I just started using the anomaly ratting bot and have a few suggestions that I think would improve it a lot
This is the bot I’m using: FrontendWeb.Main
Check that the always active modules are actually active every X ticks. It has happened on few occasions that the bot missed to activate one of the modules, like a prop mod or armor hardener and it doesn’t seem to re-check or re-enable it after warping in to an anomaly.
When hostiles are detected in local, first thing to be done should be to align to closest station, then recall drones. Recalling drones first and waiting for them to come back before even issuing align or dock is time wasted that puts the ship in danger of getting tackled.
Add configurable random delay for undocking after hostiles detected. It looks very unnatural to undock the instant hostiles leave the system. Also it may be exploited by them, if they immediately burn back to gate and jump back and catch you while bot is already warping back to anomaly.
Add a new configurable item for providing prioritized rats, so that the rats whose names include configured string should be targeted first. This mainly for killing off frigates using the most dangerous modules like EWAR, webs and warp disruptors.
Recall drones if drone shields are gone. Right now the bot seems to ignore drone health and drones keep dying when targeted.
It also happens that the bot tries to leave the anomaly and go to the next one with last rat still alive. Something is wrong there.
Thank you for sharing your experience. I think it makes sense for me to spend some time adapting the bots to your ideas.
That sounds like the appearance of the module button on your setup is too different from the training data set. If the appearances differ too much, this can result in misclassification of the module state.
We can probably fix this issue by adding training data representing how it looks in your game client.
If you link the session recording archive of your session, I can take a look and adapt the bot to your game client.
The module activation may be due to me running with no 3D space background to reduce resource usage? (CTRL+SHIFT+F9)? Technically using this should make it easier for bot to recognize modules, not harder, since the backround is always static black.
I will look into saving up session logs for reference. Will you be able to create a test version with some of those improvements which I can test? I looked into changing the code myself since I know PHP and Python, but this ELM script looks like black magic