Data Official Smogon University Usage Statistics Discussion Thread, mk.2

Status
Not open for further replies.
| 52 | Zapdos | 3.27145% | 140008 | 3.662% | 109669 | 3.774% |

Antar can we PLEASE PLEASE PLEASE PLEASE PLEASE use 1825 stats for OU because I like UU and Zapdos
 

bludz

a waffle is like a pancake with a syrup trap
is a Community Leaderis a Community Contributoris a Tiering Contributoris a Contributor to Smogon
Community Leader
First of all, there's a decent amount of theorymonning behind where the last four EVs go in a lot of cases (important for Life Orb / Leftovers numbers or for optimizing bulk), but I see what you mean. I, at one point, did work on combining similar spreads, but figuring out where to draw the line is annoying (for instance, should Hasty and Naive be combined?). The best solution is probably to change the entire way I report EV spreads and natures and do graphs that show how often each Pokemon runs:
  • Max EVs, Boosting nature
  • Max EVs, Neutral nature
  • Neutral nature, no EVs
  • Hindering nature, no EVs
and then report the two most common EV numbers (enforce, say, that the EVs cannot be within 8 points of each other..?)

This type of report would be most useful for opposing damage calcs (you see that your attack is a guaranteed KO for 252 Neutral, and that only 12% of that Pokemon run 252+, so you decide to go for it), but the flip side is that it wouldn't be useful at all for building teams (but honestly, if you're using moveset stats rather than analyses to build your teams, I think you have larger problems).

Thoughts?
Ok so I intended to bring up something like this but decided to read the thread and see if it had been addressed, lo and behold it has (twice, actually, yes I read the second one as well).

I think my issue is the separation of too many similar spreads. So for example a Landorus-T spread with 4 HP EVs moved into Def, or 8 SpD EVs into Spe. The real issue that arises is we have about 4-5 different spreads with 8-10% usage or so and then about 50% "other" which can't really be interpreted at all.

While I think the above suggestion is great I'm not sure how much time or effort it would require, and doesn't solve issues of mixed spreads that are similar. I have an idea for a proxy which could allow players to discern breakdowns in a more general way: Nature percentages.

Generally a pokemon can run many different spreads that are similar to achieve the same effect, but most of the time its nature will remain the same. There are exceptions like some bulky Scizors running Adamant and so on, but in general you can distinguish spreads between bulky (Impish/Bold/Calm/Careful/Relaxed/Sassy) and offensive (Timid/Jolly/Naive/Hasty/Modest/Adamant) simply by looking at the natures. The other thing is Natures won't get fragmented the same way as EV spreads and we can get a much clearer look at the full breakdown - I assume there could still be an "other" category but it would probably be much closer to something like 10% than 50%.

I also have another question. Since Megas are being tiered separately now, are you going to change the way you derive usage stats? For the above example, maybe I am interested in seeing how many Mega Scizors are bulky and how many are offensive, this still can't be done because Scizor still conglomerates the Mega and non-Mega form. It's also a bit annoying for Charizard - although this is a special case - because if I want to see how many Charizard Xs run Earthquake, I can't find that data.

I am still interested in total item breakdowns on pokemon like Scizor - how many run Scizorite, Band, LO, Lefties, etc, but the full breakdown seems less important because we're almost discussing 2 different pokemon (and the tiering process now reflects that).

PS: I think this is an excellent resource, so thanks for your time and effort

EDIT: well ok I see that Mega usage is done separately now so that answers my question. Still think Nature %s would be great!
 
Last edited:
So for the 1825 weighted stats, is 1825 the baseline?

Also is there way you could put a win percentage stat on there based on how many games that pokemon won and lost?
Heck, it'd also be an actual good use of the "Real" usage stat. And a good statistic to cite when discussing suspects/possible suspects

sorry for the double post, it's been a whole day and I doubt anyone would notice an edit to my post when it doesn't even alert anybody.
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
MagikaripIsOP , win percentage is a meaningless statistic, since we do matchmaking on our ladders. In a perfect world with a perfect matchmaking algorithm (and a perfect distribution of player skills), everyone's W-L ratio would be 50/50. Average weight and viability ceiling are MUCH better metrics for how "winning" a Pokemon is.

Folks, it'll be a few more days for the March stats. Total screw-up on my part--I accidentally reran FEBRUARY'S stats rather than running March, and I only just now noticed as I went to upload the stats. Sorry.
 
Last edited:
MagikaripIsOP , win percentage is a meaningless statistic, since we do matchmaking on our ladders. In a perfect world with a perfect matchmaking algorithm (and a perfect distribution of player skills), everyone's W-L ratio would be 50/50. Average weight and viability ceiling are MUCH better metrics for how "winning" a Pokemon is.
Everyone's win ratio would be 50/50, but the Pokemon won't be.

Clash Royale and various other games have skill based matchmaking but the developers still balance things (cards in Royale's case) based on it's win ratio. Clash Royale and Smogon have very similar ranking systems but that doesn't stop cards from being overpowered. It is similar to how Mega Kangaskhan or Soul Dew Lati@s should theoretically make the user have a higher chance to win games on the OU ladder.

