Three Proposals Regarding Tiering

[Great Sage]

A few weeks ago, X-Act, Doug, me, and some other people were discussing possible changes to the tiering system. What we discussed boiled down to three general proposals. Note that this is only a preliminary test of public reaction, not anything to be acted upon immediately. Still, feel free to discuss these things a bit.

1. Instituting a weighted usage system
This proposal will serve as the backbone to the second proposal. Through our discussion, we decided that the fairest way to accumulate these statistics is to do something similar to what is done on the Official Server, but with only the winner of a battle's Pokemon counted, and with all Pokemon on the winner's team counted, regardless of whether he used the Pokemon or not. For example, if the winner had a Garchomp, but it didn't appear in the battle, it would still be counted in the usage statistics, because by putting a Pokemon on a team, a person demonstrates intent of using that Pokemon.

Because the second proposal is rather sensitive to seemingly minor changes in usage, it is critical that only wins count towards usage; assuming a person takes a single hour to knock off the statistics each day, and that he gets a battle on the ladder once a minute, which is an extremely conservative assumption, that person would be able to shake the usage of not one, but six Pokemon by nearly 2000 each.

2. Defining when a Pokemon is overcentralizing, using weighted stats
Through our discussion, we decided that when the most popular Pokemon is used 30% more than the next most used Pokemon, it should be considered overcentralizing. The 30% figure is admittedly arbitrary, and is open for debate. However, there is a possibility that a bloc of Pokemon is overcentralized, which wouldn't necessarily be reflected by comparing only the first Pokemon to the second Pokemon. Therefore, using a three-month aggregate of usage statistics, we would look at the ratios between the first and second, the second and third, and the third and fourth Pokemon. The extent we do this to is also debatable. If any of the ratios exceeds 1.3, the first Pokemon, and the first Pokemon only, is banned. This way, we can not only gradually remove overcentralizing blocs, because if the second Pokemon is truly overcentralizing, it would likely become considered so during the next review, but also avoid banning Pokemon that are only used frequently because of the existence of a truly overcentralizing Pokemon.

3. Creating a new tier
We noticed that the BL tier is, in a word, enormous; we agreed that, ideally, the size of faux tiers should be about the size of Ubers, or around 20 Pokemon at most. Pokemon like Mesprit and Flygon could easily belong in a new balanced and playable tier between UU and OU, cobbled together with a majority of currently BL Pokemon and some currently UU Pokemon.

It is almost an inevitable fact that when a pool of over 270 Pokemon is crammed into three tiers and three faux tiers, and when the total size of the top tier and the top faux tier is less than 70, that the lower tiers either include enormous faux tiers, or are playable tiers with long tails of rarely used Pokemon. By introducing a new tier and a new faux tier, this congestion may be alleviated.

Part of the reason to introduce this new tier between OU and UU is marketing; the public remains adamant that DP's UU must conform with ADV UU's power standards, therefore suppressing the fluidity of the BL tier. The new tier would be capable of allowing many Pokemon that can form among themselves a balanced environment, and can depress the current UU's power, hopefully mollifying the public to some extent.

 

[Aldaron]

I like the idea of only counting the winner Pokemon, and counting all of them, as usage in battle is definitely influenced by what I can potentially use, not just what I am using.

I also agree with using weighted stats, but I'm iffy on the 1.3 line. I know you said it was arbitrarily chosen, but I think we can more effectively determine a "good" line after viewing these statistics.

BL's tier size is a joke. I've been lamenting how it was created for a while now, and a lot of my complaining has been because of our tendency to conform to the desires of the public wanting a similar uu tier (in power terms..."generally speaking" lol)

Hopefully the (expedient) implementation of this "bl" ladder will allow us to lay out the next "nu" tier, and allow us to see what we have to do after that.

 

[Great Sage]

The point is, even if UU and NU are laid out nice and neatly, it's still almost inevitable that they can't contain the 200 or so Pokemon they have to handle without stuffing a large bunch of them into faux tiers.

 

[Aldaron]

That's why I said "allow us to see what we have to do after that" because there is no reason to only have 3 tiers.

If NU ends up having "200 Pokemon," we can easily create a lower tier.

