OU GSC OU Viability Rankings mk. 4

Mr.E

unban me from Discord
is a Two-Time Past SPL Champion
tbh my opinions haven't changed a whole lot post-SPL from whatever they were last time, other than maybe I wanna give Golem a little more credit in general (and I don't remember what I ranked it without digging up the other topic) and Rhydon the leg up on Machamp/Marowak for having more substantial defensive utility
 
Wanted to rank 83 pokemon just to 1 up choolio from last year, so I did. Although, if I'm being honest, anything lower than 50-60 I was just throwing darts at.

Seems a bit silly to ask myself if Entei is better than Shuckle or some shit, but whatever.
 
I'm very new to GSC but I've noticed something pretty prominent in the meta game from just a dozen games. that skarmory can be slapped on most teams and works very well as a snorlax check and a very good phazer in the mid game. I've also noticed after taking a look at the viability rankings that suicune could be a bit high for its actual worth in game. I understand that it can work on stall teams but I don't know what else it would be used for in offensive teams or even BO. for these reasons maybe skarmory could be moved up a level from where it is now. and suicune down from A+ -> A or A- I would love to hear thoughts and explanations why I'm inevitably wrong on both of these topics.
 

Zokuru

The Stall Lord
is a Tiering Contributor
I'm very new to GSC but I've noticed something pretty prominent in the meta game from just a dozen games. that skarmory can be slapped on most teams and works very well as a snorlax check and a very good phazer in the mid game. I've also noticed after taking a look at the viability rankings that suicune could be a bit high for its actual worth in game. I understand that it can work on stall teams but I don't know what else it would be used for in offensive teams or even BO. for these reasons maybe skarmory could be moved up a level from where it is now. and suicune down from A+ -> A or A- I would love to hear thoughts and explanations why I'm inevitably wrong on both of these topics.

To be quick, if you use Skarm outside of stall, or at least a team able to lose a bit of momentum and spin, you're more or so losing to Zap sooner or later, also Suicune is a big staple in stall teams, and a massive pain to deal with for most Pokemon used to pressure stall teams, it doesn't matter if you can't use it in every team, when it's played it's always very good, as good as most A+ rank. A lot of A+ mons aren't ranked for their potential in stall btw .. and Cune is A-
 
I think one argument against Cune, is that even in stall it is getting used less, as it often competes with Zap for the resttalk tank spot. It definitely plays a very different role, but I think the increasing popularity of that kind of stall team as opposed to more "traditional" archetypes could go against Suicune. One example is I think heal bellers I see less of. Zapdos fits a bit better on healbellless stall, as a self-sufficient tank with spikes immunity. A lot of little subtle things like that contribute to my seeing less of Suicune around, amongst better players/teams these days.

On the more offfensive side, I want to say that current meta trends are more and more squeezing Marowak, and especially Misdreavus out of use, ass they don't fit with the really fast paced, put down spikes quickly, capitalize quickly, centered offensive archetype that has become the rage. You need to open up some kind off hole for Marowak to operate best, as opposed to something like Nido which threatens earlygame with utility, as well as being a late game monster with spikes up. Misdreavus is useful trapping or utility spinblocking, both for offense or stall, but will never carry the same kind of versatile instant offense threat that Gengar always does with it's insane movepool to advance offensive gameplans.
 
Last edited:
It's time again for the results of the annual GSC OU VR update survey. Thanks to everyone who took the time to contribute to this project and I hope you enjoy perusing the results.

Collection of data
- This year I used a SurveyMonkey ranking question instead of a Google Form, since the ranking question type is absent in Google Forms.
- While the survey did result in fewer people doubling up on rankings, the form was quite awkward because the drag-and-drop feature didn't work as one would expect.
- Last year's VR was used as the basis for what Pokemon to include as options in the ranking question, however prior to conducting the actual survey, I surveyed a smaller sample of players to determine Pokemon that should be initially excluded from the form. Pokemon that fewer than 1/3 of those surveyed wanted kept were removed. This resulted in the removal of Poliwrath, Tauros, Raichu, Electabuzz, Lanturn, Crobat, Mr. Mime, Hitmonlee, and Slowking from the form.
- This year, none of the Pokemon were mandatory to rank.
- The link to the survey was posted publicly this year.
- I opened a discussion channel for those who had submitted rankings again this year.
- Upon making a submission, it was possible to view anonymised results.
- It was also possible for players to make edits to their rankings after submitting the survey.
- The survey was left open for about two weeks.

Formation of rankings
- This year, vapicuno did most of the work in forming the rankings.
- I did give some input into how to name and split up the tiers, particularly with regard to B tier and below.
- I removed one submission which was suggesting to add Celebi to the rankings among other odd decisions.
- The order was determined by an outlier-compensated mean ranking.
- In the cases where players had manually set multiple Pokemon to be the same rank accidentally, I either asked them what they intended or resolved the issue alphabetically (these cases for the most part came up in the rank-50s).
- Any Pokemon which was not ranked by at least half of the players making submissions was removed from the VR (Slowbro and Blastoise).
- Any Pokemon that was included in the free-text box at the end of the form by at least 5 players was added to the VR (Tauros and Hitmonlee).
- See vapicuno's post below for more detailed information.

Flaws
- The biggest flaw in my opinion was the fact that the initial order of the Pokemon was based off the existing VR rather than being randomised, inevitably inducing bias. This was avoided last year but was impractical this year due to the awkwardness of using the SurveyMonkey ranking question with randomised order.
- The form's drag-and-drop feature did not work as one would expect it to. It took some getting used to and many players voiced their frustrations to me about this.
- It was slightly awkward to provide the intended ranking for Pokemon not listed on the form. Most people managed it but it involved some manual work on my end.

The link to the data in spreadsheet form is here.

S1 Rank
01
Snorlax​
S2 Rank
02
Zapdos​

A Rank
03
Raikou​
04
Cloyster​

B1 Rank
05
Gengar​
06
Skarmory​
07
Exeggutor​
08
Tyranitar​
B2 Rank
09
Nidoking​
10
Machamp​
11
Steelix​
12
Golem​
13
Vaporeon​
14
Forretress​
B3 Rank
15
Starmie​
16
Suicune​
17
Jynx​
18
Marowak​
B4 Rank
19
Rhydon​
20
Misdreavus​
21
Umbreon​
22
Heracross​

C1 Rank
23
Miltank​
24
Blissey​
25
Tentacruel​
26
Espeon​
C2 Rank
27
Charizard​
28
Dragonite​
29
Jolteon​
30
Porygon2​
31
Houndoom​
32
Quagsire​
33
Moltres​
34
Alakazam​
35
Smeargle​

D Rank
36
Muk​
37
Meganium​
38
Clefable​
39
Piloswine​

E1 Rank
40
Kangaskhan​
41
Sandslash​
42
Scizor​
43
Donphan​
44
Shuckle​
45
Ampharos​
46
Kingdra​
47
Entei​
48
Aerodactyl​
49
Articuno​
50
Omastar​
51
Venusaur​
E2 Rank
52
Gligar​
53
Jumpluff​
54
Pikachu​
55
Tauros​
56
Nidoqueen​
57
Qwilfish​
58
Typhlosion​
59
Ursaring​
60
Hitmonlee​

Visualisation of results and individual rankings
As with last year, I have put all the rankings together into a single image. In the New VR column, I have formatted it so that green indicates a rise and red indicates a fall when compared to the old VR. Blue indicates no movement. The pale green and red in everyone's rankings indicate what they ranked higher (green) and lower (red) than the new VR.



