Lower Tier Shifts

One thing I've always disliked about low tiers is how susceptible they are to the whims and the "flavor of the month" trends of tiers above them. The resulting erratic tier shifts cause a lot of problems for low tiers, where a potentially defining Pokemon rises to a higher tier due to a reason like OLT, or someone popular influencing ladderers, or simply because something is temporarily trendy. In particular, a big problem is that certain Pokemon tend to hover right at the border of the usage cutoff, sometimes dropping to and rising from a tier multiple times within one generation.

I'll give some recent examples of this problem that I'm familiar with:
  • Rotom-W dropped to UU at one point. Being the great Pokemon it is, it had a massive effect on UU, which is fine. Only three months later, however, it rose again to OU, having an equally massive effect on UU, all within an incredibly short timespan. During that three months, a lot of tiering decisions had to happen because of that first tier shift, which were then made pointless by the second tier shift.
  • Seismitoad rose from NU to UU literally only because of Rotom-W. However, that same tier shift, Rotom-W rose to OU. So NU was robbed of a great Pokemon for no reason whatsoever. Seismitoad is probably going to fall to NU this month (3 months after it rose). It's just pretty senseless that these tiers experience this level of volatility.
  • Amoonguss rose from UU to OU. The impact of this rise arguably straight up caused a Pokemon to be banned from UU (Breloom), and likely had a very large role in the ban of another (Azumarill). Amoonguss seems very likely to fall to UU again this month, from what I've been told.
  • Mamoswine rose from UU to OU because a youtuber gathered enough people to spam it on the ladder; now, three months later, it's very likely to fall again.
  • In ORAS, Azelf was UU for the whole generation. As the generation neared its end, OLT happened, and people spammed that bird team non-stop on the OU ladder for 8 weeks. This alone was enough to push Azelf to OU. It would have fallen again but the generation ended right then, making Azelf permanently OU in gen 6 because of a fluky 2 month period that didn't represent its true "value" in gen 6. Obviously, past gen lower tiers aren't hugely important, but the main point is that the repeated use of ONE team over a short amount of time was enough to have a major impact on the tier below it. OLT in particular is a very large part of the problem I'm addressing.
Obviously, with how tiering works, lower tiers are always going to be subject to this kind of fluctuation, to a certain extent. Drops happen, which is great; they're a big reason that people like lower tiers, and if something drops and is broken, then it gets banned. But if a key Pokemon rises, now you might have to re-evaluate the entire metagame, which is a lot more complicated and might involve multiple tiering decisions. And all that might be for nothing if the Pokemon just drops 3 months later.

-----

My proposal is to eliminate "rises". Get rid of Pokemon moving back up to higher tiers entirely. Once a Pokemon's OU usage moves it to UU, it stays in UU for the rest of the generation, regardless of whether its OU usage goes back up to the cutoff.

First of all, this would have no effect on the higher tiers that would otherwise be "taking" Pokemon from a lower tier. OU would not be affected by, say, Amoonguss being UU but having OU-level usage. It would obviously always be available to use, and if it is good enough to be used, then people will use it whether it's classified as OU or UU. As an OU player, this would have absolutely no effect on you.

Most importantly, this would provide some much needed stability to lower tiers. This would solve all of the problems I talked about before, and allow UU, RU, NU, and PU to make better, longer term decisions - decisions that won't be made useless because of a later tier shift or effectively reversed by another tier shift. Instead of UU making decisions based on a metagame with Amoonguss that would eventually be made pointless by Amoonguss leaving, UU could spend its time making decisions that permanently improve the tier.

This added stability would especially be appreciated with the official tournament circuit running them pretty much year round now (SPL, Snake Draft, Grand Slam - which has seen a ridiculous growth in size). The shifts I talked about above can all feasibly happen within the timespan of any of these official tours (12 weeks). Anyone who played SPL probably saw their UU player during week 3 say "Fuck, Rotom-W just rose...all my teams are useless." A more stable tiering system would be appreciated not only by lower tier communities, but also by anyone involved in these tournaments.

One thing that I should mention is that this system shouldn't take place immediately when a tier begins in a generation. There's a lot of fluctuation that happens in the first 12 months or so of of a new gen lower tier, because the tier above it isn't stable yet. For example, gen 7 UU had Clefable, Keldeo, Hawlucha, Mew, and more when it first started. These are now safely considered OU staples and will probably (as far as I know) have OU-level usage for the rest of the gen. That's just something I wanted to point out - that this system would ideally only be enacted 12+ months into a tier's existence.

