Depth in Strategic Games

[droqen]

Regarding multiple authors'*
"Depth in Strategic Games"

* Frank Lantz, Aaron Isaksen, Alexander Jaffe, Andy Nealen, Julian Togelius

[droqen]

I read this paper years ago, before newforum, but my thought pattern has been brought back to it by Splatoon 3 (See: Salmon Run Next Wave).

My remembered image of this paper:
It presents a spectrum of how decisionmaking occurs, either through raw search or heuristics, and 'depth in strategic games' requires a balance of both.

My new Salmon Run-fuelled picture:
Heuristics and raw search can never be applied directly to an even moderately complex game system, but are rather applied to the player's model.

I'm going to root around in the paper to see if this is addressed at all. Is there terminology I can harvest?

[droqen]

Depth is often referred to by game developers and in scholarly research but to our knowledge no attempts have been undertaken to make a thorough and rigorous investigation into the property to which is refers. The purpose of this paper is to lay the groundwork for such an investigation. We are attempting to establish a foundation, clarify the important questions, and suggest directions for further study. We are not at this time proposing final answers to the central question.

Very reasonable. I like this introduction.

[droqen]

We aim to develop a precise definition of d [the proposed formal depth property] that is psychology-independent. It should not make special reference to how humans learn or what humans find interesting or challenging[.]

I'm not sure about this. Can 'heuristics' be separated from 'how humans learn'?

[droqen]

[..] consistent under any conditions and would remain true for any intelligent, problem-solving process, whether human, non-human, or mechanical.

Attempting to discover a property does not adhere to my world view at the moment (more on that sometime, somewhere); I'll likely read this paper from the perspective that depth must be a phenomenal experience despite what is said here.

[droqen]

Searching for a needle in a haystack is a difficult problem (in the sense of being resource intensive) but it's not the kind of problem that requires cleverness and creativity, the kind of problem that rewards life-long learning and can support a large, long-term community of serious, dedicated players. Those are the features we are interested in explaining.

We can frame d as the capacity for a game system to allow for a ranked population of strategies that provide partial/approximate solutions. [..] a game that requires a great deal of computational resources to play perfectly, and also allows for many intermediary strategies along the way.

'Computational resources' in humans involves some interesting brain developments. What are those developments? That's what I'm really interested in, myself!

[droqen]

A person doesn't necessarily gain computational resources. I expect to get into the 'heuristics' part soon enough, the part I remember most clearly, but it's interesting that it's framed this way first.

Rather than gaining computational resources what does a person do? Do they commit more computational resources to the task at hand? (Is committing computational resources a difficult task in and of itself?) Do they become more efficient at allocating their computational resources to the correct places? Is this its own 'deep' task, with its own 'strategy ladder'?

[droqen]

Is the nature of strategy not dependent on the hardware on which it runs? I suppose the existence of NP-HARD problems argues that 'algorithms' are a sort of universal truth. Turing machines and whatnot. Hmm. I want to interrogate that.