Gamasutra is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.


Gamasutra: The Art & Business of Making Gamesspacer
View All     RSS
July 24, 2021
arrowPress Releases







If you enjoy reading this site, you might also want to check out these UBM Tech sites:


 

(User Research) AI designs in Battle royale games

by Yongcheng Liu on 06/18/21 10:19:00 am

The following blog post, unless otherwise noted, was written by a member of Gamasutra’s community.
The thoughts and opinions expressed are those of the writer and not Gamasutra or its parent company.

 

First of all, let's start with the definition of battle royale games. Its original name is actually tactical competitive game, which is named after the famous Japanese movie "Battle royale-Survival Program".

It’s called “chicken-eating game” in Chinese, mostly because of the reason that the top game in this category-PUBG, after survival from 100 players, there would a sign of “winner winner, chicken dinner” to celebrate the winning. This originally means winning the title in cock fighting. Cocks who lose the cockfighting will basically die on the spot, and can even be eaten as dinner by the winner.

After that, "Knives out", "Fortnite", "Apex" and some other br games were developed with different versions based on the core mechanism of battle royale game, and even the mode of battle royale underwater appeared in the market.

This is also related to the reason why the origin of battle royale games was introduced earlier. Any game under this category of br requires a great number of players to stay concurrent online at the same time. why? For example, each game requires 100 players and if 100 games need to be matched at the same time, then at least 10,000 players need to be online concurrently. Once the number of online players drops, the match time and game diversity will be affected (which may often be matched to play against the same players).

On the other hand, battle royale game will always be a fair competitive game. Just like League of Legends, any skin you buy in the store, no matter they are epic, legendary, or just ordinary ones, will not affect your controls and technic level in-game. Your control of champions depends only on personal ability and experience. The same goes for battle royale games, especially when most br games are FPS or TPS, the fairness ion-game will be more prominent. Compared with the sale of guns and grenades with different bonuses in terms of damages, the commercialization of the battle royale game will only be limited to character or weapon skins,  other stuffs that do not have any bonuses in-game. This also gave birth to Battle pass to a certain extent (originating from Fortnite). From this perspective, if a battle royale game is not played by many players, its commercialization is bound to face a huge failure, and this failure will come really fast.

 

In this situation, combined with the research of new players by different game companies, AIs (that is bots) in battle royale game appeared. Although it may not have much to do with commercialization, AI plays a very important role in-game in terms of fast matching, training newcomers to get started quickly.

 

This article will mainly introduce how European and American players perceive AI in battle royale games, and what their demands for AI look like.

 

The first part is the perception of AI.

Currently, most battle games played by European and American players are actually a handful of them. They are all popular games in the market such as PUBG, Fotonite, Apex, and CoD: Warzone. After our research, we found that players' perception of AI in these games also tends to be similar. Details as follows:

1.1. AIs are easy to spot

In most battle royale games, the existence of AI is extremely easy to recognize. Especially in Fortnite, the bot will even look at the sky for a while for no reason. It will attack back only after being hit by the players, and the attack back comes at a particularly slow rate; in addition, in Fortnite, the existence of construction function allows the bots to be recognized easily.  Because AI basically does not build and may shoot a few shots after encountering people, and seldom builds to make shelters to retreat or attack.

 1.2. A similar behavior  pattern

This is more common in players' AI evaluations of PUBG, and the same is true for Fortnite. The main similar activity patterns are: 1) Not able to utilize cover. When real players encounter enemies in-game, they will usually shoot first or go directly into the cover, or just shoot and then use cover repeatedly. In any case, real players will use cover more frequently and skillfully. However, AI will hardly use cover, even if it finds that it’s at a disadvantage or is severely disabled, it will not find a cover to heal. They basically do not retreat. Even if they receive much damage, or in the circle, the AIs will not retreat with a high probability. 2) Unintended movement between open spaces and buildings. This purposeless movement may be manifested as a slow forward and backward running without target; the sprint direction is not the path that a real person would choose (such as sprinting against the wall). 3) Weak shooting ability. When the AI ​​sees a person or is hit, they may attack back, but the shooting rate appears to be quite low. Basically, they cannot cause large-scale damage to real players, and they will be defeated eventually in most cases. Of course, this is related to the original intention of game companies to invest in AI. They probably do not want AI to be much better than real players, especially when players have not established a complete understanding of the game at the beginning; in addition, once AI settings are much stronger than real players, and at the same time, its deep learning of AI is not perfect, AI will have a great impact on players’ gaming experience. 4) Basically do not use grenades

1.3. Does not carry advanced equipment

In most battle royale games, based on the difficulty of AI combat, the equipment they carry is bound to be not very good. Otherwise, there will be situations where you can get advanced equipment without any big effort. And this is against the essential logic of the battle royale game: the final victory is achieved by constantly searching for advanced equipment and high-level skills. High-level equipment is often in the hands of a few people, and these high-level equipments are distributed in various places on the map and picked up by players. After these players are killed, the advanced equipment is transferred to the killer. Therefore, the possession of advanced equipment and the number of kills are generally positively correlated. Obtaining advanced equipment by defeating low-skilled AI will fundamentally violate this logic.

