BorgBang

[iggy_hunter]

Anyone have any interest in designing or integrating a borgbang feature. Even something as minimal as a flow chart of logic would help.

See https://nickmcconnell.github.io/Angb.../borgband.html for more info.

[fph]

I have visited that website, but it gives me no information on what "borgband" is.

[Nick]

I think it's Angband 2.9.2 with the borg, or something like that.

[iggy_hunter]

So ... it is ai for a computer to play angband ... simulating keystrokes like a player. It is a borg player. For amgbandos ... I'll provide an interface for a subroutine to be a player ... that means someone need to have a flow chart of logic on how a player plays angband. It will be awesome because angbandos has spectating mode that we can watch.

Next level would be competition for who makes the best borg.

[iggy_hunter]

From Tangeria.com

The Borg is an automatic Angband player. It plays Angband by “virtually” looking the display and pressing keys just as if it were a real person sitting in front of the computer. It can analyze the current situation of the game and tries to make intelligent decisions. There are various implementations of the Borg for different Angband versions and with slightly different capabilities.

[iggy_hunter]

If someone wants to entertain this ... provide me a pseudo logic ... I'll provide functions that provide the details easily ... like

Int currentlevel()
Int mana()
Bool isacidresistant()
Bool isnearmonster()
Int currenthealth()

I believe you can also pull Nick's borgband repo for the logic he used

[Pete Mack]

The borg was a terrible angband player--it spent nearly allvits time scumming level 1-3 for speed potions and thevlike.

[iggy_hunter]

I'll provide the resources for u to make a better one? Deal?

[Julian]

These days, I'm curious how a neural net would take to Angband. (Though really one ought to start with Rogue, as it's a simpler problem, and there's a well-known bot that was very good at it.)

[iggy_hunter]

Yeah ... I'm not up to speed with neural nets ... would be an awesome idea/learning experience

[eschiss1]

The borg was a terrible angband player--it spent nearly allvits time scumming level 1-3 for speed potions and thevlike.

Still won more often than I did.
That said, the APWBorg compiled on Angband 3.4.1 for the Mac keeps casting nonexistent spells (it thinks resist heat and cold is no.7 in the spellbook rather than 6, e.g.) and other issues and since no more recent APWBorg has ever been compiled for any Mac OS and more recent Angbands (certainly not for the very different Mac OS 13- though Angband 3.4.1 does work with Mac OS 13...) I anyway am sort of stuck with this problem, I guess. (I am _guessing_ that there's no quick borg.txt fix for "what keystrokes correspond to which spells"...)

[Estie]

These days, I'm curious how a neural net would take to Angband. (Though really one ought to start with Rogue, as it's a simpler problem, and there's a well-known bot that was very good at it.)

This. I so would love to watch it learn (that is, watching some of its games after 1k, 10k etc attempts).

I guess it wouldnt take too much hardware for it to become better than any human, too. What would it do different ?

[iggy_hunter]

I'm willing to entertain the idea of adding a neural net module for a Borg ... but someone else would have to lead this initiative.

[Pete Mack]

It's worth noting that the original borg was a very competent Rogue player. Expert system yes. Look ahead, yes. AI/Neural net? No.

[iggy_hunter]

I was just looking at this one.

https://dotnet.microsoft.com/en-us/apps/machinelearning-ai/ml-dotnet