Observations
Differences when compared to the old VR that I found particularly notable include:
-
Cloyster (-) again closed in even further on Raikou's #3 spot. Last year Raikou's average rank was 3.192 and Cloyster's was 3.769, this year Raikou's is 3.333 and Cloyster's is 3.697. Spikes remain as important as ever, and Cloyster is decidedly the best option for putting them up. Cloyster has remained consistent at ~70% usage in both of the previous SPLs, and it doesn't look like it will be budging from that usage point anytime soon. On the other hand, Raikou is used about half as often at 38.5%, although the competition for its slot tends to be more fierce. Some top players who rated Cloyster above Raikou include ABR, FriendOfMrGolem120, sulcata, and Jorgen.
-
Gengar (+1) has surpassed Exeggutor by a substantial amount due to its importance as a spinblocker and high utility threat, overtaking Exeggutor. Players have reacted to its surge in popularity by using Earthquake more and more, including on Snorlax, Tyranitar, and Machamp, and usage of Nidoking, Golem, Rhydon, and Quagsire is also rising. Nevertheless, it would appear that the importance of being able to keep Spikes has cemented Gengar's spot in the top 5 for the time being. Just about all of the top players rated Gengar as #5 in their rankings, a notable exception being BKC who rated Skarmory above it.
-
Exeggutor (-2) has dropped considerably, falling as far as 7th from its previous 5th spot. This is contrary to its results in SPL, where although it had slightly lower usage, its win rate increased substantially. It perhaps had a bad year after Jynx's usage shot up, and it still faces competition from others like Nidoking, Machamp, and Gengar for its slot somewhat, but surprisingly it has been overtaken by Skarmory in addition to Gengar. The increase in the popularity of Ground-types and Earthquake that is currently taking place should provide an opportunity for Exeggutor to make a comeback next year -- we have even seen SubSeed and other more innovative sets in use recently. Blightbringer and Royal Flush rated Exeggutor as high as 5, but most players tended to rate it 7.
-
Nidoking (+3) has continued its rise and this year has broken into the top ten. Nidoking, along with its fellow common sleeper Jynx, have massively influenced the metagame over the past couple of years and have spearheaded the rise of Thief offense in modern GSC. Nidoking in particular has shown possibly the highest proficiency of any non-Electric type at using Spikes to pressure foes and break open teams with its sleep, coverage, access to Thief, and Toxic + Electric immunity. While the metagame has adapted with measures such as Sleep Talk Snorlax, even this more durable Snorlax struggles to take hits if its Leftovers have been stolen. It has nearly doubled in usage between SPL X and SPL XI and has increased from 54% win rate to 66% win rate. Very few top GSCers rated it below 11 and it has consistently been ranked in the 8-11 range, with a few notable exceptions such as myself and Zokuru rating it at 6 and 7.
-
Steelix (-3) was previously a top ten mainstay, but it has truly fallen a long way since last year. From over 33% usage and a 57% win rate in SPL X to 12.5% usage and a 25% win rate in SPL XI, Steelix is widely acknowledged to have gotten considerably worse in the current metagame. The primary reason I see for this is its competition with Golem and Tyranitar, both of which fit on offensive and defensive teams alike. Steelix's weakness to Fire and its low usefulness after taking a chunk of damage from Fire- or Water-type attacks has caused its popularity to plummet. The commonness of spinners such as Golem and Starmie on opposing teams is also terrible news for Steelix, since it relies on Spikes to be effective with its standard set. Nevertheless, it has managed to contain its fall to three spots below where it was last year, despite ratings of as low as 15, 16, and 18 from McMeghan, ABR, and KratosMana, thanks to relatively high ratings from players such as Fear and choolio.
-
Golem (+10) was by far the biggest mover this year, jumping a ridiculous 10 spots up to reflect its widespread adoption by top GSC players. In SPL X it was used just thrice (3% usage) whereas in SPL XI it was the ninth most used Pokemon at almost 24% usage, with a respectable 47.8% win rate. It has well and truly rolled into the mainstream this year and will likely remain popular as time goes by as teams increasingly look to incorporate Rock- and Ghost-types as well as Rapid Spin on their teams. By now, everyone is aware of the role compression Golem offers a team, but FriendOfMrGolem120 went as far as to rate Golem 6th while KratosMana, sulcata, Jorgen, ABR, BKC, dice, and McMeghan all rated it in their top 10 Pokemon. Time will tell if Golem will crack the top ten in years to come.
-
Starmie (+1) didn't increase its rank by a huge amount, but took over Suicune as the defensive Water-type of choice for many teams for its access to Rapid Spin, while also finding a place as an offensive threat on certain teams with its recent Substitute Nightmare moveset. It has become substantially more popular since last year while remaining statistically effective, with a 75% win rate in SPL X and a ~62% win rate in SPL XI. While no-one ranked it extraordinarily high, players such as Jorgen and Fear rated it as high as 10 and 12. On the other hand, Jimmy Turtwig rated it all the way down at 25.
-
Marowak (-3) dropped an entire three places this year to end up at 18, as Jynx held its spot at 17 and spinners Starmie and Golem overtook it. Marowak's lack of ability to switch in as well as its susceptibility to Spikes are likely responsible for this drop, as well as the advent of RestTalk Skarmory, which is quite capable of shutting down many of its sets. However, I feel as though it would be foolish to expect this downward trend to continue with Starmie and numerous Pokemon that are weak to Ground on the rise and Exeggutor and Jynx on the decline.
-
Quagsire (+5) was this year's second biggest mover, leap frogging Moltres, Clefable, Muk, Smeargle, and Alakazam to land at 32. Quagsire's unique combination of resistances alongside STAB Earthquake and the looming threat of Belly Drum give it viability despite its low BST. While it hasn't seen substantial usage in SPL, it gained recognition after Quagsire teams were brought successfully in several recent high level matches. McMeghan ranked it 24 and several others ranked it at 25.
-
Moltres (+3) has similarly seen successful use in high level matches, including in SPL. It blazes a niche through its ridiculous power under Sunny Day, despite lacking options to get past Water-types that resist Fire. Metagame shifts away from traditional threats like Vaporeon and defensive behemoths like Suicune and replaced them with Pokemon that Moltres has no problem dealing over 70% to in a single hit. Eeveeto rated it as the 26th most viable Pokemon with ABR and sulcata close behind at 27.
-
Smeargle (-4) has suffered since the Sleep Trap ban in 2018 and continues on a downward trend. While it still has a firm spot on Agility Pass teams, the unreliability of said teams and Smeargle's reliance on the opponent getting caught out by its other options has left it in a difficult spot. While Jorgen and Century Express still believe it has more potential than most of the other low rank Pokemon, others such as Bomber. and myself rated it 39 and 41.

I'm very interested to hear others opinions on the updated rankings, so please feel free to post your thoughts.
 

vapicuno

你的价值比自己想象中的所有还要低。我却早已解脱,享受幸福
is a Site Content Manageris a Forum Moderatoris a Community Contributoris a Tiering Contributoris a Top Contributoris a Past WCoP Champion
IMPORTANT: Please use the Smogon classic theme instead of Smogon Dark to view this post. This provides a white background necessary to view the graphs in this post because of the png transparency.

Hi everyone,

I have worked with Earthworm on this year's VR update, which is going to be pretty comprehensive. This update will be based on an extension of my original methodology described here, taking inspiration from a previous post by Jorgen . Thanks to Bomber. Eeveeto d0nut ABR Hardboned Excal Lojh() Hyogafodex Diophantine We Three Kings sulcata Fear Jimmy Turtwig HSA Earthworm KratosMana McMeghan choolio THUNDERCEBRA Meriksn Blightbringer Mr.378 Zokuru Descending CMx Royal Flush FriendOfMrGolem120 asbdsp BKC Jorgen Century Express dice Djokra . And thank you Earthworm for providing this data to work with. I must profess that I am not a GSC player, so I'm going to leave it to you guys to interpret the results. On the bright side, that also means I carry no preconceived biases to the tiering.

