Tiering for Generation 8

Status
Not open for further replies.

Hogg

grubbing in the ashes
is a Tournament Director Alumnusis a Site Content Manager Alumnusis a Social Media Contributor Alumnusis a Community Contributor Alumnusis a Top Tiering Contributor Alumnusis a Battle Simulator Staff Alumnusis an Administrator Alumnus
Hello, all!

With our first set of usage stats coming out tomorrow, I wanted to give everyone an idea of how we plan to implement lower tiers this generation.

First, a major change: after much discussion, Smogon is adjusting its usage threshold for Generation 8 to 4.52%, using unweighted months. In addition (and as was previously discussed in Policy Review), we are fully removing quickrises.

A lot of numbers and terms get bandied about when we talk about tiering: 3.41%, 4.52%, quickdrops, quickrises, etc. But ultimately, our usage-based tiering has been based for more than a decade on one simple proposition: a Pokemon is considered Overused if it has at least a 50% chance to appear in 20 randomly selected games.

In practice, this has meant that anything with a usage of 3.41% or above over the course of a tier shift period has been considered OU, while all those below that usage fall to UU. There's more to it than just that, including weighted months, quickdrops, quickrises, etc., but ultimately it has all been based around that core assumption.

When we were looking at refining our tiering thresholds, we decided that we would still work off of that same basic concept: a Pokemon would be considered OU if it had a 50% chance to appear in T games. We looked at different ways we could adjust that T value to better reflect our modern tiering processes. What we settled on was T=15, which results in a 4.52% usage threshold.


Why are we changing this? Well, last generation there were serious concerns with how disruptive tier shifts could be to lower tiers. While drops could always be handled via bans, rises could completely destabilize a tier, causing major impact... impact that then needed to be reset in a few months when the risen element fell again. Even a single popular sample team could end up causing major ripples in the tiers below due to nothing more than a passing trend. This was especially true if it occurred during the months weighted more heavily. While rises are always going to be a part of any usage-based tiering process, we wanted a solution that made them a bit less frequent.

So what does it mean to increase the tiering threshold? Well, the end result will be smaller tiers overall, where rises are less frequent but drops are more common. And ideally by going with an unweighted metric for tier shifts, eliminating quickrises and keeping quickdrops fairly low, we would also favor shifts based on consistent usage over a sustained period of time, rather than spikes in usage during certain periods. Also, a happy side effect of smaller tiers is that it allows lower tiers to be more practical despite the reduced Dex size.

In addition, I'd like to announce our planned schedule for how we will implement lower tiers this generation. Following one month of usage stats from the tier above, a lower tier will enter into its alpha period. Alpha will represent a lower tier's "open" period, with no bans or suspects. It will remain in alpha for one month, after which it will automatically enter its beta period. During beta, tier leadership and councils will work to rapidly get the tier into shape, generally involving council votes for quick bans. Once the tier settles, it will exit beta, and future suspects will generally be by public tests except in cases where rapid action is needed.

This means we will be looking at the following schedule for the formation of lower tiers:
  • December 1, 2019: UU Alpha is released
  • January 1, 2020: UU exits Alpha. RU Alpha is released.
  • February 1, 2020: RU exits Alpha. NU Alpha is released.
  • March 1, 2020: NU exits Alpha. PU Alpha is released.
  • April 1, 2020: PU exits Alpha.
Once PU exits its alpha period, we will return to our standard three-month tiering schedule (incorporating the changes as referenced above).

Thanks for reading. If you have any questions, please feel free to reach out to me or to your tier leadership.
 

Marty

Always more to find
is a Site Content Manageris a Battle Simulator Administratoris a Programmeris a Member of Senior Staffis a Community Contributoris a Top Researcheris a Top Tiering Contributor
Research Leader
Since DLC was announced for Generation 8, we've had to adjust the continuation of the above schedule slightly. Once the first DLC pack comes out we would essentially have to go back to the "turbo shifts" that were the norm from December through April, in order to get all the added Pokemon into their rightful tiers in a reasonable amount of time. Pokémon HOME also came out last month and a few dozen more Pokemon became available, so it doesn't make much sense to start up the standard three-month tiering in May for the month and a bit that would remain to the release of DLC pack 1.

We will continue doing rises and drops based solely on the previous month's usage stats until a narrower release window is announced for the second DLC pack for Sword and Shield, then decide how to proceed from there. Thanks for understanding, and enjoy the ride!
 