On the other hand, if AI's machine learning training can reach a level that is closer to real players, and the difficulty of killing becomes relatively higher, then the equipment in their hands will be correspondingly better.

 

1.4. The place of appearance is relatively fixed

The place where AI appears in the battle royale game is basically set relatively fixed. For example, on an empty space, a PUBG player would suddenly encounter an AI; or in a building, a Fortnite player encounters an AI that rushes into the door while looting. These are the more common locations for AIs to show up.

 

The second part is how players feel about the current AI design.

2.1. Fighting with AI is not difficult at all, with low accomplishment in the same time

European and American players basically have the same view on the AI in battle royale game: that is, the difficulty of combat is low, which is also related to the lack of advanced equipment with AI as described above. As long as you encounter them, you can kill them, unless you just stand with no fight back. This setting is not a good phenomenon for some strong PVP players and strong competitive players. When they find bots can be killed very easily, this will have a negative effect on their sense of accomplishment in pursuit of controls and skills improvement in-game. Especially in the face of players who want to continue to learn combat skills, cover using skills, AI can provide them with very few training opportunities.

 

2.2. Important way for newcomers to become familiar with the game

At the same time, most European and American players would think that if they were newcomers of a game, they would hope that there is some AIs in this game, and they would even accept the first game or the first three games after the tutorial part, the existing opponents All are AI. This has great benefits for them to be familiar with game mechanics, master the equipment levels.

2.3. At the same time players can get a certain sense of battle participation and killing

The presence of AI in the early stage of the game can give players a certain sense of participation in battle. Whether for those who choose to camp or choose to fight positively, after collecting a certain amount of equipment, encountering and killing enemies can enhance their combat participation and enjoyment of defeating enemies. Presumably, many players have encountered a situation where they landed on a relatively far place when they were playing the battle royale games. After looting for a while, they did not encounter anyone, and they had already reached the finals without killing anyone.

 

2.4. Basically they can accept the existence of a certain amount of AI

As mentioned earlier, in terms of new gamers who want to be more familiar with the game and shortening the matching time, European and American players basically accept the existence of a certain amount of AI. Generally speaking, although AI sometimes seems a little awkward, it will have a positive effect on the game most of the time.

 

Then the third part is expectations for AI design.

3.1. Deep learning of real player activities, empowering AI with a variety of different behavior combinations, and increasing the difficulty and diversity of combat.

As mentioned earlier, European and American players can easily identify the bots in battle royale games such as PUBG and Fortnite. It is also because they have very fixed behavior patterns, and every real player in these games may have a different activity which is different from AIs. Therefore, the use of machine learning and deep learning to empower AI, after learning a large number of real players' activity data, will largely solve the problems of easy identification, fixed behavior patterns, and low combat difficulty.

3.2. For different levels of games and different stages of the game, put AI with different levels and abilities.

For different stages of the game, such as just landing, running into the circle, etc., European and American players have different demands of how AI appears; in addition, in the ranked mode, the existence of AI is generally not accepted unless they are well-trained Ais with a higher level of control ability and more suitable difficulty of corresponding ranks.

3.3. Even though it is in order to shorten the matching time, 10% of the AI is acceptable to most players.

As mentioned earlier, the existence of AI is necessary for battle royale game players to shorten the matching time. But generally speaking, it is acceptable that about 10% of players to be AI, and more Ais will affect the player's feeling of accomplishment in killing and skill improvement.

 

The above are conclusions obtained in this research. The main research methods are public opinion analysis and large-scale in-depth interviews with players.

 

 


Related Jobs

Bytro Labs GmbH
Bytro Labs GmbH — Hamburg, Germany
[07.23.21]

Lead Game Designer (f/m/x)
Bytro Labs GmbH
Bytro Labs GmbH — Hamburg, Germany
[07.23.21]

Senior UI Artist (f/m/x)
Insomniac Games
Insomniac Games — Burbank, California, United States
[07.22.21]

Senior Designer
Legends of Learning
Legends of Learning — Baltimore, Maryland, United States
[07.22.21]

Senior Systems Designer - $150k - Remote OK





Loading Comments

loader image