Comments on: Reverse updates

By: Wouter Lindenhof

Wouter Lindenhof — Tue, 26 Oct 2010 11:14:35 +0000

That is actually quite a good idea. It does rise another issue, how would you give the plots a mathematical value? The plots are created by the computer (to find a solution for the current state of affairs).
There are actually quite a few solutions for this problem (story templates).

By the way I did find this paper: Story generators: Models and approaches for the generation of literary artefacts which might be of some interest to you.

By: Don Bloomfield

Don Bloomfield — Tue, 26 Oct 2010 01:12:16 +0000

I’d like to see if you could think of it in terms of path-finding, where an action (going forward) causes a series of consequences that are found using a pathfinding algorithm (Say A*, but that may not be best) – the first action causes a “ripple” that starts a series of events searching through relationships / events around the first, looking for a path that causes the best sum of “plot” or “benefit” and keeps causing ripples outward until the initial “importance” or “strength” is spent.

Might enable you to pull this idea off…

By: Wouter Lindenhof

Wouter Lindenhof — Thu, 10 Jun 2010 06:54:22 +0000

real time updates wouldn’t be a problem if all interactions where one on one, properly distributed over time and containing limited information. If on a party (or Marketplace) something big happens then the amount of updates would be enormous, poorly distributed and most likely contain many (Redundant) information. Let me say that if the player is there at the moment in time no matter what kind of update you use, be it forward (real time) or reversed updates, you will have a performance hit.
But what if the player is nowhere near that place (another city)? Real time updates would cause a performance hit that the user will not be able to explain.

The idea behind reversed updates basically say: “We deliver only on demand and if there is no demand we won’t update and when we update we will have relevant information first available and less relevant information later”.

I will try and write out a scenario later which handles all cases (real time forward updates, delayed forward updates, reversed updates) including an estimation how much it would cost in terms of performance.

By: Thijs

Thijs — Thu, 10 Jun 2010 05:56:37 +0000

I thing you are over-complicating things again. In the above example, you state that for each action you would need to update a lot relations for each action taken by any character (be it the hero of an NPC).
However, in a realtime simulation (or even turnbased, based on your timespan), you would only need to update the relations as they take place. If the hero buys the sword, it would take time for the shopkeeper to go buy his food. After that it would take time for the next action to happen, because the NPC’s will actually have to take that action.
If you let them perform these actions, the relations will be updated on a one on one bases each time, as one character interacts with the next. Even if the miller goes out of business each person in the town will only be affected when they go buy bread (or if an other person tells them.)
This would make the system a lot more manageable and far less complex. Even if you have a lot of relations between character, the effect of any action would slowly spread as the character would interact with each other.