 

[gengod]

Part of the reason to introduce this new tier between OU and UU is marketing; the public remains adamant that DP's UU must conform with ADV UU's power standards, therefore suppressing the fluidity of the BL tier. The new tier would be capable of allowing many Pokemon that can form among themselves a balanced environment, and can depress the current UU's power, hopefully mollifying the public to some extent.

As of right now the public is more adamant about combining the current UU and BL to create what should have been the original UU tier. Only a few people think that the DP UU's power standards should conform with the ADV power standards.

I don't think there is much need to create any new tiers because as we reform the BL and UU tiers its more likely that there are going to be far fewer pokemon in the BL faux tier than they are now and if a new tier has to created (although its rather unlikely) then I agree with Aldaron as we can simply make another tier after NU.

 

[Blue Kirby]

I'm with Maniac on this one. The notion of DP UU needing to confirm to RS UU really doesn't seem all that widespread to me. Simply reforming our current UU/BL stance will allow us to move forward, as then we'll finally be able to deal with NU.

 

[X-Act]

About the overcentralising thing, I've been studying it a bit. I literally don't know how to state objectively that a Pokemon is overcentralising. However, I've seen a pattern in the usages of Pokemon, which should help me make a definition.

As I've been saying before, usages of Pokemon seem to conform to an exponential distribution. In other words, the Pokemon's usages seem to be just a geometric progression. If Pokemon #1 is used a times, then Pokemon #2 is used ar times, Pokemon #3 is used ar^2 times, Pokemon #4 is used ar^3 times, and so on. Of course, they are not an exact G.P., but they seem very close to one, with a correlation of at least 0.96 or better (a correlation of 1 would be a series that corresponds exactly to a G.P.). Apart from that, since we know that the sum of the percentage usages is always 100% (or 1), we can infer that a = (1-r) / (1 - r^n) where n = 498 is the number of Pokemon. Hence we can state that u_i, the percentage usage of the ith Pokemon, is roughly equal to (1-r)r^(i-1) / (1-r^n) for a 'parameter' r.

Now, the first thing I need to answer is: why do Pokemon usages always conform almost perfectly to a G.P.?

Secondly, I've found that the first 40 or so Pokemon are roughly always over the expected value that the G.P. tells us it should be, while the next 80 Pokemon or so are always roughly below the expected value. This is sort-of akin to a cosine wave: If you imagine a cosine wave having a periodic time of 160, it would be above the x-axis between 0 and 40 and below the x-axis between 40 and 120. Again, I don't know why this happens, but it seems to happen many times.

The reason why I'm trying to find a pattern in the usages of Pokemon is because if one Pokemon is used much more than the pattern should suggest, it should be grounds for being called overcentralising.

I'll let you know how that goes.

 

[Hipmonlee]

I agree with everything, except the 1.3 thing. I mean, I worry about us getting locked into a system that might end up failing.

I mean, how would we introduce pokemon? We have seen with both Garchomp and Deoxys that they can take a long time to establish themselves. If Deoxys breaks the game, but is used slightly less than Gengar, because of laziness, and people's principles or due to the fact that people didnt realise its power until halfway through the month.

I just worry that using a statistical method, when we dont fully understand the statistics, means that our method might prove flawed under certain circumstances, and we might not realise it until we had done some significant damage to our rules.

That plus the possibility that we end up banning a different every month and never stop, or end up never banning anything because of the 1.3 figure we chose arbitrarily.

Have a nice day.

 

[X-Act]

That's why I'm researching the usage patterns of Pokemon: to make a better (and, hopefully, correct) definition of overcentralisation. That means that I don't agree with the 1.3 thing either... even though I suggested the method originally (although I didn't come up with the number 1.3).

I'm also considering changing the 75% cut-off point of OU to something else, for two reasons. The first is that Doug's usages now count also Pokemon that are NOT used in battle but that were in teams (unlike Colin's usages, on which the 75% was based). Secondly, studying the usages of Pokemon might give me a better cut-off point anyway.

 

[imperfectluck]

I agree entirely with only winning teams' pokemon being counted... I've discussed "poisoning" the statistics pool before (although I didn't do it!) and I'm sure some people here remember that 1600+ ranked Sunkern player (me!)

I also think that creating a brand new tier for those Mespirits and Pinsirs and so on to shine is a pretty good idea, who ever said that we have to have only Uber, OU, UU, and NU tiers for 4 main metagames? (As BL was never meant to be a metagame) With a hundred+ more pokemon than in DP, it's certainly viable that there would be even more discrepancies between Pokemon and therefore a need to create more than 4 main tiers.