-----

I understand that this is kind of a huge change to make - after all, Smogon has been all about "usage based tiering" since day 1, and with this proposal, some Pokemon would fall into a tier below the one that their usage dictates. However, I really think this would be to the benefit of everyone involved. Lower tiers don't have to deal with Pokemon dropping and rising and dropping and rising, and thus can make more lasting decisions and progress their tier far more than they can do now. Higher tiers are still not affected by something dropping out of the usage cutoff for their tier. I would just like to see what people's thoughts on this are and gather a general public opinion.
The current tiering system causes too much instability in lower tiers. Pokemon drop and rise based on temporary trends and that really fucks with UU through PU (see my examples in bullet points). If we got rid of Pokemon rising back up to higher tiers, we could eliminate this and solve a lot of consistency issues in lower tiers without affecting OU or really hurting anything.
 

power

uh-oh, the game in trouble
is a Pre-Contributor
One thing that I should mention is that this system shouldn't take place immediately when a tier begins in a generation. There's a lot of fluctuation that happens in the first 12 months or so of of a new gen lower tier, because the tier above it isn't stable yet.
This is the trickiest point in a proposal; how do we decide when the fluctuation is over and the tier is now stable? For example, in BW2 OU, Mew is good enough to warrant 3.41% usage or more in pretty much every tournament, but BW2 OU didnt stabilize on this meta facet until after the tier ended.

There are bound to be Pokemon that will rise to OU after the 12-month mark that will stay there for the rest of the generation; look at Zapdos in ORAS, for example.

I think that instead, we should try to verify whether a Pokemon has the continued usage to remain in a tier before rising. What I'm proposing is that after a generation is one year old, we should require that a Pokemon meets the 3.41% criteria for two consecutive 3-month usage cycles rather than one. I think this deals with your issue without causing misrepresentation of a Pokemon's power level when there is indeed a permanent metagame development in OU.
 

g

blood tastes like iron
is a Tutor Alumnusis a Site Staff Alumnusis a Team Rater Alumnusis a Super Moderator Alumnusis a Community Contributor Alumnusis a Tiering Contributor Alumnusis a Top Contributor Alumnusis a Smogon Media Contributor Alumnusis a Battle Simulator Moderator Alumnus
This is something i really thought about a lot when i was still involved (draft of a proposal really similar to this one i made a few months ago). I don't know if what you proposed is the best approach to solve the situation (even tho i can't think of real practical problems considering a tiering council exists) and my idea was that tiering updates should happen way less frequently, for example after the biggest tournaments or like 2 times a year, with a higher cutoff for stuff to actually move.
From what i gathered a lot of people think that the volatility of these changes is actually part of the fun and for smaller communities with not that many players that are trying to experiment all the time things can actually get quite stale. I will be interest to read what you think
 

Bughouse

Like ships in the night, you're passing me by
is a member of the Site Staffis a Forum Moderator Alumnusis a CAP Contributor Alumnusis a Tiering Contributor Alumnusis a Contributor Alumnus
I think this is a bad proposal for a few main reasons:

1) smogon tiers are meant to more or less produce a unique list of viable Pokémon in each tier. Watching an UU game, you will see entirely different Pokémon from an OU game. Sure some UU Pokémon might get used in OU but it’s at a low frequency. This policy breaks that. If a Pokémon drops to UU even once, let’s say... mega heracross and mega medicham being UU due to aegislash. Then Aegislash gets banned and both would rise except now they don’t. They’re still UU. But they’re also very good in OU. And so now UU games and OU games look similar.
2) suppose tiering councils step in to prevent this first problem? Well, now we have a smogon with larger banlists. People don’t think of bans by usage as bans really, so while it’s the same outcome as what we currently have it looks worse.
3) it’s not an easy bright line rule to explain to the public.
 
I think that instead, we should try to verify whether a Pokemon has the continued usage to remain in a tier before rising. What I'm proposing is that after a generation is one year old, we should require that a Pokemon meets the 3.41% criteria for two consecutive 3-month usage cycles rather than one. I think this deals with your issue without causing misrepresentation of a Pokemon's power level when there is indeed a permanent metagame development in OU.
I think this is a great idea and appreciate the input. Originally when I was writing this, I was thinking about the idea of raising the "3 month, 3.41%" criteria to make it harder for Pokemon to rise and fall. While raising the percentage may solve the problem to an extent (although I'd have to dig deeper into the numbers), it would also make it much harder for Pokemon to fall, which is something no one wants. However, extending the 3 month criteria to 6 months ONLY for rises (but still letting Pokemon drop after only 3 months of weighted 3.41% usage, as they do currently) sounds like a really great solution. It's a much less drastic change than what I proposed, and keeps the current philosophy of strictly usage based tiering. At worst, a 6 month turnover rate for Pokemon like Rotom-W and Amoonguss that hover around the border of 3.41% is much more manageable than a 3 month one.

