Posts from Tanta Simulator

The bot defeated the first 5 ghosts with Tyrax. SMOrc

https://www.twitch.tv/videos/617792611

Watch at the end; the beginning is just a match.

I watched the end of the stream. Impressive, congratulations. And through what software interface does your bot interact with Heroes? I'm also working on developing AI in Heroes, although I'm taking a much less promising path. I'm just very interested in doing this. I'm simply trying to improve the original AI while preserving its style. That is, if we take the analogy with chess, to raise it from the initial level of play on a Dendy console to the middle level.

I don't know what should surprise me more, conquering the ghosts or an Necromancer playing HotA?)))) Well, now let's conquer the factory! SMOrc

Integrating this into the game would be a separate creative task :)
I thought – maybe something like this would be useful in my project after all. Let's say, so that AI heroes could also effectively manage their stacks on the battlefield.

But then, I probably wouldn't want the bot to be trained "too well," because that would contrast too sharply with its actions on the map. ))
Is it theoretically possible to create a bot that is "trained, but not too well" – and at the same time, universally, in general for combat? (or not quite universally, but for a wide range of possible combat situations).

In general, it would be a very interesting task to train in a situation of confrontation with an intelligent opponent – a human. That is, training based on the example of a human playing against a human (or a human playing against himself).

Started testing the workshop using a guide.
https://www.twitch.tv/videos/625851558
https://www.youtube.com/watch?v=GOuzFVMKuwI
The discrepancies started on the second turn.
Also tried to exploit the Graveyard with one of the hero builds, using Elves. A similar discrepancy occurred around turn 5.
Since I don't have the motivation to learn HotA and come up with exploits, I'm asking the community:
is what I'm doing potentially useful to anyone? Of course, you can answer privately on Discord, rather than officially announcing a desire to create exploits.

I'm not interested in HotA as a game version; there's little use for the bot in that mode. If a group of people gathers (I'm counting on 3+ people, at least one of whom knows how to program and can continue writing scripts), we'll create a cozy Discord where we can work together to develop the bot into a few "solid" scripts that work not only on my computer but also as a normal program that everyone can run.
Then I can promise to continue developing the core and create a normal program with a UI, rather than something that's launched from an IDE.
If not, then I've satisfied my own interest. So that it doesn't just sit on my computer and get lost in the abyss of history, I've uploaded it to GitHub under the MIT license (meaning you can steal it, modify it, and even not remember that I started it).
https://github.com/Ihromant/hotaabuse/tree/master
That's all for now.

I looked at both discrepancies. The first thing that comes to mind is that Deemer has a Meteorite, which neutrals usually take into account. As for Shiva, it's quite possible that the +speed for birds played a role, but that's not certain.

Via screenshot. The first video contains all the horrors of this implementation.
When I finish implementing the workshop takeover, I'll upload it to GitHub. I'm interested in the viability of the idea, but so far, the idea is very viable.
I probably misled you a bit with the title about darkness; right now, it's not AI, but a simple script-based clicker. There's no AI here; AI is programmed by a person based on the principle of "see the situation - do the next thing."

This is a bot playing the game; I just launch it. Kappa

potentially, does someone actually need what I'm doing?