Okay, TLDR stuff first:
The average outlier-compensated ranks from everyone are
01 Snorlax
02 Zapdos
03 Raikou
04 Cloyster
05 Gengar
06 Skarmory
07 Exeggutor
08 Tyranitar
09 Nidoking
10 Machamp
11 Steelix
12 Golem
13 Vaporeon
14 Forretress
15 Starmie
16 Suicune
17 Jynx
18 Marowak
19 Rhydon
20 Misdreavus
21 Umbreon
22 Heracross
23 Miltank
24 Blissey
25 Tentacruel
26 Espeon
27 Charizard
28 Dragonite
29 Jolteon
30 Porygon2
31 Houndoom
32 Quagsire
33 Moltres
34 Alakazam
35 Smeargle
36 Muk
37 Meganium
38 Clefable
39 Piloswine
40 Kangaskhan
41 Lanturn
42 Sandslash
43 Scizor
44 Donphan
45 Shuckle
46 Ampharos
47 Kingdra
48 Entei
49 Aerodactyl
50 Articuno
51 Lapras
52 Omastar
53 Venusaur
54 Gligar
55 Jumpluff
56 Pikachu
57 Tauros
58 Nidoqueen
59 Blastoise
60 Qwilfish
61 Typhlosion
62 Ursaring
63 Hitmonlee
64 Slowbro
65 MrMime
66 Electabuzz
67 Magmar
68 Raichu
69 Arcanine
70 Poliwrath
71 Dodrio
72 Victreebel
73 Electrode
74 Hypno
75 Chansey
76 Primeape
77 Scyther
78 Crobat
79 Magneton
80 Gyarados
81 Bellossom
82 Politoed
83 Feraligatr
84 Kabutops
85 Girafarig
86 Vileplume
87 Slowking
88 Pinsir
89 Granbull
and considering only pokemon that were ranked by >50% of players, and adding Pokemon that 5 or more people believe should be ranked, we get the reduced list that Earthworm has worked with in his graphic,
01 Snorlax
02 Zapdos
03 Raikou
04 Cloyster
05 Gengar
06 Skarmory
07 Exeggutor
08 Tyranitar
09 Nidoking
10 Machamp
11 Steelix
12 Golem
13 Vaporeon
14 Forretress
15 Starmie
16 Suicune
17 Jynx
18 Marowak
19 Rhydon
20 Misdreavus
21 Umbreon
22 Heracross
23 Miltank
24 Blissey
25 Tentacruel
26 Espeon
27 Charizard
28 Dragonite
29 Jolteon
30 Porygon2
31 Houndoom
32 Quagsire
33 Moltres
34 Alakazam
35 Smeargle
36 Muk
37 Meganium
38 Clefable
39 Piloswine
40 Kangaskhan
41 Sandslash
42 Scizor
43 Donphan
44 Shuckle
45 Ampharos
46 Kingdra
47 Entei
48 Aerodactyl
49 Articuno
50 Omastar
51 Venusaur
52 Gligar
53 Jumpluff
54 Pikachu
55 Tauros
56 Nidoqueen
57 Qwilfish
58 Typhlosion
59 Ursaring
60 Hipmonlee
There is essentially no difference in these lists for the purposes of this VR, as we are analyzing just the top few tiers.

and the aggregate VR tiers obtained are
S1: :Snorlax:
S2: :Zapdos:
A: :Raikou::Cloyster:
B1: :Gengar::Skarmory::Exeggutor::Tyranitar:
B2: :Nidoking::Machamp::Steelix::Golem::Vaporeon::Forretress:
B3: :Starmie::Suicune::jynx::Marowak:
B4: :Rhydon::Misdreavus::Umbreon::Heracross:
C1: :Miltank::Blissey::Tentacruel::Espeon:
C2: :Charizard::Dragonite::Jolteon::Porygon2::Houndoom::Quagsire::Moltres::Alakazam::Smeargle:
D: :Muk::Meganium::Clefable::Piloswine:
E1: :Kangaskhan::Sandslash::Scizor::Donphan::Shuckle::Ampharos::Kingdra::Entei::Aerodactyl::Articuno::Omastar::Venusaur:
E2: :Gligar::Jumpluff::Pikachu::Tauros::Nidoqueen::Qwilfish::Typhlosion::Ursaring::Hitmonlee:

Let's go through the whole process.

First the data is cleaned by compensating outliers 1 standard deviation away from the edge of the percentiles expected to contain +/- 1 standard deviation of a normal distribution. This is a modification of the conventional interquartile range (IQR), which I have not chosen to use because 50% of the sample doesn't capture the full variation from what I've seen. The compensation is done by bringing these points to the edge of this extended range. This results in mostly zero, but sometimes one or two outlier corrections. We then plot the outlier-removed data as a function of the integer rank to obtain this graph.

VR Tiering Decisions
1588502000050.png
and we can zoom in to the top 26 mons where tiering decisions are expected to make more sense,
1588502012207.png


Up to Tyranitar the tiers are pretty well defined, but tiering seems really difficult after that; there are lots of overlaps between tiers and looking for jumps in the mean ranking is possible but not easy, We turn to hierarchical clustering to help obtain the tiers. We form a dissimilarity matrix where the distances between Pokemon X and Y are given by the following: Take the rate at which voters ranked Pokemon X over Pokemon Y, take the logit transform as is done in logistic regression of a Bernoulli-distributed variable, and take the absolute value. Performing what we call a Ward linkage, this yields a dendrogram of the following sort, where the clusters (what we are going to call tiers) formed by setting a reasonable threshold are represented by different colors, and the dissimilarity between each cluster can be thought of as the vertical height of the nearest branch that connects the two clusters.

In other words, the Tentacruel-Blissey tier is a lot closer to the Charizard-Moltres tier (connecting height 20, note the log scale) than the Gengar-Exeggutor cluster (connecting height >40). Note that the order is not preserved by the algorithm, and hierarchical clustering throws away information, so this is just a rough guide to defining the tiers.

1588502027976.png

We want to verify the validity of the clusters obtained from the dendrogram, so we next plot the dissimilarity matrix and draw out the tiers specified. I choose to split Snorlax and Zapdos up into two distinct tiers; I suspect that has to do with the scaling of the Ward linkage at small cluster sizes.

To read the dissimilarity matrix, note that zero (the darkest value) corresponds to equal number of people voting in favor and against the Pokemon on the Y axis > X axis, and the higher the value, the more one-sided the voting becomes. In other words, the darker, the more indistinguishable the Pokemon on the X and Y axis become, and a well-defined tier would be a fully dark square (read my methodology thread for explanations).

1588502042820.png


Note that although the Gengar-Tyranitar tier was closer to Raikou/Cloyster than Nidoking-Forretress in the dendrogram, I choose to trust my eyes on this and group it in the B tier due to the substantial overlap with the latter. This yields the following subdivision which Earthworm and I have decided on

S1: :Snorlax:
S2: :Zapdos:
A: :Raikou::Cloyster:
B1: :Gengar::Skarmory::Exeggutor::Tyranitar:
B2: :Nidoking::Machamp::Steelix::Golem::Vaporeon::Forretress:
B3: :Starmie::Suicune::jynx::Marowak:
B4: :Rhydon::Misdreavus::Umbreon::Heracross:
C1: :Miltank::Blissey::Tentacruel::Espeon:
C2: :Charizard::Dragonite::Jolteon::Porygon2::Houndoom::Quagsire::Moltres::Alakazam::Smeargle:
D: :Muk::Meganium::Clefable::Piloswine:
E1: :Kangaskhan::Sandslash::Scizor::Donphan::Shuckle::Ampharos::Kingdra::Entei::Aerodactyl::Articuno::Omastar::Venusaur:
E2: :Gligar::Jumpluff::Pikachu::Tauros::Nidoqueen::Qwilfish::Typhlosion::Ursaring::Hitmonlee:

Numerical ranks represent partial tiers, whereas letter ranks represent a more complete separation. I choose to adopt numerical subranks because there is no reason a priori to believe that Pokemon are grouped in viability by a tripartite scheme of +/-.

Metagame Shifts

This chart shows the difference between this and last year's VRs, together with the uncertainties in the means (not the standard deviation, but the standard deviations divided by sqrt(N-1)).
1588502059675.png
A better way to understand how significant these changes are so as not to mistake changes occuring as due to pure chance is to plot the z-score,

1588502069064.png

where the Y axis means number of standard deviations away from zero. To recap, 0.5, 1 and 2 standard deviations are about 69%, 84%, and 98% significant (one-sided), meaning roughly that for a z-score of 1, we expect that this change to have occurred due to chance 100%-84% = 16% of the time. Therefore, trust the data on the left than on the right.