If people do have a problem with the huge change in philosophy that my proposal would require, then this seems like an excellent compromise.
 

lydian

Formerly lyd
is a Top Social Media Contributoris a Top Artistis a Forum Moderatoris a Community Contributoris a Top Contributoris a Top Smogon Media Contributor
Moderator
I’d just like to leave my input on the topic. While I’m by no means experienced in lower tiers, the exact reason I don’t play them as often is that rises and drops do constantly shift the meta in a unnecessary way. Of course, unhealthy drops can be accessed by bans, but losing a healthy mon via a rise that was key to balance the metagame can be very frustrating, at least from my point of view. I wouldn’t say that the lack of rises would bring staleness because, well, look at OverUsed! It’s a constantly shifting metagame even though the last change on its banlist came with a quickban to Naganadel. The OU metagame of right now is vastly different from that of, say, two months ago.

I’d also like to address Bughouse’s points in this post.
  1. I don’t think this is by any means bad, I’d say not all that many Pokémon would overlap and even if they do, they’d likely play different roles because of the different meta games, and that’s just something interesting from a viewer standpoint as well. For instance, Gastrodon runs a Specially Defensive set in OU and a Physically Defensive one in PU. Moltres prioritizes Defensive sets in UU as opposed to more Offensive ones in RU. Ferroseed runs a more Physically Defensive set in NU as opposed to a more mixed spread in PU.
  2. Didn’t you say in your example that Mega Heracross and Mega Medicham would have enough usage for both OU and UU? Well a UU ban would make them OU again, not UUBL. I don’t get how this would be any different from they simply rising. You could say it gives the council more work to do, but sometimes a drop could be worth keeping.
  3. I completely agree with you on this point and this is exactly where the issue lies in the no rise proposal. I was willing to post about this topic before seeing your post, but now I’ll try to develop a bit on this argument. So basically, from a technical standpoint removing rises is a great idea, or at least that’s my point of view. But it wouldn’t be all that for whoever is getting onto Smogon, let’s use another example. Suppose this policy of no rises was already implemented when Assault Vest Amoongus popped up and gathered enough usage for OU. Well, of course it wouldn’t rise because rises don’t exis anymore, so it would remain in the UnderUsed tier but, usage wise, it would be over used. I hope you get the point, but basically this format would create a discrepancy between usage and tiering. Which would make calling OU over used controversial, because Amoongus is over used, but it’s not OU. I’m bad at explaining things so sorry if this wasn’t clear. But basically what I’m saying is that cancelling rises makes the whole usage tiering thing hypocritical to some extent. So if we want to completely cross out rises, we would have to at least rename the tiers not to contradict themselves.
With that being said, I think Sedertz’s solution is the best one, considering the fact that eliminating rises as a whole can and will cause a lot of controversy. But even then it could be unhealthy for a couple tiers, for instance, I’m not sure how useful Gliscor is to balance the UU metagame, but it is likely rising to OU and it should, at the very least, stay OU for a good while, making the more months solution not all that great if Gliscor is keeping something in check in UU.

All in all, both dodmen’s original idea and Sedertz’s solutions are excellent alternatives, but like almost everything in life, they come with a downside.
 

Lord Death Man

i cant read
is a Community Contributoris a Tiering Contributor
  1. I don’t think this is by any means bad, I’d say not all that many Pokémon would overlap and even if they do, they’d likely play different roles because of the different meta games, and that’s just something interesting from a viewer standpoint as well. For instance, Gastrodon runs a Specially Defensive set in OU and a Physically Defensive one in PU. Moltres prioritizes Defensive sets in UU as opposed to more Offensive ones in RU. Ferroseed runs a more Physically Defensive set in NU as opposed to a more mixed spread in PU..
Pokemon run different sets in different tiers because different tiers have threats, and a proposal like this could very easily blur the lines between different threats being in different tiers. Is phys def gastrodon really that different from spdef gastrodon? They run the exact same moves, and they check entirely different threats because the metagame around them is different, not just because of their EV spreads. And, despite your claim that Moltres runs different sets in UU and RU, Flynium is the most used item for both up until 1760 stats - that's a massive portion of the playerbase using Moltres exactly the same between tiers.

