Resource Simple Questions/Simple Answers: Random Battles Edition

Ester98 said:
Hola, wanted to ask because I struggle finding the right thread for it: where would someone go to suggest a new randomized format? My case is also very specific as it would be VGC random battles, so it's a bit of a crossover
Click to expand...
VGC rands is a concept that's been discussed internally before. The closest we've come to an implementation was spotlight cycle 18, which had, as far as I can think, basically all the trappings of VGC besides a lower level average and specifically designed sets, the latter of which would be the biggest bottleneck towards it. If you have any suggestions beyond that, you can lay it out here and we'll keep it in mind, but I don't see it happening soon, especially given potential-dev-timesink Champions is just around the corner.
 
Sauzels said:
VGC rands is a concept that's been discussed internally before. The closest we've come to an implementation was spotlight cycle 18, which had, as far as I can think, basically all the trappings of VGC besides a lower level average and specifically designed sets, the latter of which would be the biggest bottleneck towards it. If you have any suggestions beyond that, you can lay it out here and we'll keep it in mind, but I don't see it happening soon, especially given potential-dev-timesink Champions is just around the corner.
Click to expand...
I actually coded a first version of it in my personal server.

Instead of having it as a classic randbats, I took Limitless results data and used usage statistics to extrapolate whole teams.

Unsure if it counts as spam, if not I could send my private server where I have it implemented and my twitter post; I got a lot of positive feedback from the community, although I'm aware of some limitations
 
Sami said:
when using the /rwr command, what is the Z-score supposed to mean?
Click to expand...
In statistics, the Z-score (also "Standard score" if someone wants to find more about it) is the number of standard deviations a sample is away from the mean.
Here it's used as one would in a statistical test: with the null hypothesis that the "real, underlying" winrate is 50% for a given mon, we can calculate how deviant, and therefore how unlikely, the actual sample is. It is thus a measure of evidence that a pokemon's winrate is not 50%. However, importantly, it is not a measure of how far from 50% the real winrate is.

The Z-score, and thus the evidence that something is not 50%, is larger when the deviation is further from 50%, and when more games have been played with a mon. Specifically it's calculated as the winrate's deviation from 50% multiplied by twice the square root of the number of games.

For example, Gen9 Mamoswine currently has a 50.95% winrate with Z-score 3.07 which indicates that its real winrate is likely above 50% (p~=0.001), but it's not very far from 50%. Gen2 Yanma, on the other hand, has a 58.7% winrate with Z-score 1.18. It thus might very well have a very high winrate, but the evidence is quite weak (p~=0.12). Indeed Yanma might in reality have a winrate lower than 50%. The difference between them is in the sample size. We have 26322 registered games of Gen9 Mamoswine, but only 46 for Gen2 Yanma this month.
 
Why doesn't the randbats command say the most likely items a mon has, especially for cases like Weakness Policy. Asking players to remember item generation is fine but asking them to remember game-losing random items is plain awful as a user experience. Why is there little clarity around this in the UI?
 
bwarang said:
Why doesn't the randbats command say the most likely items a mon has, especially for cases like Weakness Policy. Asking players to remember item generation is fine but asking them to remember game-losing random items is plain awful as a user experience. Why is there little clarity around this in the UI?
Click to expand...
It's because items aren't explicitly coded to be linked to each set, but decided as the result of the role and moves generated. Putting that in the command would require the server to generate item probabilities based on the set generation code every time, which would be a lot of strain. Storing the probabilities would make the command itself more complex as well.
There is a browser extension that can do all that work and add it to the in-battle hovertext, and you can also read this thread for more info about how roles affect items, although yes it doesn't cover certain niche cases.
 
You must log in or register to reply here.
Back
Top