Analysis of Camps

The VR tiering results are quite intriguing. Why are tier separations more ambiguous than last year's just looking at the mean and deviation in ranks, yet still quite clear in the dissimilarity matrix? I suspect, especially looking at the large deviations of Steelix and Golem, that there are some very strongly divided opinions that cause distributions to become bimodal. There are hints of this behavior in the heatmap, where we would expect Gaussian-distributed variables to form a smooth hourglass shape around the diagonal line, but we are getting oddities in interactions like Exeggutor-Golem, Steelix-Rhydon, and Umbreon-Starmie/Suicune. Here are some lines of thought I pursued:

Cloyster-Raikou Split
In this analysis, I simply clustered voters based on their rankings of the first four Pokemon, which in effect boils down to Cloyster and Raikou because almost everyone is in agreement with Snorlax and Zapdos. Then, I extrapolate to find patterns in the rest of the OU Pokemon.
1588502116514.png
from the Dendrogram, we identify a camp from Djokra-d0nut and from FOMG-ABR (I'm ignoring Lojh()-Jorgen, that group is too small). We then plot their mean rankings of each Pokemon for the two camps,
1588502127082.png
Note again the error bars are uncertainties in means, not the actual deviations (ie divided by sqrt(N-1)). Sorting the data to see the most significant changes, and using the z-score for that, we obtain
1588502136334.png


The following is a dissimilarity matrix for Pokemon now, divided by the two camps. To read it, note that when you see a strongly red square, then the camp being analyzed frequently ranks the corresponding Pokemon on the Y axis more favorably than that on the X axis. Another way to see this is by looking across the diagonal line. A strongly red square should be accompanied by a strongly blue square reflected across the line, and we can say the camp prefers the Pokemon on the Y axis of the red square more than that of the blue square. For example, in the data below, the Jorgen camp prefers Cloyster to Raikou.
1588502149025.png
Some things you might have missed out from looking simply at the z-score chart above are the non-adjacent preferences; Jorgen's camp prefers Starmie to Marowak, Umbreon to Marowak, Miltank to Heracross, and Porygon2/Houndoom to Charizard/Dragonite.

Tyranitar-Exeggutor Split; Skarmory-Gengar Split
If we now extend the analysis to include B1, we find that there are three camps,
1588502160033.png
I'm going to ignore the CMx-hyogafodex camp for being too small. The camps can be summarized in the relative ranking chart,
1588502170628.png
Notably, the Jorgen and Lojh() camps are split among Exeggutor and Tyranitar, while the Royal Flush camp differs from the other two in placing Skarmory above Gengar. Let us look at these three camps more closely with the z-scores,
1588502177530.png

1588502186861.png

1588502194927.png


And the corresponding dissimilarity matrices for Pokemon now, divided by the three camps
1588502204302.png
where we observe higher tier preferences of Gengar and Tyranitar over Skarmory and Exeggutor, and lower-tier preferences in favor of Golem over Exeggutor, Rhydon over Marowak, Dragonite over Porygon2, and Charizard over Smeargle.
1588502214001.png
where we observe high-tier preferences of Cloyster over Raikou, Gengar and Exeggutor over Skarmory over Tyranitar, and lower-tier preferences in favor of Forretress and Suicune over Golem, Blissey over Heracross, Jolteon and Porygon2 over Quagsire and Moltres.
1588502221563.png
where we observe higher tier preferences of Raikou over Cloyster, Skarmory and Exeggutor over Gengar and Tyaranitar, and lower tier preferences in favor of Steelix over Nidoking, Misdreavus over Golem, Starmie and Rhydon, and Umbreon over Starmie too, and Miltank over Heracross.

B-C Tier Camps
We now move to camps in the B-C tier. Notably, there are two evenly-split camps,
1588502247045.png
with the following relative ranks
1588502260675.png
and the following sorted z-scores,
1588502270419.png

and the corresponding dissimilarity matrix,
1588502278830.png
Notable preferences are a very strong preference for Golem over Steelix, and also Gengar > Skarmory, Nidoking > Machamp, a strong affinity for Rhydon and Forretress, and a preference for Blissey and Espeon > Miltank.

Camps across all tiers
Finally, we answer the question: are there overarching camps across all the tiers that can be considered mostly OU?
1588502295503.png

1588502302531.png
The dendrograms are strikingly similar to the results from splitting the B-C camp, indicating that most of the variance can indeed be captured by how people view the lower-tier Pokemon. I will simply show the data but not explain it due to the similarities with the B-C division.
1588502313832.png

1588502320915.png

1588502330366.png
Instead, I want to focus on the four-way split provided by the second dendrogram. This split is demonstrated in the relative rankings,
1588502340821.png
Noting that the sulcata and FOMG camps were grouped under one camp, and choolio and Bomber were grouped under the other camp, we note the following relative differences:
Between sulcata and FOMG camps: Suicune, Jynx, Starmie, Nidoking, Houndoom, Espeon are favored; Steelix, Tyranitar, Golem, Miltank, Heracross are disfavored.
Between choolio and Bomber camps: Machamp, Vaporeon, Marowak, Misdreavus, Umbreon, Miltank, Exeggutor are favored; Cloyster, Gengar, Nidoking, Golem, Tyranitar, Jynx, Tentacruel, Espeon are disfavored.

with corresponding differences wrt the rest of the field:
1588502349612.png

1588502356489.png

1588502363864.png

1588502370265.png

and here are the dissimilarity matrices:
1588502379328.png
This is really interesting to tease; there are a couple of major shifts. The sulcata camp prefers Skarmory and Exeggutor to Tyranitar, is really low on Steelix, rates Forretress, Starmie, Suicune and Jynx higher than even traditionally top 10 Pokemon. Notable preference for Umbreon over Rhydon, Blissey and Espeon over Miltank.
1588502393787.png
the FOMG camp has a pretty strong preference for Gengar over Exeggutor and Tyranitar over Skarmory; the camp likes Golem and Rhydon; the camp also prefers Heracross over Umbreon, Jolteon over Charizard, Jolteon over Houndoom, and Moltres over Smeargle.
1588502401179.png
The choolio camp strongly dislikes Golem and Nidoking, and favors Machamp, Steelix, Marowak, Misdreavus and Umbreon. Notable preferences are Skarmory and Exeggutor over Gengar, Marowak and Umbreon over Nidoking, Dragonite and Jolteon over Tentacruel and Espeon; in fact, there is a huge preference for Dragonite and Jolteon in this camp.
1588502409550.png
Probably the camp most consistent with the average rankings, the Bomber camp however really likes Steelix and dislikes Golem. There are some mild preferences for Vaporeon, and notably Tentacruel is preferred to Blissey, and Miltank over Espeon.

Individual Analyses

For those who are interested to see whose S to C rankings are closest to theirs, you can refer to the chart below. The numbers inside the box go from -100% (full anticorrelation) to 100% (full correlation). They are sorted by the S to C dendrogram order (and the light squares represent the camps).
1588502420967.png


And finally, these are the relative ranks of everyone. Blue = disfavor, Red = favor. Cyan lines demarcate tier cutoffs.

1588502448051.png


Closing Remarks

The large number of graphs may seem daunting, and to people who aren't quantitatively trained, this may be really confusing. I recommend just glancing over the spoilers on the first read, only thoroughly analyzing them after you've gone through the more important graphs that have been left unhidden. I'm interested to know what you can infer from these trends even though I don't play the metagame, and I hope this can generate some discussion.

Finally, I have attached the Jupyter notebook for this in a zip file. You'll need python and the associated packages to open and run it, and the easiest way to do that is to install the anaconda library here.
 

Attachments

the Bomber camp however really likes Steelix and dislikes Golem.
I can explain.

premise: i dont like rspin golem. i would put rspin in but hey we have just four slots to fill: eq, explosion, fireblast/flamethrower, roar, curse, rockslide, bodyslam, counter..

that said, i can compare Golem and Steelix considering a similiar role. Both are excellent baitkiller for bulky water and grass (such as Suicune, Starmie and Exeggutor, and therefore a great support for those who benefit from their elimination). just steelix can do this job better considering it can eat surf from suicune (and ohko at +1, same for Cloyster i didnt mention above), starmie (ohko guaranteed no curse needed), vaporeon (ohko after spikes) and even exeggutor is ohkoed after spikes damage. in addition steelix has toxic immunity wich is a huge boon when you want to shuffle and has better matchup vs electrics.

Golem has enormous offensive power, slightly better SDef, resistance to Fire, access to Flamethrower to keep away Skarmory from its boom targets (wait..and here is my biggest concern with the "standard" spin set), higher speed than Snorlax, Forretress (fb ohkoed), Rhydon and Steelix itself - Steelix and Golem 3HKO each other with STAB Earthquake, but Golem wins because he is faster (if steelix curses guess what? fb could be useful again). Golem at +2 ohkoed skarm with explosion (Steelix +3), bs has 30% to paralyze the targets and threaten them with boom. counter, well do you know how it works.

but it spins so ofc let it spin. it's not what stands out in my eyes looking at its movepool, but its ok, i guess it does its job more or less.
 
