A Theory of Depth for Game Design

You videos are so good!

Game Overanalyzer never drops a ball. This might be one of the most relevant videos for game designers from your channel. Thank you, I enjoy listening to your videos during the walk, washing dishes, or when waiting for the bus.

He doesn't miss.

Thank you for this video. I always love how you reference books, articles, and GDC talks that I can look up for more info.

I checked your sources in the description but couldn't find the one for the GDC talk you show at 7:52. Only for the 2 GDC talks that showed after. Could you say what talk its from please?

if you hosted a podcast with game designers and discussed these topics it would be amazing.

Elucidating as usual.

A sidenote of yours really got me to think, and that's how momentum can be used to feed into combat systems.

I used to primarily think about the relation of movement and combat in terms of positioning, so giving players the means
to get a combat or stealth advantage through cover and height which is fine and dandy, but if the player can feed their
velocity into the combat system, suddenly every little piece of the movement system that allows people to get a little
bit faster translates into a combat advantage – at the price of foregoing slow and methodical combat and thus opening
themselves up to attack. And it can of course be modified by other factors like carried weight/loadout.

I've been struggling with how to enable both slow and methodical combat and fast movement and verticality in the same
game space without fast movement only being useful for quick, zero-risk escapes (a pattern I have come to detest) and
I think this has been a key point towards that, so thanks!

To add my own sidenote: I think an area that's especially interesting is feeding small and direct actions into wider world
systems. The example I always give for this as it's how my interest in game systems engineering started is ecosystem
simulation.

Imagine a bog-standard RPG; you're in the starting village and oh boy, there's wolves about in the forest surrounding
it. The townsfolk task you with the eradication of a pack that's been attacking travelers, so you do and collect your
reward. And you go on killing pretty much all the wolves in the surrounding area, because it's a fantasy RPG and
that's what you're supposed to do, right?

But you have removed a keystone predator and now rabbits, rats and other small rodents multiply without being
kept in check and start eating and shitting in the towns granaries, spreading famine and disease. So you stop
killing the wolves who now have an overabundance of prey and start multiplying accordingly. At first, they don't
attack people because there's so much easier prey. But as the prey is diminished by the burgeoning wolf population,
the wolves start to go hungry and now you have an even worse wolf problem and a town population that's sick and
weak. And systems-wise this isn't even that hard to do. Lotka-Volterra is a well-established function to model
predator-prey relationships (tho I would modify it for more control) and infection is a function of population size,
density and current infection ratio.

Hunger itself is simple to implement, but it would imply more logic (plant->harvest->store->process->consume)
that is probably the most involved ones out of these.

The real challenge here is balancing these so things don't just explode everywhere all the time. I think the basic
implementation to enable this wouldn't be all this complicated tho – essentially you just feed weighting curves
into the functions that can be edited graphically and use this to make ranges on some axes chosen by algorithm
implementation more or less stable. For example defining a high weight for intermediate population densities
of both predator and prey species while having 0 weight on both ends of the curve so these locations in state
space either don't happen at all or are quickly (and more randomly than weighted ones) bounced back from.

A normal distribution for these would likely be a good default, but I think this gets really interesting if you have
multiple defined equilibriums on the same axis. That way you could define a whole range of distinct variations
for locations. If we go back to the previous example with the town and the wolves the rather relaxed starting
town with a small wolf problem vs. the depressing place of death and disease it became could be two such
distinct variations that are stabilized like this. But it being systemic, there would be a plethora more emergent
scenarios that are delightfully impossible to foresee as developer. 🙂

Depth is the property of a game to resist minmaxing