I'm against the idea of councils actively curating lower tiers, and I think a proposal like this asks them to either do that, or to do nothing. I think we're better off locking major tournaments to set metagames at the beginning of the tour (or at a specific point before the tour starts) rather than mess with how we do tiering in a major way for the sake of better tournaments. Honestly, tours should always have been meta-locked from the beginning.

If we were to wait 12+ months in a gen to implement this policy, we would need a solution to tier instability for tournaments anyway. And I think the best solution is to lock the tiers played in tournaments for the full length of the tour (and let the ladders develop their own meta). Tournament play and ladder play has a stark enough divide at this point that I don't think it's unreasonable to say that only a minority of teams are tested on ladder, outside of particularly big tournaments like the first two or three weeks of Slam, where tiers are unlikely to have major shifts anyway.
 

DoW

formally Death on Wings
is a Pre-Contributor
I think this is a great idea and appreciate the input. Originally when I was writing this, I was thinking about the idea of raising the "3 month, 3.41%" criteria to make it harder for Pokemon to rise and fall. While raising the percentage may solve the problem to an extent (although I'd have to dig deeper into the numbers), it would also make it much harder for Pokemon to fall, which is something no one wants. However, extending the 3 month criteria to 6 months ONLY for rises (but still letting Pokemon drop after only 3 months of weighted 3.41% usage, as they do currently) sounds like a really great solution. It's a much less drastic change than what I proposed, and keeps the current philosophy of strictly usage based tiering. At worst, a 6 month turnover rate for Pokemon like Rotom-W and Amoonguss that hover around the border of 3.41% is much more manageable than a 3 month one.

If people do have a problem with the huge change in philosophy that my proposal would require, then this seems like an excellent compromise.
How about two numbers, one to rise and the other to fall, then? To clarify: currently, a pokemon being used once in 20 games ( = 3.41% usage) means it is in the tier, regardless of what has happened previously. But is it worth considering a setting where a pokemon in UU must be used perhaps once in every 15 games for it to rise to OU, but then its usage in OU has to fall to maybe once in every 25 games for it to drop back down? This way, there would be more stability to the system, while still causing the tiering to be based on usage, with all the positives this provides. However, this option would require careful thought into the numbers picked, to ensure that there was enough stability without preventing changes from happening as quickly as they really should. The idea of a pokemon's tier being based on where it was previously may also raise some concerns, possibly.


Another possibility would be to make moving a pokemon down a tier (OU -> UU) equivalent to moving a pokemon up a tier (UU -> UUBL or OU). By this I mean that the usage statistics give us the default result, but can be overruled by councils / suspect tests with justified reason. In all the scenarios the OP lists, the council could either step in or hold a suspect test, which would resolve the issue if the pokemon in question was kept in the lower tier; however it would not prevent pokemon that were slow to find their niche or that only became good due to another pokemon's ban from rising to the relevant tier. It would also solve the age-old issue of a pokemon remaining in a higher tier purely because people see it listed in that tier on the PS! teambuilder.


I'm not a fan of any options that involve large timeframes. For example, one way to resolve all the issues listed in the OP would be to increase the statistics taken into account from 3 months to 6 months, preventing short-term swings and fashions from massively changing the shape of the tiers. I think this, and others involving large numbers of months, are worse options - one of the great things about usage-based tiering is that it accurately reflects the current metagame, and acknowledges that this will change over time. This is one of the things that keeps the games from being dull, and I think giving up on it would be a big mistake.
 

power

uh-oh, the game in trouble
is a Pre-Contributor
While raising the percentage may solve the problem to an extent (although I'd have to dig deeper into the numbers), it would also make it much harder for Pokemon to fall, which is something no one wants.
I wanted to point out that there's no law that forces us to use the same number for both falls and rises. In the past, when Quickrises existed along Quickdrops, we enforced different cutoffs since they were "Quick" and thus only intended for major metagame shifts.

Although this logic doesn't extend to the normal three month tier update, we could still easily use slightly separated cutoffs. For example, we could enforce a 4% cutoff for a Pokemon to rise and a 3% cutoff for a Pokemon to drop. (This is just an example; numbers could be chosen later.)

Enforcing different cutoffs sacrifices a tiny amount of logical consistency for a great deal of stability; I think it's worth exploring.
 

Adaam

Light Machoke
is a Tiering Contributor Alumnusis the 8th Grand Slam Winner
Not much to add here, but UU got hit with another temporary loss with Serperior rising due to people spamming a veil RMT. Considering it's trash in OU otherwise, it seems pointless to have UU adapt to losing it only for it to inevitably drop next tier shift. I think increasing the rise % is a good solution to handle random spikes in usage that don't accurately reflect the metagame.
 
