Jump to content

  • Log In with Google      Sign In   
  • Create Account


Member Since 25 Nov 2005
Offline Last Active Today, 01:59 AM

Posts I've Made

In Topic: Storing Signed Distance Fields

22 June 2016 - 03:24 AM

  • Do you need to combine elementary distance functions and render their sum, product etc. (for example, "metaballs" composed of several spheres)?
  • Do you really need to keep around SignedDistanceFunction objects, i.e. old brushstrokes? Forgetting them after applying an "edit" once would seem the most efficient course of action.
  • What do you need to store in your voxels, apart from the 1-byte material code?
  • How are you going to render your voxels? Their representation has to be suitable.

In Topic: A game suitable to training a neural network?

20 June 2016 - 02:02 AM

A racing game with simple graphics and rules would be slightly more complex, but much more fun for humans, than the suggested cursor chasing.

Simple 3D or pseudo-3D graphics from the point of view of the vehicle might allow the use of screenshots as input, like in most experiments with Atari 2600 games, but a more abstract input might be more effective for the NN and more convenient for designing graphics.

For example, if the course is represented as a "tube" of passable terrain around a center line, you can adopt a coordinate system consisting of arc length (between the projections on the center line of the current player position and of the point of interest) and distance from the center line; the NN input can then consist, in addition to car positions and velocities, of track shape given as curvature of the center line and width at many predetermined arc length offsets ahead of the player.

In Topic: Using Pathfinding Efficiently

13 June 2016 - 04:03 AM

I strongly recommend using a profiler; in particular, line_profiler, which works with Python 3 and gives line-by-line data. 

In this particular case a profiler would give you not only timing data, but also an immediate indication of abnormal counts of function calls. For example, in a map with enough impassable walls A* should expand only the nodes on the optimal path and the adjacent walls; if it expands more you have a bug

In Topic: RPG combat maths

13 June 2016 - 02:34 AM

As you describe it, "luck" consists of doing more damage with any attack: it is very incoherent with having separate skills for different attack traits.


In some games, particularly tabletop roleplaying games relying on other types of character building and advancement, there is no place for skills or other stats attached to specific attack "traits"; what attacks are possible and their damage could be collapsed to two trait-oblivious skills or stats (e.g. "magical power" and "magical lore") or one (e.g. "magical aptitude"), or even to no skill at all (all characters are equally good at spellcasting).

"Luck" at dealing damage in general, if present as a stat or skill, would be opposed to something completely different, e.g. "luck" at obtaining good answers from divination magic or contact with the gods or "luck" in interactions to lead and persuade other people.


On the other hand, in a complex, combat-oriented CRPG most players would expect the harmonization to go the opposite way: skill at dealing damage should be specific to some subsets of attacks, in the same way as skill at pulling them off.

For example, "fire attack damage" paired with "maximum fire spell level" and/or "fire spell failure probability by spell level" (or individual spells bought as separate skills, if appropriate).

With the choice between improving "luck" and improving spell sophistication there would be a natural way, on top of the specialization towards different magical traits, to distinguish between a character build that can do something nasty but only at low power and a character build that can hit hard but only in simple ways, with the obvious convergence path of doing a lot of damage with complex attacks.

Bonuses to all damage should be reserved for buffing effects, including magical items.

In Topic: Good tools for indoor-map editing / architecture

10 June 2016 - 03:51 AM

I never tried Max, Maya or other "big guys" except for Blender, which -I've been told- has a map-editor or something. But the few times I tried it, I ran quickly out of patience as I couldn't even figure out how to move the camera hehe. Well, anyhow, advise???



Blender isn't meant to be learned by impatiently poking around and seeing what happens; so many things are "hidden" in sensible but non-obvious keyboard shortcuts, menus, and windows.that the beginner's ignorance becomes a roadblock

Invest time in a course, as suggested, or in following along books and tutorials. I've rarely found the official Blender documentation satisfactory, as it is usually badly edited, incomplete, and/or obsolete.