 

[obi]

In addition to only counting the Pokemon on the winner's team, I recommend also weighting it by the rating of the loser. This means that teams which beat highly ranked players will count for more than teams that beat lower ranked players.

 

[Great Sage]

As I've said on Shoddy, I agree highly with that idea. Beating an idiot and beating a good player shouldn't count identically towards a Pokemon's worth.

 

[Tangerine]

In addition to only counting the Pokemon on the winner's team, I recommend also weighting it by the rating of the loser. This means that teams which beat highly ranked players will count for more than teams that beat lower ranked players.

Isn't this rewarding luck? If a lower rank player beats a higher ranked player - it is likely more due to luck than their skill. Also considering how low of a ranking you start when you first make a new alt (and since everyone uses an alt anyway) the ratings will be a bit skewed.

Anyway, I agree with the OP - and as people have said, minus the 1.3 line - the number should be decided after we look at statistics a bit - maybe we can use Garchomp as a precedent.

 

[obi]

The rating of the winner is irrelevant. It also doesn't "reward" anything. This proposal has nothing to do with determining the ratings of the player, just the rankings of each Pokemon. To that effect, taking 'luck' into account is actually a good thing. This is because certain Pokemon can take advantage of 'luck' better than others, and those Pokemon ought to be rated higher, all other things being equal.

 

[Tangerine]

This is because certain Pokemon can take advantage of 'luck' better than others, and those Pokemon ought to be rated higher, all other things being equal.

I'm pretty sure most of the times you get lucked out is not by Pokemon like Absol or Kingdra but I guess you have a point!

The rating of the winner is irrelevant. It also doesn't "reward" anything. This proposal has nothing to do with determining the ratings of the player, just the rankings of each Pokemon.

Gotcha. I misunderstood. If that's the case I think it's a pretty good idea

 

[Cynthia]

Just as a side note, if NU is to be based off of UU usage statistics there likely needs to be a faux tier between NU and UU to balance the NU metagame.

 

[Aldaron]

of course cynthia

 

[X-Act]

I've been thinking about the following reasoning.

The Uber tier is just a ban list for the Standard metagame.

The Uber, OU and BL tiers together is a ban list for the UU metagame.

The Uber, OU, BL, UU and the ban tier for NU together is a ban list for the NU metagame.

So the OU tier's only function is really that of stating which Pokemon are banned from UU, exactly like the BL tier. The only difference between OU and BL is that the OU tier bans Pokemon from UU because of usage in the Standard metagame, while the BL tier bans Pokemon from UU because of power.

So, to summarise:

Uber - A ban list for the Standard metagame based on being too powerful for it.
OU - A ban list for the Underused metagame based on usage in the Standard metagame.
BL - A ban list for the Underused metagame based on being too powerful for it.
UU - A ban list for the Neverused metagame based on usage in the UU metagame.
"Ban list for NU" - A ban list for the Neverused metagame based on being too powerful for it.

All tiers are ban lists, then.

Now, I need to know just how big the OU and UU tiers need to be. Remember, the larger they are, the more Pokemon are banned in the metagame immediately below them. This will help me also for this thread.

By the way, I'd call 'Uber', 'BL' and 'Ban list for NU' as 'SBL', 'UBL' and 'NBL' respectively - short for "Standard Ban List', 'Underused Ban List' and 'Neverused Ban List'.

 

[Hipmonlee]

I suggested this a while back, but it was met with all around disapproval.

I think people want an OU tier that is an indication of what is commonly used, so they can know what to build teams to counter or whatever.

Granted that if we have useage stats then that is kinda unnescessary, but I dunno, nobody else agreed with me.

Have a nice day.

 

[X-Act]

The OU tier IS a list of Pokemon that are commonly used in Standard, actually. The Pokemon that are commonly used in Standard are then banned in the Underused metagame. That is the sole function of the OU tier, and hence the sole reason for finding what is commonly used in Standard: to ban those Pokemon from UU.

 

[Hipmonlee]

Yeah but it's not the sole function. People wanted it to be a quick reference to see what pokemon are commonly used as well..

I agree with you personally, this is my devils advocate.

Have a nice day.