I would like to see what needs to be done to get this moving forward. However, there hasn't been a whole lot of discussion yet, so I would like to address potential problems that might arise from a change like this.

Right now the options are:

1) a Pokemon must meet the 3.41% criteria for two consecutive 3-month periods in order to rise
2) a Pokemon must meet some higher percentage for the 3-month period in order to rise
3) get rid of rises
4) no change

I prefer option 1 out of all of these, I think the pros and cons are pretty clear for each option. Can anyone foresee any issues in tiering, overall policy, etc, that would stem from this change?
 

david0895

Mercy Main Btw
While I prefer the elimination of the rises, I agree that the 2 consecutive 3-month periods are the best option, because if we negate the rise of a pokemon that initially dropped at the start of gen and then raised, it will be played on multiple tiers for the rest of it.

Especially if we count that the all tiers needs to be cleaned from all the broken stuff that drops that could negate their viability...
 
I honestly feel like option 1 would be the best as choosing a higher threshold for rises would be a subjective process (right now nobody is sure what the threshold would be). Also choosing a higher threshold doesn't necessarily mean a mon will stay in the tier despite it being promoted in a higher tier RMT. RMTs are hot for a few months, and 6 months is long enough to ensure factors like this won't affect the tier.
 

power

uh-oh, the game in trouble
is a Pre-Contributor
Before taking a vote, I'd like to have option 2 be fleshed out a bit more - what would be a better percentage to use?
I proposed this option in my earlier post, so I'll elaborate quickly.

A Pokemon with 3.41% usage has a 50% chance of being seen at least once in 20 games.

The new cutoffs I envisioned for rises and drops would be:

3.7776% cutoff for Rises - This represents the chance that a Pokemon would be seen once in 18 games.
Retain the 3.4064% cutoff for Drops - This represents the chance that a Pokemon would be seen once in 20 games.

I've provided a small Table of Cutoffs: TLs can feel free to choose whatever cutoffs they prefer if the above option is not preferred. 19 as the cutoff for rises also seems like a good and less drastic option.

1532167145047.png
 

Pearl

wide
is a Tournament Director Alumnusis a Site Staff Alumnusis a Community Leader Alumnusis a Community Contributor Alumnusis a Tiering Contributor Alumnusis a Contributor Alumnusis the 7th Grand Slam Winneris a Past SPL Champion
Most of the aspects that I consider important in the making of this decision have already been brought up, so this is not going to be too long. Excuse me if not all of my statements are properly backed up, since raw numbers are far from being my area of expertise. With that said, I feel like times have changed since the current cutoff for rises/drops was initially adopted; information gets spread a lot faster, metagame trends impact the ladder a lot harder and tiers are a lot more volatile overall, a fact that isn't exclusively a consequence of rises and drops alone. As it stands, option B with 17-18* needed (as per the table above) or option A with 19-20* game needed are the best options to me (numbers subject to change depending on what more experienced people have to say), with A sounding slightly better overall due to the point Arifeen brought up earlier in his post (though I think that the cutoff number could still be raised slightly even if this method is adopted).

* edit: Meant either number for both rises and drops (since I'm personally unsure at the moment)
 
Last edited:

MZ

Ban all DLC
is a member of the Site Staffis a Community Leaderis a Community Contributoris a Tiering Contributoris a Top Contributoris a Top Smogon Media Contributoris a Top Team Rater Alumnusis a Battle Simulator Moderator Alumnus
PU Leader
My biggest overall concern is when things will rise and then immediately drop the next month/shift. Both options seem like they would reasonably address it, but option 1 still leaves too much potential for essentially the same thing to happen, with scenarios like Altaria in gen 7 PU (rose for 2 shifts and then dropped back down) not actually changing. Basically I agree with Hogg down to not really knowing usage stats well enough to suggest the actual threshold.
 