Last edited:

Mr.E

unban me from Discord
is a Two-Time Past SPL Champion
tfw too lazy to participate this time and possibly no longer considered in the long term :blobsad:

Charizard still about a billion ranks too high, too little love for the UU Grassers.

Nidoking, along with its fellow common sleeper Jynx, have massively influenced the metagame over the past couple of years and have spearheaded the rise of Thief offense in modern GSC.
Ironically, non-LK Nidoking usage is higher than ever. People seem to be valuing its sheer versatility more lately than relying on it as their team's Sleep user.

premise: i dont like rspin golem. i would put rspin in but hey we have just four slots to fill: eq, explosion, fireblast/flamethrower, roar, curse, rockslide, bodyslam, counter..
I used to use Golem back in the day as a generic offensive threat. Explosion is much better in GSC than it is in RBY, of course. In that sense, I think Golem is pretty competitive with Rhydon/Wak/Champ if people chose to use it that way instead of as a Spinner, which notably locks it out of Rock Slide (RBY tradeback) and makes it much more comparable to Steelix as a defensive pseudo-Electric counter that is completely helpless against Zapdos. Compared to Steelix, Golem still has the advantage of noticeably higher base damage and it isn't vulnerable to both of Snorlax's common coverage moves (though it gets reamed by the odd Surf, plus HP Water Raikou is becoming much more common in response), at the cost of basically just Toxic immunity. I do like Golem, but I think people have gone a little crazy with it.

Also, I'd kinda like to see some mono-Iron Tail Steelix get some action.
 
Last edited:
...In that sense, I think Golem is pretty competitive with Rhydon/Wak/Champ if people chose to use it that way instead of as a Spinner, which notably locks it out of Rock Slide (RBY tradeback) and makes it much more comparable to Steelix as a defensive pseudo-Electric counter that is completely helpless against Zapdos.
Golem does get HP Rock which can be used with Rapid Spin without issue, though it does come with a drawback:
Golem HP Rock vs. Zapdos (HP Ice DVs): 180-212 (46.9 - 55.3%) -- 16.2% chance to 2HKO after Leftovers recovery
Golem Rock Slide vs. Zapdos (HP Ice DVs): 195-230 (50.9 - 60%) -- 89.8% chance to 2HKO after Leftovers recovery


Plus the lack of flinching. Though if Zapdos has a non-Ice Hidden Power, HP Rock's 2HKO odds increase considerably, and every time Rock Slide misses you might regret not having HP Rock's 100% accuracy
 

Mr.E

unban me from Discord
is a Two-Time Past SPL Champion
If Zapdos has a non-Ice Hidden Power, Golem is potentially getting OHKOed by it anyway so... :blobshrug:
 

wyc2333

A=X+Y+Z Y: Hard Work
quick thoughts:
1588901919364.png
Skarmory (0)
If perfect is what you're searching for Then just stay the same
although resttalk skarm with curse and dpeck can be forced out by phazers, it is a useful set
https://replay.pokemonshowdown.com/smogtours-gen2ou-480259
tox skarm is rly a thing atm. hp fire may worth mentioning in oo bc it pairs well with tox
https://replay.pokemonshowdown.com/smogtours-gen2ou-472039
https://replay.pokemonshowdown.com/smogtours-gen2ou-477912
 

Jorgen

World's Strongest Fairy
is a Forum Moderator Alumnusis a Community Contributor Alumnusis a Contributor Alumnusis a Past SPL Champion
Meganium is just a tanky defensive thing that doesn't die. Apart from Jynx, Heracross, and the odd Fire-type, it's capable of holding the line against pretty much any major OU threat. Especially non-Ice Beam Vaporeon (which is most of them).

Until it runs out of Synthesis PP, anyway. Leech Seed helps with that, though. Although being especially vulnerable to status because it doesn't use Rest or have a status-immune typing really *hurts* with that.

I would say the standard Meganium set is something like Razor Leaf / Reflect / Synthesis / Leech Seed. It generally doesn't accomplish much apart from holding the line and trying to waste PP. It's similar to Charm Umbreon or Curse Skarmory in that regard, except Meganium is a lot shakier against last-Poke Curselax, which is frankly the main thing you want out of such a "hold the line" Pokemon. Growl might therefore be preferable to Reflect or Razor Leaf, in all honesty. However, Razor Leaf lets it pretend it's Suicune by being able to chase things away with a Super-Effective STAB move, and Reflect shores up the shortcomings of that approach. That may not be a *good* reason for the standard, but it's the reasoning that makes the most sense to me.

Meganium's options for extracting value out of the midgame are all pretty mediocre. The standard set might pair well with Spikes to force switches with Leech Seed? STAB Razor Leaf is actually a nice way to deal with all the major Rapid Spinners, barring Forretress, although nothing really stops Zapdos or Toxic Skarmory from coming in and trying to status you, which kind of negates Spikes. Maybe Meganium offers unreliable paralysis support with Body Slam over Razor Leaf? Maybe it just slaps Toxic on that fourth moveslot and calls it a day? You could also try dual screens, or do something completely different and attempt to actually do damage with a Swords Dance set, albeit without STAB. Or you try to catch something with Counter, that could be good. Maybe you pair it with a Swords Dance set and kill a Skarmory that absent-mindedly Drill Pecks you? Or even better, you bait Snorlax into trying to land a crit Double-Edge in trying to break a standard wall set.

Overall, the current metagame assigns a lot of value to being able to extract an advantage out of the midgame, and Meganium does not really do that. It's a lot shakier than Umbreon and Skarmory when it comes to PP stalling Snorlax, so it doesn't do its "main" job as well as you'd hope, either. It's similar to Miltank, except Miltank's main job was always Heal Bell, and its ability to check Snorlax with Growl was never good per se but rather just a way to check two boxes at once. So Meganium is pretty mediocre overall if you ask me, although I definitely recognize it has some very appealing traits, and some relatively unexplored options outside of its standard bread-and-butter set!
 

Mr.E

unban me from Discord
is a Two-Time Past SPL Champion
A recently successful Meganium team really put the spotlight on Synthesis Eggy for me operating in a similar capacity, actually. Eggy is physically tankier (almost but no dice), Stun Spore is much more reliable than Body Slam, and of course it's actually a legitimate offensive threat too. Stun-Leech is one of the best paralysis spreaders in the game.

As for Meganium itself, I think Jorgen largely covered it: it's just a brick. But it's a brick with slightly different utility than its peers and it's one of the few that doesn't allow Cloyster to just walk in for free on it. SD can be threatening too because Meganium is so self-sufficient but status ruins it and of course Skarm hard walls it, if the opponent has it.
 
Last edited:

Jorgen

World's Strongest Fairy
is a Forum Moderator Alumnusis a Community Contributor Alumnusis a Contributor Alumnusis a Past SPL Champion
Snorlax Double-Edge vs. Exeggutor: 153-180 (38.9 - 45.8%) -- guaranteed 3HKO after Leftovers recovery

Snorlax Double-Edge vs. Meganium: 138-163 (38 - 44.9%) -- guaranteed 3HKO after Leftovers recovery

Meganium is actually slightly tankier on the physical side. But the point about Exeggutor having much better offense and support still stands! I assume you used a set like Psychic / Seed / Stun Spore / Synthesis? With Synthesis replacing Substitute on the SubSeed set? I never could get into "slow" Egg sets like that personally, although it always seemed like it should work.
 
I'm back once again to present the results of this year's GSC OU viability ranking update. Thank you to the nineteen other players who took part this year.

Collection of data
- Last year's VR was used as the basis for what Pokemon to include as options in the ranking questions.
- The order of the Pokemon was randomised for each submitter, however I grouped the top 35 Pokemon together and the bottom 25 together to save submitters a bit of time.
- Pokemon could be added to or removed from the list as the submitter wanted.
- None of the Pokemon were mandatory to rank.
- The link to the survey was posted in GSC Discord servers this year, but not publicly on the forums.
- It was possible for players to make edits to their rankings after submitting the survey.
- The survey was left open for about two weeks.

Formation of rankings
- vapicuno determined the cutoffs for the ranks and subranks.
- The order was determined by an outlier-compensated mean ranking.
- Pokemon had to be ranked by at least five players to be included. The end result of this was that the same set of Pokemon remained on the VR.
- See vapicuno's post below for more detailed information.

