysis became easier. Also, when you have a reflection meeting with your teammates, it seems that you can have a smooth discussion if you reduce the amount of information only to the subject you want to discuss.

** This might be even better! ?? Comments like this are welcome! ** ** Since it is developed as a personal hobby, there has never been a review from design to testing, so it may be out of sync with demand. .. ..

■ End

How was it? This time, I started to analyze videos. If the number of multiplayer players in PES increases in the competition scene, I think there will be demand. Even in other games, if the game emphasizes placement and movement, I think this kind of analysis is essential.

The next task is to return to the data analysis report creation because the tester data has been accumulated.

■ Small words

The other day I clearly read my article and saw a service that would have imitated the idea lol The principle is rolling as much as you want, but it was a "service for PES" lol I don't have any rights, but I wanted a word. .. ..

I don't like the effort, so I'll write the article sparingly on technical topics. However, the URLs of the referenced technologies and sites will be summarized in Reference URL ♪

Stay tuned for the next article!

■ Reference URL

** Hough transform is interesting so I recommend it! The guy who thought absolutely metamorphosis! ** **

I tried to automatically extract the movements of PES players with software