The reasons I said I preferred the 6 month period over the higher percentage are:

  • The number of games thing is arbitrary as fuck. Why in particular is 1 in 17 or 1 in 18 an accurate representation of a Pokemon's value in a tier? If we've been using 1/20 this whole time, we might as well continue using it. Moreover, if we're saying that below 1/20 usage is acceptable for a Pokemon's "value" to drop, why is a different percentage necessary to send its value back up? (sorry for poor wording)
  • More importantly, the higher percentage cutoff option wouldn't even necessarily solve the problem; for example, Mamoswine's usage went from 1.2% in January to 5.3% in February, and was back down to 3.0% by April. Azelf in gen 6 went from regular 2.3% usage to an incredible 6.1% usage for the first month of OLT. Those 5.3% and 6.1% numbers aren't even high enough to be on sedertz's list and probably require like 1/12 and 1/10 game values, which is quite ridiculous to set the bar for rises. And of course Arifeen linked another example of the higher percentage option failing. The point is that trends like the ones I talked about in my original post (OLT, popular person influence, RMT etc) can send a Pokemon's usage to well above reasonable values for a rise cutoff, only to fall back below the current cutoff 3 months later, which is exactly the problem we are trying to eliminate. However, I can't find a single example of one of these trends lasting for 6 months. And honestly if a Pokemon is used in OU 1/20 times over a 6 month period, I think it's safe to say at that point that the Pokemon's value is OU.
 
User Nailec wanted to add the following:
I see you're rethinking the tiering policy concerning rises and I'd like to contribute to the discussion with a new idea and some analysis on the current system and the other proposals. I don't have a badge yet but I have a good knowledge of how usage stats are calculated and how tier shifts result from it.

The current algorithm used to produce usage stats is made to predict what will be the usage for each pokemon in the following month, meaning it aims to get a picture of the metagame as accurate as possible at the time it’s ran. That works extremely well at the beginning of a generation and allows to form all the tiers rather quickly. However with time passing the main concern is having stable tiers, so what we’d want the algorithm to do is getting a longer range picture of the metagame, to prevent "noise" in rises and drops.

Here are what I think are few flaws of the current method of determining the tier shifts:
  • Tier shifts wouldn’t be the same if they were done at different months. Amoongus rose from UU to OU in April, when its weighted usage for the previous 3 months was 3,6 %. If the months where tier shifs are calculated were February and May, these percentages would have been 2,91 % and 3,3 %, both under the cut for rising.
  • Rises happen right after meta defining changes. For example, this month Torterra, Sandslash and Stoutland rose despite their time being viable in NU being gone since Gigalith was banned.
  • The 20-3-1 weighting helps trend based shifts. This month Serperior reached 3,89 % in usage, after being under 3 % the previous months. It seems excessive to have it rise already.
I've read the different solutions proposed, all with their advantages and drawbacks and here's my opinion on them:
  • Increase the cut for rises. With a slightly higher cut for rises Amoongus and Serperior wouldn’t have been back in OU however this alone doesn’t solve anything since Mamoswine and Ambipom incidents would have still happened.
  • Get rid of rises. If we did this a year ago, Palossand would have more than 10 % usage in NU and PU, which looks difficult to understand from a player’s point of view and to me having a pokemon a major force in 2 different tiers is against the tier principle.
  • Only rise if above the cut in 2 consecutive. While this looks like the favorite of many this not only still has the same problems (Qwilfish and Altaria in PU as Megazard mentionned) but also add a layer of complexity on an already complex method. I believe a rule such as « being over the cut for 3 consecutive months » would solve the same problems as this one while being able to keep the rhythm of changes.
With this being said I come with my own proposal: Several consecutive months over/under the cut to rise/drop (using unweighted stats).
First of all, I like the idea of keeping the same methodology for rises and drops so it doesn’t look like something was thrown together just for some very rare cases. It is also very easy to understand and solves most of debatable shifts so let’s have a look at how it affects rises since January using 3 consecutive months for rises and drops as an example :
  • January
    • Rotom-Wash (UU → OU) : Still rise, but 1 month later
    • Kommo-o (RU → UU) : Still rise but 1 month later
    • Rotom-Mow (NU → RU) : Still rise but 2 months after
    • Seismitoad (NU → UU) : Its 3rd month above the cut happens 1 month after Rotom rose, difficult to tell if this would have still happened, but in the end it still has solid usage in UU so it seems at the right place
    • Qwilfish (PU → NU) : Its usage was linked to Emboar’s, I might miss something from 2017 NU but blame NU users for letting it drop in the first place. With Emboar banned Qwilfish obviously dropped back.
    • Silvally-Steel (PU → NU) : Still rise, but 1 month later
  • February
    • Ambipom (NU → UU) : No rise since it was above the cut only 1 month.
  • March
    • Altaria (PU → NU) : Still rise, mostly because of Venusaur, not sure it would have made sense to prevent it. It dropped after the latter was banned, remaining 5 months above the cut.
    • Mamoswine (UU → OU) : No rise, only 2 months above the cut
  • April
    • Amoongus (UU → OU) : No rise, only 2 months above the cut
    • Miltank (PU → NU) : Still rise
    • Palossand (PU → NU) : Still rise
  • July
    • Gliscor (UU → OU) : No rise yet (2 months so far)
    • Serperior (UU → OU) : No rise yet (1 month so far)
    • Gastrodon (PU → OU) : Still rise
    • Moltres (RU → UU) : No rise (only 2 months, but not in June)
    • Toxicroak (NU → RU) : No rise yet (1 month so far)
    • Sandslash (PU → NU) : No rise (only 1 month)
    • Torterra (PU → NU) : No rise (only 1 month)
    • Stoutland (PU → NU) : Still rise
    • Ferroseed (PU → NU) : No rise (only 1 month)
This of course has a major drawback, it gives shifts every single month, which is against the stability of tiers we're trying to have. So in order to fix this, pokemons matching the requirements would stack in a sort of buffer (the actual BLs for drops), until tier leaders decide to implement the changes for their respective tiers. This could happen at any month (not necessarily every 3 months) and not even all the tiers together.