Flaws
- The only real flaw this year in my opinion was that there was bias towards the Pokemon already on the VR. As a result, nothing was added or removed. I'll perhaps consider doing a pre-ranking survey again next year, if I am around to do this again.

The link to the data in spreadsheet form is here.

S1 Rank
01
Snorlax​
S2 Rank
02
Zapdos​

A Rank
03
Cloyster​
04
Raikou​
05
Gengar​

B1 Rank
06
Tyranitar​
07
Skarmory​
08
Nidoking​
09
Golem​
10
Exeggutor​
B2 Rank
11
Starmie​
12
Machamp​
13
Forretress​
14
Vaporeon​
15
Jynx​
16
Steelix​
B3 Rank
17
Marowak​
18
Misdreavus​
19
Suicune​
20
Umbreon​
B4 Rank
21
Rhydon​
22
Heracross​
23
Miltank​
24
Blissey​
25
Tentacruel​

C1 Rank
26
Espeon​
27
Alakazam​
28
Jolteon​
C2 Rank
29
Charizard​
30
Moltres​
31
Smeargle​
32
Dragonite​
33
Quagsire​
34
Houndoom​
35
Porygon2​

D Rank
36
Muk​
37
Meganium​
38
Clefable​
39
Piloswine​

E1 Rank
40
Scizor​
41
Kangaskhan​
42
Articuno​
43
Shuckle​
44
Aerodactyl​
45
Donphan​
46
Entei​
47
Venusaur​
48
Kingdra​
49
Sandslash​
50
Ampharos​
51
Omastar​
52
Tauros​
E2 Rank
53
Qwilfish​
54
Ursaring​
55
Typhlosion​
56
Pikachu​
57
Jumpluff​
58
Nidoqueen​
59
Gligar​
60
Hitmonlee​

Visualisation of results and individual rankings
As with last year, I have put all the rankings together into a single image. In the New VR column, I have formatted it so that green indicates a rise and red indicates a fall when compared to the old VR. Blue indicates no movement. The pale green and red in everyone's rankings indicate what they ranked higher (green) and lower (red) than the new VR. I used a different method to produce it this year, so it may require some serious bandwidth--sorry about that!



Observations
Differences when compared to the old VR that I found particularly notable include:
-
Cloyster (+1) finally overtook Raikou for #3, which was a long time coming in my opinion and is a sign that offense is taking the front seat in the metagame. Literally everyone ranked Cloyster between 3 and 5, and only choolio and false ranked it #5--would be interested to hear why.
-
Tyranitar (+2) has jumped up above both Skarmory and Exeggutor into 6th place, which I'm fairly sure is its highest ever ranking. It's a respectable threat on its own and is probably the most well-rounded Normal-resistant phazer on offensive teams as well as a decent choice for Firelax answer on defensive teams. Earthquake Tyranitar is very popular thanks to its ability to threaten Gengar more immediately and its solid coverage against common Pokemon like Raikou, Nidoking, Golem, and opposing Tyranitar. This all-around utility is probably what made this jump possible.
-
Nidoking (+1), while only going up one position in the rankings, was remarkably close to overtaking Skarmory this year. Yet another sign that offensive strategies are trending in the metagame. It achieved a mean ranking of 8.4 despite its highest being 7 (Zokuru, McMeghan, and false being among those that ranked it 7) and likely would have overtaken Skarmory if choolio had ranked it higher than 11.
-
Golem (+3) again jumped up the rankings and managed a mean rank of 9.75 to pierce the top 10. This year seemed to be a contest for who could rank Golem highest, with TonyFlygon ranking it as the 5th most viable Pokemon (over even Raikou), barely edging out TDK who ranked it 6th. It's clear that some of the most successful and influential players rate Golem among the best Pokemon in the metagame. Anyone who has been watching high-level tournament play would have seen the effectiveness of aggressive Spikes control, and it's yet to be seen if an opposing metagame trend will decrease that strategy's viability. However, Golem was also one of the most divisive Pokemon, with Zokuru ranking it 16th and choolio 20th. If TDK's SPL performance and Golem's (unmatched for its usage rate) 60%+ win rate don't convince you of Golem's viability, I don't know what else will.
-
Exeggutor (-3) has continued its fall, dropping from 7th to 10th this year, and down from 5th the previous year. Once again, this is contrary to its 60% win rate in this year's SPL. It is indisputable that Jynx's rise has had something to do with this, but Exeggutor's actual performance doesn't seem to reflect how players are evaluating it. That said, it has continued its trend of low usage in SPL games--this year it was used the same number of times as Jynx, and Jynx had an absurd 80% win rate. It fell despite choolio ranking Exeggutor 4th and both FriendOfMrGolem120 and ABR ranking it 6th, thanks to players such as Hyogafodex rating it 15th, cherryb0ng rating it 14th, and Diophantine and Oibaf rating it 13th. However, it currently looks as though it will remain in the top 10 barring a substantial rise from other Pokemon.
-
Starmie (+3) has finally shot up the rankings, with players utilising a wide variety of sets successfully and proving its strength. While last year's trend of Nightmare Starmie wasn't as prominent this year, it remains a solid choice for a team slot with access to high-utility moves such as Thunder Wave and Rapid Spin, and it is a very effective Substitute user. Even Confuse Ray Starmie has been seeing use on the ladder recently. sulcata and false ranked Starmie 7th and 9th respectively, whereas ABR, TC, and cherryb0ng were some of the biggest Starmie doubters, ranking it 15th and 16th.
-
Steelix (-6), in perhaps the most significant double-down from last year, dropped an astounding 6 positions on the rankings, despite a 66% win rate in SPL and a respectable 12 uses. Whereas Golem was the Pokemon people were competing to rank highest this year, Steelix seems to be the target of those competing to rank a Pokemon the lowest. While choolio ranked Steelix at 6th and Earthworm ranked it 11th, sulcata took no issue with ranking it 24th, and Jorgen ranked it 21st. In my view, Steelix is still a perfectly viable choice, being a high-utility defensive Pokemon capable of meaningful offense. Its glory days may be over, but I hope we at least see a return to the top 15 for Steelix in future years.
-
Alakazam (+7) somehow managed to climb the most spots this year, despite only being used twice in SPL (albeit with a 100% win rate). One of Fear's trademark Pokemon, Alakazam is a wickedly fast special attacker with just enough coverage and utility options to cause big headaches for weakened teams. I think this climb can be attributed to an overall increase in appreciation for the value of offensive Psychic-types alongside Gengar--Jynx and Gengar are also trending upwards, and Alakazam makes for a great partner for the latter (and even the former, if you're Zokuru). While it lacks the defensive utility of Pokemon like Nidoking, timely use of Alakazam can clearly open up a game or clean it up. TDK ranked the spoon-wielder at 20th, and ABR at 21.
-
Smeargle (+4) managed to reverse its fortunes this year, being one of the few cases to buck the trends of last year in a significant way. While it fell four places last year, it has regained all four of them this year to return to 31st, presumably off the back of players such as gorgie using it in SPL. While it was only used four times, it did at least win three of those games, and the even the occasional sight of Smeargle succeeding seems to be enough to pull it back from the depths and leave it solidly a C rank Pokemon this year when it looked destined for D last year.

Looking forward to hearing everyone's thoughts on the update!
 

vapicuno

你的价值比自己想象中的所有还要低。我却早已解脱,享受幸福
is a Site Content Manageris a Forum Moderatoris a Community Contributoris a Tiering Contributoris a Top Contributoris a Past WCoP Champion
IMPORTANT: Please use the Smogon classic theme instead of Smogon Dark to view this post (link to profile settings here for convenience). This provides a white background necessary to view the graphs in this post because of the png transparency.

Hi everyone,

I have worked with Earthworm on this year's VR update as per last year. Again, this update will be based on an extension of my original methodology described here, taking inspiration from a previous post by Jorgen . Thanks to Hyogafodex Oibaf Siatam Earthworm Mr.378 Jorgen TC d0nut choolio Diophantine vani TonyFlygon TDK McMeghan Zokuru sulcata cherryb0ng false ABR FriendOfMrGolem120 for your rankings. And thank you Earthworm for providing this data to work with. I must profess that I am not a GSC player, so I'm going to leave it to you guys to interpret the results. On the bright side, that also means I carry no preconceived biases to the tiering.

