Data Official Smogon University Simulator Statistics — December 2013

Status
Not open for further replies.

darkie

just remember no caps when you spell the mans name
is a Site Staff Alumnusis a Smogon Social Media Contributor Alumnusis a Super Moderator Alumnusis a Live Chat Contributor Alumnusis a CAP Contributor Alumnusis a Contributor Alumnusis a Smogon Media Contributor Alumnusis a Battle Server Moderator Alumnus
#52
So, apparently the best counter to Heatran according to the stats is Kingler
If I understand the stats correctly, it's not that the best counter to Heatran is Kingler, but more that of all teams with Kingler, they use their Kingler as a Heatran counter more than any other Pokemon.
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
#53
HikozaruYes, based on the month's data, that's what the stats show.

It's not quite as darkie put it, since I can *try* to use Exeggutor as a Heatran counter as much as I want, and it still won't work, but what it says is that, of all the Pokemon that have ever faced off against Heatran, the one that has statistically done the best is Kingler.

There's an element of garbage in -> garbage out going on here, but it's really more of a case of small number statistics more than anything (I do try to filter out low occurrences, but I have to put the cutoff somewhere).
 
#57
Not sure if it's been asked already, but is it possible that we could get 1850 moveset statistics? It's quite disconcerting to see that the most common Rotom-W is specially defensive in this meta, and I have reason to believe that it's at least in part due to Showdown's suggested spread not changing since last gen.
 
#58
The "changes vs last month" stats for OU compare this month's OU stats to (presumably) last month's Gen V OU stats. While this is certainly interesting, does the data that compares last month's Pokebank OU stats to this month's OU stats exist in an easily obtainable format? It would be interesting to see, especially for Pokemon such as Mandibuzz and Bisharp that had significantly higher usage in December than in November. I could always just manually subtract the usage for each Pokemon, but that would be tedious and very time consuming.

Also, I don't know if this has been pointed out yet, but "Nintendo Gen V Metagames" is misspelled in the stats posts.
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
#59
Not sure if it's been asked already, but is it possible that we could get 1850 moveset statistics? It's quite disconcerting to see that the most common Rotom-W is specially defensive in this meta, and I have reason to believe that it's at least in part due to Showdown's suggested spread not changing since last gen.
1850 moveset stats tend not to look significantly different, but I'll check this one for you (VM/PM me if I don't post it by the end of the day today.

The "changes vs last month" stats for OU compare this month's OU stats to (presumably) last month's Gen V OU stats. While this is certainly interesting, does the data that compares last month's Pokebank OU stats to this month's OU stats exist in an easily obtainable format? It would be interesting to see, especially for Pokemon such as Mandibuzz and Bisharp that had significantly higher usage in December than in November. I could always just manually subtract the usage for each Pokemon, but that would be tedious and very time consuming.
It *should* be comparing last month's Pokebank OU to this month's Pokebank + post-merge OU... Clearly something went wrong. Will investigate.

Edit: It was comparing to October's stats, in which "ou" referred to Gen V OU. Should be fixed by the time you read this.

Also, I don't know if this has been pointed out yet, but "Nintendo Gen V Metagames" is misspelled in the stats posts.
::headdesk::

Unfortunately, I can't change it without reposting the ENTIRE thread (due to XenForo not playing nice with spacing). It'll be correct next month.
 
Last edited:
#60
I wrote a program that takes the list of pokemon on your team and outputs team recommendations based on the checks and counters listed in Antar's moveset data.

It's very rough right now, but the suggestions it's outputting are surprisingly good. I was expecting to need to ask for a more complete version of the checks & counters data (which would still make it more accurate) but so far it's working.

Example 1:



Example 2:



Example 3:



Algorithm:

Threats scores are found by using counter rates of all of your pokemon, summed, then weighted by how often that counter appears. So if Talonflame is countered by Tyranitar 83% of the time, Espeon is countered by Tyranitar 84% of the time, and Tyranitar has a useage weight of 0.58, then T-tar counters a team with those two by (83+84) * 0.58 = ~97.

Threat scores are then reduced by the counter rates of any of your team members. If the above team has Scrafty, which counters Tyranitar 85% of the time, then Tyranitar's score becomes ~12. There is no useage weighting on this step - we know these pokemon are relevant because they're on your team.

Suggestions are used by going through all threats, finding the counter rates of their counters, multiplying them by that threat's threat score, and adding that to the suggestion's suggestion score.

Code:

This isn't wrapped up to be easy to use. You'll need to understand Java. You'll also need a copy of the moveset data report that Antar provided. http://pastebin.com/gpnKx64w

I can do a couple of runs for people who can't run the code themselves, but if it gets to be a hassle or blows up the thread, I'll stop.

I have some ideas for how to make this more powerful and more accessible, but I want to stop and see if anyone else is interested in this first.
Dear squashed im pretty sure i speak for nearly everyone when i say that we would love this made more publicly and accesible as it would help a lot of people out with there team building (or atleast the start of the team building) and would generally make this website better in what its supposed to do.
 
#61
Yikarur Photofluid

It's a moot point. Smogon has decided to discontinue the "Pre-bank" tiers so as to focus on the "Pokebank" metagames. Since they won't appear in stats ever again, it really doesn't matter what you want to call them.
oh why that? Is there a reason?
"Pre-Bank" is what the cartrige people played for months and it's really fun and different and I don't understand why such a good metagame is abandoned like that :(
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
#62
oh why that? Is there a reason?
"Pre-Bank" is what the cartrige people played for months and it's really fun and different and I don't understand why such a good metagame is abandoned like that :(
Because we didn't want to divide the playerbase, and if we could only have one metagame, it would be the one that didn't exclude previous-generation Pokemon. If you'd like, you can propose that "Pre-Bank" be an "Other Metagame," but it will no longer be supported as an official one. This is not up for discussion, especially not in this thread.
 
#63
For anyone following the team suggestions from my post earlier in this thread: https://github.com/tolmar/Pokemon-Showdown-Client

If you download this, open testclient.html, and set your team in the teambuilder to OU, you'll find that pokemon after your first are sorted by suggestion order. If anyone wants to host it properly, talk to me. Otherwise, I sent the patch to Zarel and it may or may not be integrated into PS at his discretion.
 
#64
Apologies if it has already been pointed out but Chesnaught has Illuminate instead of Bulletproof in the data (at least for OU and Ubers, didn't check the others).
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
#65
Apologies if it has already been pointed out but Chesnaught has Illuminate instead of Bulletproof in the data (at least for OU and Ubers, didn't check the others).
Illuminate = no ability. Not sure how that happens, but that's what it means.
 
Status
Not open for further replies.

Users Who Are Viewing This Thread (Users: 1, Guests: 0)