Thanks a lot for reading this, I hope you'll share the idea with the others. If you need me to detail any point or have any question I'll be happy to answer. Have a nice day !

Nailec
The point about weighted stats is a good one and one that I hadn't even thought about. If Serperior's usage had spiked in one of May or April, it wouldn't have risen. But since it happened to be in June, the brief increase in usage was enough to push it to OU despite having low usage the previous two months (due to the 1-3-20 weighting method). Like Nailec said, this just adds to the problem of short-term trends having an unnecessarily large role in tier shifts.
 

Quite Quiet

1 + 1 = 3
is a Tournament Directoris a member of the Site Staffis a Super Moderatoris a Community Contributoris a Tiering Contributoris a Contributor to Smogonis a Top Smogon Media Contributoris a Battle Simulator Moderator Alumnus
TFP Leader
Just as a side note, I don't have time to do the math for this, but would it not also be an alternative to change how much emphasis the current month has on the 3-month period. 1:3:20 values the current month 20 times as much as the one furthest back. This, as others said, helps well-timed short term trends to make a much larger impact on the averaged usage than it otherwise would. I don't recall right now where the 1:3:20 came from (probably X-act or Antar if I had to guess though) but is there an alternative weighting like 4:8:12 that we could use to lessen the impact temporary trends has on the three month averages. I'm not too fond of needing 6 months to raise a Pokemon because it seems like it'd impact the tiering process due to Pokemon that are rightfully used in a higher tier not as a trend (Contrary Serperior in gen 6, if the quickbans didn't happen, for example).
 

Bughouse

Like ships in the night, you're passing me by
is a member of the Site Staffis a Forum Moderator Alumnusis a CAP Contributor Alumnusis a Tiering Contributor Alumnusis a Contributor Alumnus
there's really not a huge reason to weight at all. the only point to weighting is to be responsive to trends and catch them earlier rather than later... which is exactly what this thread is seeking to eliminate.
 

phantom

is a member of the Site Staffis a Community Leaderis a Community Contributoris a Top Contributoris a Tiering Contributor Alumnus
RU Leader
Figuring out the threshold for rises would be ideal, but I don't think it should come at the cost of slowing down any progress this thread is making...which seems to be what it's doing. I think proceeding with some sort of decision or vote and then addressing the possible percentage for rises would be ideal, provided it's the course of action taken.

As for my thoughts, I think 6 month tier shifts and removing weight would be the best. Raising the threshold for rises is something that would not properly deal with issues at hand. Sudden spikes in usage that result from ladder tours/suspect tests/short-lived trends still have the potential to influence usage stats such that raising the threshold would be meaningless. Furthermore, given how consequential rises are for tiering, it only makes sense that they occur only after an extended period of time to prove that whatever is moving up remains good far beyond the duration of its trend. I also don't think raising the threshold would matter much if the third month is as heavily weighted as is. As for the way the months are currently weighed, I think removing weight has the potential to be beneficial for boths drops and rises since it prevents breakout trends in the third month from suddenly exploiting a fault in the system, and it also reduces the influence of third parties who would attempt to "spam" a particular pokemon in the third month at the cusp of dropping to keep it in its tier, which has happened quite a few times in the past.
 

Quite Quiet

1 + 1 = 3
is a Tournament Directoris a member of the Site Staffis a Super Moderatoris a Community Contributoris a Tiering Contributoris a Contributor to Smogonis a Top Smogon Media Contributoris a Battle Simulator Moderator Alumnus
TFP Leader
Proxyposting for Ohmboy Tesla:
I've been following Antar's 'Usage Stats' updates and more recently the 'Lower Tier Shifts' threads pretty closely for a long time now and I would like to offer my insights and a slightly different perspective to what has already been proposed.