Okay, TLDR stuff first:
The average outlier-compensated ranks from everyone are
01 Snorlax
02 Zapdos
03 Cloyster
04 Raikou
05 Gengar
06 Tyranitar
07 Skarmory
08 Nidoking
09 Golem
10 Exeggutor
11 Starmie
12 Machamp
13 Forretress
14 Vaporeon
15 Jynx
16 Steelix
17 Marowak
18 Misdreavus
19 Suicune
20 Umbreon
21 Rhydon
22 Heracross
23 Miltank
24 Blissey
25 Tentacruel
26 Espeon
27 Alakazam
28 Jolteon
29 Charizard
30 Moltres
31 Smeargle
32 Dragonite
33 Quagsire
34 Houndoom
35 Porygon2
36 Muk
37 Meganium
38 Clefable
39 Piloswine
40 Electabuzz
41 Scizor
42 Kangaskhan
43 Articuno
44 Shuckle
45 Aerodactyl
46 Donphan
47 Entei
48 Venusaur
49 Kingdra
50 Sandslash
51 Ampharos
52 Omastar
53 Tauros
54 Qwilfish
55 Ursaring
56 Politoed
57 Chansey
58 Typhlosion
59 Pikachu
60 Jumpluff
61 Blastoise
62 Nidoqueen
63 Gligar
64 Hitmonlee
65 Lanturn
66 Magneton
67 Flareon
68 Raichu
69 Mr. Mime
70 Crobat
71 Electrode
72 Slowbro
73 Dugtrio
74 Girafarig
75 Lapras
76 Weezing
77 Golduck
78 Feraligatr
79 Victreebel
80 Hypno
81 Gyarados
82 Persian
83 Magcargo
84 Magmar
85 Bellossom
86 Poliwrath
87 Venomoth
88 Ninetales
89 Scyther
90 Arcanine
91 Primeape
92 Hitmontop
93 Granbull
94 Slowking
95 Hitmonchan
96 Sudowoodo
97 Vileplume
98 Kabutops
99 Dodrio
100 Dewgong
101 Corsola
102 Delibird
103 Sneasel
104 Kingler
105 Pinsir
106 Raticate
107 Tangela
108 Aipom
109 Farfetch'd
110 Lickitung
111 Ledian
112 Wigglytuff
113 Octillery
114 Azumarill
115 Arbok
116 Rapidash
117 Murkrow
118 Mantine
119 Dunsparce
120 Parasect
121 Ariados
122 Seaking
123 Togetic
124 Butterfree
125 Beedrill
126 Xatu
127 Stantler
128 Pidgeot
129 Ditto
130 Furret
131 Sunflora
132 Fearow
133 Noctowl
134 Yanma
135 Wobbuffet
and considering only pokemon that were ranked by 5 or more players, we get the reduced list that Earthworm has worked with in his graphic,
01 Snorlax
02 Zapdos
03 Cloyster
04 Raikou
05 Gengar
06 Tyranitar
07 Skarmory
08 Nidoking
09 Golem
10 Exeggutor
11 Starmie
12 Machamp
13 Forretress
14 Vaporeon
15 Jynx
16 Steelix
17 Marowak
18 Misdreavus
19 Suicune
20 Umbreon
21 Rhydon
22 Heracross
23 Miltank
24 Blissey
25 Tentacruel
26 Espeon
27 Alakazam
28 Jolteon
29 Charizard
30 Moltres
31 Smeargle
32 Dragonite
33 Quagsire
34 Houndoom
35 Porygon2
36 Muk
37 Meganium
38 Clefable
39 Piloswine
40 Scizor
41 Kangaskhan
42 Articuno
43 Shuckle
44 Aerodactyl
45 Donphan
46 Entei
47 Venusaur
48 Kingdra
49 Sandslash
50 Ampharos
51 Omastar
52 Tauros
53 Qwilfish
54 Ursaring
55 Typhlosion
56 Pikachu
57 Jumpluff
58 Nidoqueen
59 Gligar
60 Hitmonlee
There is essentially no difference in these lists for the purposes of this VR, as we are analyzing just the top few tiers.
The aggregate VR tiers obtained are
S1: :Snorlax:
S2: :Zapdos:
A: :Cloyster::Raikou::Gengar:
B1: :Tyranitar::Skarmory::Nidoking::Golem::Exeggutor:
B2: :Starmie::Machamp::Forretress::Vaporeon::Jynx::Steelix:
B3: :Marowak::Misdreavus::Suicune::Umbreon:
B4: :Rhydon::Heracross::Miltank::Blissey::Tentacruel:
C1: :Espeon::Alakazam::Jolteon::Charizard:
C2: :Moltres::Smeargle::Dragonite::Quagsire::Houndoom::Porygon2:
D: :Muk::Meganium::Clefable::Piloswine:
E1: :Scizor::Kangaskhan::Articuno::Shuckle::Aerodactyl::Donphan::Entei::Venusaur::Kingdra::Sandslash::Ampharos::Omastar::Tauros:
E2: :Qwilfish::Ursaring::Typhlosion::Pikachu::Jumpluff::Nidoqueen::Gligar::Hitmonlee:

Let's go through the whole process.

First the data is cleaned by compensating outliers 1 standard deviation away from the edge of the percentiles expected to contain +/- 1 standard deviation of a normal distribution. This is a modification of the conventional interquartile range (IQR), which I have not chosen to use because 50% of the sample doesn't capture the full variation from what I've seen. The compensation is done by bringing these points to the edge of this extended range. This results in mostly zero, but sometimes one or two outlier corrections. We then plot the outlier-removed data as a function of the integer rank to obtain this graph.

VR Tiering Decisions
2021_GSC_OU_VR_S_to_C_Ranking_Linear.png
and we can zoom in to the top 35 mons where tiering decisions are expected to make more sense,
2021_GSC_OU_VR_Top_35.png


Up to Gengar the tiers are pretty well defined, but after that there are lots of overlaps between tiers and looking for jumps in the mean ranking is possible but not obvious to the eye. We turn to hierarchical clustering to help obtain the tiers. We form a dissimilarity matrix where the distances between Pokemon X and Y are given by the following: Take the rate at which voters ranked Pokemon X over Pokemon Y, take the logit transform as is done in logistic regression of a Bernoulli-distributed variable, and take the absolute value. Performing what we call a Ward linkage, this yields a dendrogram of the following sort, where the clusters (what we are going to call tiers) formed by setting a reasonable threshold are represented by different colors, and the dissimilarity between each cluster can be thought of as the vertical height of the nearest branch that connects the two clusters.

In other words, the Tentacruel-Blissey tier is a lot closer to the Marowak-Umbreon tier (connecting height 10, note the log scale) than the Espeon cluster (connecting height 30). Note that the order is not preserved by the algorithm, and hierarchical clustering throws away information, so this is just a rough guide to defining the tiers.

2021_GSC_OU_VR_Dendrogram.png

We want to verify the validity of the clusters obtained from the dendrogram, so we next plot the dissimilarity matrix and draw out the tiers specified. I choose to split Snorlax and Zapdos up into two distinct tiers; I suspect that has to do with the scaling of the Ward linkage at small cluster sizes.

To read the dissimilarity matrix, note that zero (the darkest value) corresponds to equal number of people voting in favor and against the Pokemon on the Y axis > X axis, and the higher the value, the more one-sided the voting becomes. In other words, the darker, the more indistinguishable the Pokemon on the X and Y axis become, and a well-defined tier would be a fully dark square (read my methodology thread for explanations).

2021_GSC_OU_VR_Dissimilarity.png


Note that although the Tyranitar-Steelix tiers were closer to Raikou/Cloyster than Marowak-Tentacruel in the dendrogram, I choose to trust my eyes on this and group them together in the B tier due to the substantial overlap with the latter. I have also chosen to place Smeargle in C2 and Charizard in C1, because the dendrogram makes it clear they belong there. I'm not sure why the algorithm decided against that, but I'll go with common sense.

