terrain graphics probabilities
Moderator: Forum Moderators
terrain graphics probabilities
The current terrain-graphics.cfg controls the probabilities for random terrain tile images. Variations, details, etc. Looking at the contents of the file, it seems that some of the WML authors and/or graphics people did not understand the workings of multiple rules with probabilities. In some places I see probabilities like {20% 20% 20% 20% 100%}, which tells me that the artist probably intended for the 5 variants to occur equally often. The problem is that due to compounding probabilities, the actual random distribution is more like {20% 17% 13% 10% 40%} (17% being 20% of the 80% remaining after the first rule catches 20%, etc). To accomplish the even 1/5 split that I think the artist wanted, the probabilities should be set to 20% 25% 33% 50% 100%.
Some of the rules seem to have been tweaked to achieve the look that the artist wanted. Probabilities like {26% 18% 100%} are unlikely to have been set arbitrarily or with fore-planning. These should be left alone.
And then there is the unfortunate grey area. {33% 25% 10% 100%} might have been set naively, or possibly is just the result of some left-brained tweaking. For the purpose of my correction attempt, I assume that this case was set naively.
I have posted a patch on Gna. I think the obvious naive conditions should almost certainly be fixed. I may have stepped a little too far into the grey area, but that is up to the powers that be.
https://gna.org/patch/?617
PS: I know that probabilities on rules with rotations are also broken, thats a matter for a different thread and I believe an already existing test patch. None of the rules in question here use rotations, I think.
Some of the rules seem to have been tweaked to achieve the look that the artist wanted. Probabilities like {26% 18% 100%} are unlikely to have been set arbitrarily or with fore-planning. These should be left alone.
And then there is the unfortunate grey area. {33% 25% 10% 100%} might have been set naively, or possibly is just the result of some left-brained tweaking. For the purpose of my correction attempt, I assume that this case was set naively.
I have posted a patch on Gna. I think the obvious naive conditions should almost certainly be fixed. I may have stepped a little too far into the grey area, but that is up to the powers that be.
https://gna.org/patch/?617
PS: I know that probabilities on rules with rotations are also broken, thats a matter for a different thread and I believe an already existing test patch. None of the rules in question here use rotations, I think.
Though I imagine you've already read it, confer the following discussion in the thread about new swamps:
http://www.wesnoth.org/forum/viewtopic. ... c&start=15
http://www.wesnoth.org/forum/viewtopic. ... c&start=15