I agree that all weighting should be removed as it disproportionally favors usage in months that happen to fall "on the right moment", which allows for easier manipulation of usage stats, should someone want to do that. Usage in one month shouldn't by default count less than that in another one imo. That's not the main problem this thread is trying to solve (i.e.: fluctuations in lower tiers when shifts occur), but it's a longtime problem in need of a solution, and this thread can solve both issues at once.

Manipulation can still happen as has been discussed in the thread, which is why, at the same time as removing the weights, the "window" for the stats should be increased. I would propose to raise it to 6 months. Things that rise/drop would now have to have a prolonged (non-)usage in the tier and spikes in 1 month aren't enlarged by weights. Trends are caught onto less quickly, but higher tiers can use all lower 'mons anyway, so that shouldn't be a huge problem. The long window provides the wanted stability.

1 problem remains: pokémon that hoover around the threshold. Only one solution exists for this: different thresholds for rises and drops. The current cutoff of 3,41% is the usage% a 'mon should have so that you have a 50% chance of encountering it at least once withing a day (=20 battles) of playing. People were discussing the "randomness" of other thresholds, but I believe the explanation is fairly simple. 3,58% usage (50% chance for 1 in 19) just means you have MORE than 50% chance of encountering it within a day of playing. Similar for 3,25% (50% chance for 1 in 21) meaning you have a LESS than 50% chance etc. etc.
Which exact cutoffs should be chosen is up for discussion (3,58 and 3,41 / 3,41 and 3,25 / 3,58 and 3,25), but without a form of hysteresis (= different thresholds) there will always be "hoovering" pokémon, no matter which system is in place.

Only a combination of unweighted stats, a longer timeframe, and a hysteresis for rise/drop cutoffs makes it so that 'mons rise/drop when their usage is high/low for a long enough period, without a (high) chance for manipulation and without pokémon that hoover around the usage cutoff.

Now I know a lot of what I said has already been mentioned in some form in the thread, but I can't stress enough how all 3 of those things need to happen together in order for it to work.
And this is also where I would like to go against a lot of the comments so far:

With such a system in place, I don't believe tier shifts should only occur every 3, let alone 6 months, but might happen as often as every 1 or 2 months.

Let me explain why:
Because of the large window and the hysteresis between rises/drops, pokémon simply can't rise and drop that quickly anymore. Here's a scenario:
Say pokémon X just rose a tier and the cutoffs are 3,58% and 3,41%, meaning it has at least 3,58% usage over the last 6 months. In order for it to drop again next update, it would have to go below 3,41%, but 5/6th of it's usage is still made up of that higher one. This would mean it's usage would have to drop below 3,41% = (5/6 * 3,58%) + (1/6 * X%) with X being 2,56% in the sixth month. At that point, if the usage falls below 5,56%, the drop should be justified IMO.
These are the numbers for the other cutoffs:
3,41% and 3,25% => below 2,45% usage
3,58% and 3,25% => below 1,60% usage

Tiering still needs to happen, it's the core concept of Smogon's metagames, and with an update every 6 months and a lot of generations/versions only lasting 2 years, you would only have 3 updates after the start of a generation. If you read the comments from the Usage Stats Update threads, you'll notice that a lot of players actually want pokémon to rise/drop once their usage justifies it, especially when you put stops to blatant manipulation as I mentioned. With the proposed solution, you'll start to see that less 'mons will be shifting tiers every update. With a 6 month window, if a 'mons usage just happens to start rising/declining in the "3rd month" or so, you risk missing the next update an now you're waiting 9 months to shift the tier of a 'mon that should've done so almost a year ago.

That's my input on the matter. I hope I could contribute in a meaningful way.
I don't think going shorter than 2 months is realistic, and 2 might be too short as well, but it might be a better system than 6. Consider that with 6 you can have Pikachu at 50% usage in OU for 5 months straight and still be PU, that's not an accurate representation of the tier list either. Anything we do should be hard enough to change, while still allowing for tier lists to change often /enough/ to be accurate.
 

Sam

i say it's all just wind in sails
is a Battle Simulator Admin Alumnusis a Community Contributor Alumnusis a Tiering Contributor Alumnusis an Administrator Alumnus
I looked at some of the examples, and to avoid any of these would have to be pretty high, like over 5%.

A lot of the issues seems to stem for the weighing, so I think to begin the next generation we can remove that and see how it looks. If further action is needed we can look into extending the rise/drop period.
 

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

Top