Marty

Always more to find
is a Site Content Manageris a Battle Simulator Administratoris a Programmeris a Member of Senior Staffis a Community Contributoris a Top Researcheris a Top Tiering Contributor
Research Leader
DLC pack 1 will be released next week, so we're revising the timetable. There's been discussion on how much time and resources will be wasted across all sections if we continue with monthly shifts through to DLC pack 2, and various proposals have been brought up on how to deal with this effectively. In an effort to come up with a process as fast as realistically possible, I've spent the last week and a half coming up with a usage-based solution that's been unanimously agreed upon by the affected tier leaders.

On July 1st only, usage stats for the purposes of tiering will be based on the final ~13 days of June, after DLC pack 1 is implemented on Pokemon Showdown.

Dropping from OU will be more complex:​
OU to UU -> below the 4.52% cutoff​
OU to RU -> below 1/15th of 4.52%​
OU to NU -> below 1/30th of 4.52%​
OU to PU -> below 1/60th of 4.52%​
Rises and drops from tiers other than OU will behave as usual. Also, NFEs from DLC pack 1 will start untiered.​

August 1st's shifts will be based on July using the 4.52% cutoff for all tiers, in order to get any stragglers from the previous shift moved around. Then, if before September comes around DLC pack 2 gets slated to be released late November or December, we'll talk about doing a normal 3-month shift process for September-October-November.

Here's hoping everything goes smoothly and we can try it all again for DLC pack 2!

Good question! I re-ran the last two weeks of February usage post-HOME release (17th through 29th). From there I used what lower tiers have ended up looking like today and worked backwards to find ideal multi-tier-drop cutoffs. Below is an example of this metric used on post-HOME February usage stats for the Pokemon that originally fell to UU.
Blastoise moved from OU to UU
Celebi moved from OU to UU
Cobalion moved from OU to UU
Decidueye moved from OU to UU
Gyarados moved from OU to UU
Incineroar moved from OU to UU
Ninetales-Alola moved from OU to UU
Primarina moved from OU to UU
Raichu-Alola moved from OU to UU
Rotom-Mow moved from OU to UU
Toxtricity moved from OU to UU
Virizion moved from OU to UU

Linoone moved from OU to RU

Persian-Alola moved from OU to NU
Rapidash moved from OU to NU
Stunfisk moved from OU to NU
Weezing moved from OU to NU

Corsola moved from OU to PU
Dugtrio-Alola moved from OU to PU
Farfetch'd moved from OU to PU
Meltan moved from OU to PU

(Necrozma and Venusaur manage to get used enough in OU during these two weeks to stay OU.)
 

Marty

Always more to find
is a Site Content Manageris a Battle Simulator Administratoris a Programmeris a Member of Senior Staffis a Community Contributoris a Top Researcheris a Top Tiering Contributor
Research Leader
Hi! DLC pack 2 will be released this week and the above plan worked fairly well so it'll happen again, but with some adjustments after. Here's how the rest of this year will look:

November 1st - usage stats for the purposes of tiering will be based on the last ~9 days of October, after DLC pack 2 is implemented on Pokemon Showdown.

Dropping from OU will be more complex:​
OU to UU -> below the 4.52% cutoff​
OU to RU -> below 1/15th of 4.52%​
OU to NU -> below 1/30th of 4.52%​
OU to PU -> below 1/60th of 4.52%​
Rises and drops from tiers other than OU will behave as usual. Also, NFEs from DLC pack 2 will start untiered.​

November 15th - normal tier update treating November 1-14 as a full month.

December 1st - normal tier update treating November 15-30 as a full month.

Going into 2021 we'll see how things are going and actually start up the normal 3-month shift process at some point, assuming there are no more surprise announcements.
 

Marty

Always more to find
is a Site Content Manageris a Battle Simulator Administratoris a Programmeris a Member of Senior Staffis a Community Contributoris a Top Researcheris a Top Tiering Contributor
Research Leader
Hello, let's revise the tiering timeline again!

December 16th - tier update treating December 1-15 as a full month (4.52% rise/drop)

January 1st - tier update treating December 16-31 as a full month (4.52% rise/drop)

February 1st - January usage stats (1st month, < 1.53% to quick drop)
March 1st - January + February usage stats (2nd month, < 2.28% to quick drop)
April 1st - January + February + March usage stats (3rd month, > 4.52% to rise, < 4.52% to drop)
May 1st - April usage stats (1st month, < 1.53% to quick drop)
etc.

We're almost through the DLC rough patch(es), here's to 2021! :blobthumbsup:
 
Status
Not open for further replies.

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

Top