-
Welcome to Smogon! Take a moment to read the Introduction to Smogon for a run-down on everything Smogon, and make sure you take some time to read the global rules.
-
Congrats to the winners of the 2024 Smog Awards!
Data Official Smogon University Usage Statistics Discussion Thread, mk.3
- Thread starter Antar
- Start date
- Status
How are the metagame statistics calculated? For example, how is it decided that a team is balance instead of semistall? Would it count a balance team that has tornadus-t and rotom-w as part of the volt turn category instead?
Other requests:
- win-rates and combos
- most common megas
- most common choice item and Z-users (scarf would be a huge one)
- lists of all of the above in correlation to their playstyle archetype
- removal of weird metagame categories, looks weird reading the gen7ou mg stats and seeing mono-fairy, for example.
- It would be nice if OLT's could have usage stats for the ladder cycle phases - any chance of filtering out stats out of username tags to achieve this? Would need to be taken out of the games in a week instead over an entire month.
Other requests:
- win-rates and combos
- most common megas
- most common choice item and Z-users (scarf would be a huge one)
- lists of all of the above in correlation to their playstyle archetype
- removal of weird metagame categories, looks weird reading the gen7ou mg stats and seeing mono-fairy, for example.
- It would be nice if OLT's could have usage stats for the ladder cycle phases - any chance of filtering out stats out of username tags to achieve this? Would need to be taken out of the games in a week instead over an entire month.
Last edited:
You should check out Porydex from Drk Pwnr which shows a few of the stats you request (like most common scarf in the link).
Regarding metagame statistics, I know there is an old thread from Antar explaining them all but I couln't find it (will edit the post if I do).
Finally, about adding more stats you should see directly with Marty who is in charge of it now. Although win/rate is considered a bad metric out of ladder stats according to Antar (see point 12 of the op) I think computing it from tournament games would be useful. If you are able to compile enough replays together I'll be glad to run my own scripts that compute win rate and pokemons combos and publish the results
The stalliness metric is described here, with a few other posts on the blog talking about improvements that were made.
I'm not really sure what's on the porydex, the site seems pretty hard to navigate to me. But the most common items, and most common megas and Z-users (or at least, Z-crystal / megastone holders) should be possible to work out from the chaos stats.
I haven't thought in great depth about stalliness metrics, but I can see a few potential issues with just using faints per turn. For one thing, if the whole ladder were very stally, then every team would look as though it were more stally, because offensive teams have their pokemon faint less quickly when playing against stall than against offense. It's also possible the metric would be useful for seeing how often different playstyles get KO'd, in which case that metric would be useless. And also, it would be far more luck-based with regards to which teams you played on the ladder (two extremely similar teams with just different choice of one coverage move might see very different stalliness ratings due to which teams each happened to face on the ladder, for example). And also, that would probably be harder to compute.
I don't think there's a way to see the number of pokemon that faint per turn, without access to the data (or scripts we use on the raw data), though it's possible I'm wrong about that.
Sorry, forgot I hadn't answered this! The short answer is that 100% of the battles were counted, but some are effectively ignored while many are just counted much less than a game of the level specified (e.g. between two 1500-weighted players). More details can be found here: https://www.smogon.com/forums/threads/weighted-stats-faq.3478570/ (sorry, struggling to embed links on my phone).
As for how many games were actually played at each part of the ladder, as far as I can tell we don't have a way of working that out without access to the raw data (at least, not at all easily). If you're interested in the glicko ratings then I'd expect them to basically follow a bell curve, but I honestly have no idea how other ratings would be distributed.
Is there any raw data available with full team builds? I'd like to build a teambuilder that makes good suggestions on team construction, but the current data isn't quite enough to get there as there's no relationship between the different stats.
There isn't, which annoyed me for similar reasons (though not exactly the same). There's no official plans in place to change this either.
- Status