Average weight isn't too accurate because otherwise Landorus-T would be S rank. I assume by viability ceiling you mean the 1825, 1500, etc. stats?
I am 99% sure that there are Pokemon with higher win rates than Landorus-T.

(Don't mean any disrespect in any of this obviously, I just really want win percentage as a statistic.)
 
Last edited:

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
Everyone's win ratio would be 50/50, but the Pokemon won't be.

Clash Royale and various other games have skill based matchmaking but the developers still balance things (cards in Royale's case) based on it's win ratio. Clash Royale and Smogon have very similar ranking systems but that doesn't stop cards from being overpowered. It is similar to how Mega Kangaskhan or Soul Dew Lati@s should theoretically make the user have a higher chance to win games on the OU ladder.
This kind of analysis, frankly, doesn't work, for the same reason average weight is a shitty metric. This is because, just because a Pokemon is powerful doesn't mean the average player uses it well. Let's say, for the sake of example, that Pikachu is the best, most overpowered Pokemon in the game. Then you'd have the top players using Pikachu to great effect and winning their matches. The problem is that since Pikachu is the anime mascot, you also have 100x as many players who have NO IDEA what they're doing who are using it and dragging the ratios / averages down.

What you're trying to determine is how having a Pokemon on a team helps that team win. I tackled this problem from the opposite perspective: how much is a given Pokemon holding a team back? That is, if you have a certain Pokemon on your team, what's the best you'll be able to do on a ladder? This measure is called a Pokemon's viability ceiling, and I encourage you to read more about that here: http://www.smogon.com/forums/thread...re-of-how-far-a-pokemon-can-take-you.3546373/
 
This kind of analysis, frankly, doesn't work, for the same reason average weight is a shitty metric. This is because, just because a Pokemon is powerful doesn't mean the average player uses it well. Let's say, for the sake of example, that Pikachu is the best, most overpowered Pokemon in the game. Then you'd have the top players using Pikachu to great effect and winning their matches. The problem is that since Pikachu is the anime mascot, you also have 100x as many players who have NO IDEA what they're doing who are using it and dragging the ratios / averages down.

What you're trying to determine is how having a Pokemon on a team helps that team win. I tackled this problem from the opposite perspective: how much is a given Pokemon holding a team back? That is, if you have a certain Pokemon on your team, what's the best you'll be able to do on a ladder? This measure is called a Pokemon's viability ceiling, and I encourage you to read more about that here: http://www.smogon.com/forums/thread...re-of-how-far-a-pokemon-can-take-you.3546373/
That specific example only works for Pikachu though. For most Pokemon the people that actually use the pokemon at least somewhat correctly would outweigh the clueless players.

Viability ceiling is a cool statistic though. Thanks for that and it should hold me over even if you never publish a win rate
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
That specific example only works for Pikachu though. For most Pokemon the people that actually use the pokemon at least somewhat correctly would outweigh the clueless players.
You'd be surprised. PS is *very* active, and only a small fraction of players (especially in tiers like OU) are what most of us would consider competitive.
 
http://www.smogon.com/stats/2016-03/ubers-1760.txt
| 9 | Rayquaza-Mega | 15.14961% | 65770 | 8.758% | 43256 | 8.536% |

i assume Rayquaza-Mega is any Rayquaza with Dragon Ascent? could this be changed to just Rayquaza(there's a separate entry for just "Rayquaza" which i assume is Rayquaza without DA)
| 71 | Rayquaza | 0.63578% | 19048 | 2.536% | 11925 | 2.353% |

im sorry if this it the incorrect place for this
 

dodmen

Ain't no tuition for havin' no ambition
is a Site Staff Alumnusis a Team Rater Alumnusis a Community Leader Alumnusis a Community Contributor Alumnusis a Tiering Contributor Alumnusis a Contributor Alumnus
I'm not sure how feasible this is or if it's the right place to ask, but for the moveset stats, would it be possible to have a separate section for just Nature?

Currently the "Spreads" section doesn't really give much information at all because of all the variations based on where the 4 extra EVs go, random Speed creep, etc. For example, Florges in uu-1760 looks like this:



If there was a separate section with something like "Calm: 50%, Bold: 40%, Modest: 10%", that would be a lot more helpful than what's currently shown. Is this possible/not too difficult to do?
 

Antar

is a Battle Server Administratoris a Programmeris a Super Moderatoris a Community Contributor
Official Data Miner
dodmen, I used to list Nature separately from EVs, but people seemed to find that LESS useful than the current configuration.

I've considered having a separate section for each stat, so it would look like:

Atk
252+: 10%
252: 6%
248: 1%
128: 4%
0: 58%
0-: 21%

Or something like that, but then you lose the correlation. I've also considered combining similar sets (e.g. 4/252/0/0/0/252 and 0/252/4/0/0/252) but there's a lot of strategy that goes into placing those last four EVs. The best I can recommend is that you look into the JSON files in the chaos folders, which contain the usage of EVERY spread and then maybe write your own analysis to combine them as you'd like. I don't know if you have any sort of programming background, but JSONs are super easy to work with, especially in languages like Python.
 
Sorry if this is a stupid question, but shouldn't the cutoff for dropping megas be lower than normal mons? You can only use one per team, and that's why their usage is so low compared to other pokemon.
 
Status
Not open for further replies.