Living flag game mode

[farooqkz]

In this game mode, the flag is carried by a bot or an NPC/mob. Each team has one of these bots which...

• Carries the flags
• Has 100 HP
• Walks between different stations
• Has a sword and when in danger fights with enemies which are too close
• Has unlimited frag nade and throws it at enemies which are close but not too close
• Has a magnum and when it takes a ranged shot, it shots back precisely at the shooter
• On death its team lose.

This bot is always walking on a pre defined route between the stations. Each station...:

• is a 5x5 indestructible cobblestone
• Heals the flag carrier while close to it
• Has a teamchest

The bot is always moving between stations unless in the middle of sniping. If its HP is less than maximum, upon arriving to a station, it stays there till it fully regenerations.

When team members die, they respawn at the current location of this bot.

I believe this game mode could be an extension of classic with no magnum, assault rifle and maybe shotgun. We could have bandages to heal both teammates and the flag carrier. Obviously, new maps with stations and pre defined routes should exist. Algorithms for walking, finding the way and fighting need to be researched.

An alternative game mode with similar game play would be a train with automatic turret or nade launcher.

Also see my comment below.

[farooqkz]

Flagwars game mode has been always suggested and accepted as an alternative to Minecraft's bedwars(see #953). We could have this mode as our own creative version of flagwars. We don't need to clone Minecraft games, we can go beyond.

This game mode has very high complexity for implementation. However as the first version, we could omit, reduce or simplify the fighting part. Sniping with magnum doesn't require much effort. We just calculate where the shot has come from, point to that and shoot.

However throwing nades and fighting with sword requires high level of efforts if we want to have a savilli level fighter. If we really want that level of skill, we have to research this using Machine Learning methods. Then we'll have two hard challenges: First we need good performance, second we need optimized code on CPU. Having an ML model which requires like 1GB of RAM and a GPU for inference is just out of options. Not that it's impossible. It's just not worth it.

I am a Genetic Programming researcher and I think having a code which fights well and also is light on computing requirements is feasible. But requires research, at least months of research. I would instead suggest using a combination of heuristics for the fighting part. We could have simple heuristics for the sword fight. And omit nade throwing or provide a very simple/naive one.

Also a good question is about mob vs bot. I think a mob would make the server laggy as (I think) the mob code has to run on the same thread/process as the server code. If we do a bot, we could even host it on another server. Bots are doable thanks to people at minetest-rust. Last time I tried, their code works with little modifications.

[farooqkz]

@AnonomousAccount has suggested something similar on Discord and it is very popular with only a single downvote. It is here. The text for reference:

Suggestion: A mode where, instead of a flag, there is an item with a certain amount of health that you can hit with your sword. Once you take all the health of the enemy's ""flag"" you win

[farooqkz]

See a comment here about offloading some computations to external library, and then maybe to another machine:

luanti-org/luanti#16176 (comment)