Data Official Smogon University Usage Statistics Discussion Thread, mk.3

Status
Not open for further replies.
would sampling stats take less for them to compute? why aren't we doing it right now? i am not good enough at statistics to make the calculations myself but a few posts in the March Stats thread (like http://www.smogon.com/forums/threads/march-stats.3600723/post-7308956
http://www.smogon.com/forums/threads/march-stats.3600723/post-7309192
http://www.smogon.com/forums/threads/march-stats.3600723/post-7320948) suggests that we could have >98% accuracy with just 10% of the total number of games and the error is insignificant if it means stats came out a week early :)
 
I am sincerely sorry to rub salt into the wound but as pokeisfun asked for it, would be pleasant it we get an esteemed date for the release of November Stats. I know that it must be boring to hear the same thing a couple time but be silent as a grave is not the solution in my opinion. If there is some problems to get the usages stats, maybe we should think about a better or easier way to get them (as galbia evoked it). Ask people to help you, we have a big community and there should be people capable of helping if the work is too much for only one person.

Btw thanks for ur work Antar.
 
Last edited:
Anything that drops now will be affected in the next tier shift three months later as their usage for December will be only for a third of the month. They are also more likely to drop in the quick shifts next month. That should probably be addressed.
 
So compute-time isn't the issue right now so much as having to debug the spaghetti-code Python 2 scripts I wrote as a grad student. The new Ultra formes are not something the scripts can currently handle, and on top of all of that, now I'm getting Unicode issues (which is why I never code in Py2 any more).

I really hate to say it, but it's looking like I won't be able to post November stats until close to January (flip side: I'm very hopeful I'll have stats for the first full month of Ultra-SuMo early in January).

I'm sorry--I know this is disappointing to a lot of people, and my communication has definitely not been at what I would consider an acceptable level.

I'll post updates as I have them, but it might be another week and a half before everything is done.
 
All right, so I'm hoping I have everything fixed. Stats are compiling now and should be up for November and December in early January.

My thoughts exactly. I'm sure Zarel could give you an assistant or two
I would love to have someone assist me with a stats-system rewrite. I can totally lay out backlog, do code reviews, etc. I just don't have the bandwidth to do the whole overhaul myself.
 
Update: November might be done enough to get a tier update*, but December needs to be rerun to handle some weird issues in PS logs that cropped up late in the month (specifically, some logs contain chat/bot messages that they shouldn't, and that broke the parser).

*In the process of fixing the aforementioned PS issue, I discovered a bug in my parser that borked the part of the log parser that analyzes matchups (which is used to calculate checks/counter stats). In theory it shouldn't affect the usage stats numbers that are used to generate tier updates (though the "Real" numbers are going to be completely wrong), so I should be able to hold off on rerunning November until after I post the tier update, but I need to double-check.
 

G-Luke

Sugar, Spice and One For All
is a Community Contributoris a CAP Contributoris a Forum Moderator Alumnus
Update: November might be done enough to get a tier update*, but December needs to be rerun to handle some weird issues in PS logs that cropped up late in the month (specifically, some logs contain chat/bot messages that they shouldn't, and that broke the parser).

*In the process of fixing the aforementioned PS issue, I discovered a bug in my parser that borked the part of the log parser that analyzes matchups (which is used to calculate checks/counter stats). In theory it shouldn't affect the usage stats numbers that are used to generate tier updates (though the "Real" numbers are going to be completely wrong), so I should be able to hold off on rerunning November until after I post the tier update, but I need to double-check.
Man I just want to run Mega Pidgeot viably in Tier Shift
 
Antar you are a hard working individual and we all root for you. You are a good person and a lovable individual. Don't give up!! We know you can do it. Sometimes we seem impatient but without you none of this would work so actually we're very proud and grateful to you. Thank you for everything, and a happy new year!
 
I can understand the others... but why Dhelmise? Dhelmise is terrible.
It has been part of low ladder in the beginning of UU. Since it had dropped to RU at the time, it was kinda awkward but viable. However, Devidueye dropped and threw it near the unviable line. Eventually, the Tsareena buffs made Dhelmise completely unviable. So yeah. Unfortunately, like most tiers, it has become a low ladder victim.
 
It has been part of low ladder in the beginning of UU. Since it had dropped to RU at the time, it was kinda awkward but viable. However, Devidueye dropped and threw it near the unviable line. Eventually, the Tsareena buffs made Dhelmise completely unviable. So yeah. Unfortunately, like most tiers, it has become a low ladder victim.
It was never viable.
 

cdbk

Banned deucer.
So I had a decent amount of free time, so I thought:
Hey, what if I made a visual list of my...
Tier Shift predictions for November and December!

These predictions have been made using 2 key factors:
  • Viability in the tier (and consequently metagame shifts)
  • The usage stats from October

OU

Rises:
(♂)


Drops:
(+
please low ladder)


UU

Rises:
(probably drop to RU once the January Stats are revealed)


Drops:


RU

Rises:


Drops aka I beg you, please stop clutching onto these:
+Ampharos


NU

Rises:


Drops:


Now let's just hope that this will be accurate looking at you RU
 
Last edited:
Status
Not open for further replies.

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

Top