This yields the following subdivision which Earthworm and I have decided on

S1: :Snorlax:
S2: :Zapdos:
A: :Cloyster::Raikou::Gengar:
B1: :Tyranitar::Skarmory::Nidoking::Golem::Exeggutor:
B2: :Starmie::Machamp::Forretress::Vaporeon::Jynx::Steelix:
B3: :Marowak::Misdreavus::Suicune::Umbreon:
B4: :Rhydon::Heracross::Miltank::Blissey::Tentacruel:
C1: :Espeon::Alakazam::Jolteon::Charizard:
C2: :Moltres::Smeargle::Dragonite::Quagsire::Houndoom::Porygon2:
D: :Muk::Meganium::Clefable::Piloswine:
E1: :Scizor::Kangaskhan::Articuno::Shuckle::Aerodactyl::Donphan::Entei::Venusaur::Kingdra::Sandslash::Ampharos::Omastar::Tauros:
E2: :Qwilfish::Ursaring::Typhlosion::Pikachu::Jumpluff::Nidoqueen::Gligar::Hitmonlee:

Numerical ranks represent partial tiers, whereas letter ranks represent a more complete separation. I choose to adopt numerical subranks because there is no reason a priori to believe that Pokemon are grouped in viability by a tripartite scheme of +/-.

Metagame Shifts

This chart shows the difference between this and last year's VRs, together with the uncertainties in the means (not the standard deviation, but the standard deviations divided by sqrt(N-1)).
2021_GSC_OU_VR_Rank_Improvement_Raw.png
A better way to understand how significant these changes are so as not to mistake changes occuring as due to pure chance is to plot the z-score,
2021_GSC_OU_VR_Rank_Improvement_z_Score.png

where the Y axis means number of standard deviations away from zero. To recap, 0.5, 1 and 2 standard deviations are about 69%, 84%, and 98% significant (one-sided), meaning roughly that for a z-score of 1, we expect that this change to have occurred due to chance 100%-84% = 16% of the time. Therefore, trust the data on the left than on the right.

Analysis of Camps

As per last year, I tried to look out for divided opinions. The most significant split I found was in opinions of Skarmory/Exeggutor in the S-to-B1 tiers; other than that, ABR, TonyFlygon and TDK have very different opinions from the rest in B2-to-C1.

S to B1: Skarmory-Exeggutor Split
In this analysis, I simply clustered voters based on their rankings of the first ten Pokemon. Then, I extrapolate to find patterns in the rest of the OU Pokemon.
2021_GSC_OU_VR_S1_to_B1_Dendrogram.png
from the Dendrogram, we identify a camp from vani-Diophantine and from Mr.378-FOMG. We then plot their mean rankings of each Pokemon for the two camps,
2021_GSC_OU_VR_S1_to_B1_Relative_Rank.png
Note again the error bars are uncertainties in means, not the actual deviations (ie divided by sqrt(N-1)). Sorting the data to see the most significant changes, and using the z-score for that, we obtain
2021_GSC_OU_VR_S1_to_B1_Relative_Rank_z_Score.png


I want to focus on the largest two differences - Skarmory and Exeggutor are clearly divided in opinion, but how so? We examine this through the following dissimilarity matrix divided across the two camps.

To read it, note that when you see a strongly red square, then the camp being analyzed frequently ranks the corresponding Pokemon on the Y axis more favorably than that on the X axis. Another way to see this is by looking across the diagonal line. A strongly red square should be accompanied by a strongly blue square reflected across the line, and we can say the camp prefers the Pokemon on the Y axis of the red square more than that of the blue square. For example, in the data below, the vani camp prefers Skarmory to Tyranitar.
2021_GSC_OU_VR_S1_to_B1_Dissimilarity.png

The dissimilarity matrix shows that the divide in Skarmory/Exeggutor is really extensive. The vani camp generally thinks of Skarmory as superior among the whole B1 tier and Exeggutor as the bottom of the pack. Not only is Exeggutor rated lowly, it is also frequently ranked by the vani camp below Starmie-Jynx (most of the B2 tier). On the flip side, the Mr.378 camp unanimously thinks Exeggutor is better than Skarmory.

B2-C1: ABR/TonyFlygon/TDK Split
In B2-C1, ABR, TonyFlygon, and TDK form a camp that mostly agrees internally but astoundingly differs from the rest.
2021_GSC_OU_VR_B2_to_C1_Dendrogram.png
2021_GSC_OU_VR_B2_to_C1_Relative_Rank.png
2021_GSC_OU_VR_B2_to_C1_Relative_Rank_z_Score.png

Note the huge difference in z scores, which means it's very likely differences in opinion are real and not just a result of natural variation.
2021_GSC_OU_VR_B2_to_C1_Dissimilarity.png
I'll leave it up to you guys to explain this.

Individual Analyses

For those who are interested to see whose S to C rankings are closest to theirs, you can refer to the chart below. The numbers inside the box go from -100% (full anticorrelation) to 100% (full correlation). They are sorted by the S to C dendrogram order (and the light squares represent the camps).
2021_GSC_OU_VR_S_to_C_Correlation.png

And finally, these are the relative ranks of everyone. Blue = disfavor, Red = favor. Cyan lines demarcate tier cutoffs.
2021_GSC_OU_VR_Individual_Rankings.png

Closing Remarks

The large number of graphs may seem daunting, and to people who aren't quantitatively trained, this may be really confusing. I recommend just glancing over the spoilers on the first read, only thoroughly analyzing them after you've gone through the more important graphs that have been left unhidden. I'm interested to know what you can infer from these trends even though I don't play the metagame, and I hope this can generate some discussion.

Finally, I have attached the Jupyter notebook for this in a zip file. You'll need python and the associated packages to open and run it, and the easiest way to do that is to install the anaconda library here.
 

Attachments

Jorgen

World's Strongest Fairy
is a Forum Moderator Alumnusis a Community Contributor Alumnusis a Contributor Alumnusis a Past SPL Champion
I actually ranked 134. You can see them in the aggregate rankings. The only one of those I didn't rank was Chansey.

ETA: I think the lower-rank cluster suggests a small group of serious Golem lovers. It makes the most sense to me that lower-rank movements would actually follow from preferences among the higher ranks, anyway. Loving Golem means not seeing much use for Rhydon (and in turn, not much use for a Pokemon like Suicune whose major selling point is how hard it stuffs Rhydon). Loving Golem also means wanting to find good ways to switch something into Gengar that aren't Tyranitar-based. This is consistent with their relative love of things like Alakazam, Umbreon, and Blissey, all of which either outspeed Gengar or stuff it. Of course, loving Golem may actually be a function of seeing Cloyster & Gengar as much better than the rest of the pack, which gets obscured since everyone loves those guys. But, still, this group also seems to value BP more highly than the rest; Jolteon and Smeargle are evidence of this. So there might be something deeper going on among this group. I might call this trio of ABR-aligned players "hypermodern", who value a fast-paced, spikes-centric game with the occasional curveball in an off-beat "anti-offense" Pokemon or a BP team. They are also really down on the slow, plodding, "heavy" wallbreakers that defined GSC offense in the past.

As for the Egg-Skarm divide, I think that might be more of a Skarm divide than anything else. If you think Skarm is really good, then it probably follows that you think Egg is really bad. You'd expect them to move together if Egg (the offensive Pokemon of the pair) were driving things. Ultimately, a lot less interesting imo, although it might be a good proxy for how divided people are on just how good Snorlax is (with Zokuru, the fella who put Zapdos above Lax, yet still got lumped into the Skarm-loving group, perhaps being exceptional in this regard).
 
Last edited:
I might call this trio of ABR-aligned players "hypermodern", who value a fast-paced, spikes-centric game with the occasional curveball in an off-beat "anti-offense" Pokemon or a BP team. They are also really down on the slow, plodding, "heavy" wallbreakers that defined GSC offense in the past.
Interesting. Who are the three players you're referring to and which graph have drawn this conclusion from?
 

Mr.E

unban me from Discord
is a Two-Time Past SPL Champion
Zokuru the kind of baseball guy to gatekeep 100% unanimity by voting against obvious hall-of-famers.

But he's also one of the two to rate Golem by far the lowest so he clearly knows what he's doing. :blobthumbsup:
 
Legit question, Mr.E - why don't you like Golem? I find it's one of the easiest Pokemon to slap on all types of teams and have success with. I didn't submit rankings but I definitely would've put it in the top ten. Conversely, I would've ranked Steelix - whose slot Golem competes for - quite low, as I find it to consistently be underwhelming on both offense and defense. I think it requires Gengar alongside it, or Forretress is just going to ruin you. Lix is pretty decent if it is paired with Gar, but that's restrictive as hell.
 

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

Top