Difference between revisions of "Talk:Wind"
From Worms Knowledge Base
(edited so that "adjacent" leaves out the case of being equal) |
m |
||
| (3 intermediate revisions by the same user not shown) | |||
| Line 13: | Line 13: | ||
The article is so beautifully written I'd feel bad about getting rid of Run's limited but really cool empirical analysis and replacing it with precise data. So for now I'll post my analysis here in the talk page, instead. | The article is so beautifully written I'd feel bad about getting rid of Run's limited but really cool empirical analysis and replacing it with precise data. So for now I'll post my analysis here in the talk page, instead. | ||
| − | The wind is biased against coming up two turns in a row with a wind value that is equal or "adjacent to" (within 1 quantized step of) the previous turn's wind. My preliminary tests (running | + | The wind is biased against coming up two turns in a row with a wind value that is equal or "adjacent to" (within 1 quantized step of) the previous turn's wind. My preliminary tests (running 160 billion trials) show that the chance of getting a wind "adjacent" to the previous wind is 1 in 1922, and the chance of getting the same wind as the previous turn is 1 in 5371. |
| − | The position of the active worm at the start of the turn biases the wind to a maximum of 7 steps in the opposite direction (this happens up to 1/30 of the map width from its left or right edge). Between these two extremes the bias scales linearly relative to position (though it is quantized in steps of 1/15 map width). A bias of +7 means that the wind range will only be in the range -3 to +10, with +10 being close to 7 times more likely than any other value due to range clipping. | + | The position of the active worm at the start of the turn biases the wind to a maximum of 7 steps in the opposite direction (this happens up to 1/30 of the map width from its left or right edge). Between these two extremes the bias scales linearly relative to position (though it is quantized in steps of 1/15 map width). A bias of +7 means that the wind range will only be in the range -3 to +10, with +10 being close to 7 times more likely than any other value due to range clipping. —[[User:Deadcode|Deadcode]] 02:01, 15 July 2009 (UTC) |
| − | —[[User:Deadcode|Deadcode]] 02:01, 15 July 2009 (UTC) | + | |
| + | It is possible to get a bias of 10 by constructing a leaning tower of worms, with the bottom worm on the edge of the map and each additional worm on the edge of the worm below it. The worms near the top of this tower will always get maximum wind. —[[User:Deadcode|Deadcode]] 02:30, 15 July 2009 (UTC) | ||
Latest revision as of 10:14, 15 July 2009
Wind consecutivity
|
W:A replay: Same winds in two consecutive turns |
| Download · Info |
As I was experimenting with the weapons' effect radius (which experiment, by the way, seems to be a failure), I was skipping every one of my turns that wasn't a zero-wind turn. While doing so, I had the same (smallest lefward) wind in two consecutive turns. I don't know if this could happen bacause of me using the latest RubberWorm extension, or not, but it still looks strange, and I think could modify what we thought about the wind up til now. Take a look at the replay's infos, for playback details. Balee 15:47, 8 March 2009 (UTC)
- Oh cool. That'll have to be mentioned in the article. Though, despite that it happens, it's clearly rarer than it should be! Run! 07:29, 10 March 2009 (UTC)
- Well I'm thinking about some way to check lots of turns (I mean, thousands) in the least time possible, and as far as I think now, I might be able to make something useful, like a script, for this. It will definitely take some time, though. Balee 16:34, 11 March 2009 (UTC)
Oookay, I guess Im done with the testing. It took a while, but I checked 1014 turns (come to think of it, why didnt I wait for 1024? eh...) in one single game, and two consecutive turns had the same wind only once. So I think it can be said now that while it is absolutely possible for the game to generate same winds, it has a very low statistical probability. I can upload the replay if you wish :) Balee 19:46, 22 March 2009 (UTC)
- Also as I conducted the simulation offline Im quite sure RubberWorm had nothing to do with it the previous time. Balee One minute later (UTC)
The article is so beautifully written I'd feel bad about getting rid of Run's limited but really cool empirical analysis and replacing it with precise data. So for now I'll post my analysis here in the talk page, instead.
The wind is biased against coming up two turns in a row with a wind value that is equal or "adjacent to" (within 1 quantized step of) the previous turn's wind. My preliminary tests (running 160 billion trials) show that the chance of getting a wind "adjacent" to the previous wind is 1 in 1922, and the chance of getting the same wind as the previous turn is 1 in 5371.
The position of the active worm at the start of the turn biases the wind to a maximum of 7 steps in the opposite direction (this happens up to 1/30 of the map width from its left or right edge). Between these two extremes the bias scales linearly relative to position (though it is quantized in steps of 1/15 map width). A bias of +7 means that the wind range will only be in the range -3 to +10, with +10 being close to 7 times more likely than any other value due to range clipping. —Deadcode 02:01, 15 July 2009 (UTC)
It is possible to get a bias of 10 by constructing a leaning tower of worms, with the bottom worm on the edge of the map and each additional worm on the edge of the worm below it. The worms near the top of this tower will always get maximum wind. —Deadcode 02:30, 15 July 2009 (UTC)
