• Announcements

    • Hisagi

      Rules for Ingame | Forums | IRC   04/29/2016

      Welcome to PokeMMO! This is the PokeMMO Code of Conduct, which states how we moderate the various services provided by us (Forums, In-Game, IRC). You are required to follow this Code of Conduct while using our Forums and playing the Game: 1. Rules related to "Chat" and communication with other users: a. You may not harass, threaten, embarrass or cause distress and/or unwanted attention to other persons using our Service (Players and PokeMMO's Staff). This includes posting insulting, offensive, or abusive comments about people, repeatedly sending unwanted messages, reporting players maliciously, attacking a player based on race, sexual orientation, religion, heritage, etc. b. You may not spam, flood, or repeatedly make duplicate posts. Messages composed of gibberish ("fjdklasjfld" "asdfsafdsa" "uiouoiuoiuoi") are considered spam. c. You may not impersonate any PokeMMO staff (volunteer or employed.) d. You may not solicit, advertise, or promote any services other than PokeMMO via PokeMMO's services (forums, in game, irc, etc). e. You may not upload, attempt to distribute files, or facilitate the distribution of files that contain viruses, corrupted data, or any other malicious software. f. You may not distribute or facilitate distribution of any pirated or illegal software while using PokeMMO's Services. g. You may not transmit, distribute, or facilitate distribution of any person's personal information (name, account name, telephone number, address, etc.) h. You may not transmit, post, link to, or facilitate distribution of any sexually explicit, harmful, threatening, abusive, defamatory, infringing, obscene, hateful, vulgar, racially or ethnically offensive imagery or content. i. You will follow the instructions of PokeMMO's Staff while using PokeMMO's Services.   2. Rules related to Gameplay: In addition to all rules defined by Section 1 of the Code of Conduct: a. You may not exploit errors in design, features which have not been documented, or "bugs" to gain access which is otherwise not available, or to gain an advantage over other players. b. You may not communicate any exploitable issues (defined in Section 2 Paragraph A of the Code of Conduct) either directly or through public posting, to any other users of PokeMMO. c. While participating in Player vs Player (PvP) gameplay, You may not participate in any form of match manipulation. Match manipulation is defined as any action taken to manipulate the outcome of a match, or alter or manipulate any official rankings. d. You may not use cheats, automation software (bots), macros, hacks or any third-party software which can be detrimental to other users' experience, nor will you relay or store any items for other users who are using these processes. e. You may not exploit the game or any of its parts for any commercial purpose, including without limitation: I. Gathering in-game currency, items or resources for sale outside the Game (commonly known as "Real Money Trading") II. Performing in-game services in exchange for payment outside the game. III. For use at a cyber cafe, computer gaming center or any other location-based site without the express written consent of PokeMMO.   3. Rules related to Usernames: When you choose a character name or a username or otherwise create a label which can be seen by other users when using our Services, you must abide by the following guidelines. You may not use any Username which: a. Belongs to another person with the intent to impersonate that person, including PokeMMO's Employees, Volunteers and any other user of PokeMMO's Services; b. Incorporates offensive racial, ethic, national, or cultural connotations; c. Is sexually suggestive or pornographic; d. References any form of criminal activity or drugs; e. Makes inappropriate references to the human anatomy or bodily functions; f. Uses misspellings or alternative spellings of any of the above guidelines.   These guidelines may not cover all inappropriate or disallowed names. PokeMMO reserves the right to reject any name it concludes, at its sole discretion, to be indecent, obscene, offensive, or otherwise violates the naming guidelines. 4. Miscellaneous Rules: In addition to all rules defined by Sections 1 and 2 of the Code of Conduct: a. You may not attempt to or facilitate any attempts to bypass any restrictions set in place against user accounts or characters. b. Exploits (as defined by Section 2, Paragraph A of this document) must be communicated to PokeMMO's Staff within a reasonable timeframe either via PokeMMO's Website or the e-mail address:[email protected] c. You may not use any unauthorized third-party software that intercepts or otherwise collects information from or through the Game or Service, including without limitation any software that reads areas of RAM used by the game, any program which reads and attempts to manipulate network traffic between your Game Client and the Game Server. d. You agree that You will not, under any circumstances violate any applicable law or regulation in connect with Your use of the Game or PokeMMO's Services. e. You may not disrupt or assist in the disruption of: I. Any computer used to support the Services (each "Server") II. Any other player's Game experience   5. Forum-specific policies: The rules in this section apply specifically to the forums at https://forums.pokemmo.eu/ a. You are required to use the English language when using the PokeMMO forums. Posts written in other languages are required to have an English translation appended to the post.     6. Violation Reports: If you find a player who is violating the Code of Conduct on these forums, please take the time to report it to us by using the Report button under their post, or the Report button in their profile. If you find a player in-game who you suspect is violating this code of conduct, please take the time to report it to us in the Player Reports section of this forum: https://forums.pokemmo.eu/index.php?/forum/28-player-reports/   7. Notes: Certain passages of the PokeMMO Code of Conduct have been simplified in this version. These simplifications include: The term "PokeMMO's Employees" and "PokeMMO's Volunteers" are encompassed by the term "PokeMMO's Staff"     You may find the full, legally binding version of this document at http://pokemmo.eu/code_of_conduct/ This document was last updated on June 30th, 2013.
Schwubbeldiwubb

A Statistical Approach on Team Building and Battling (WIP)

6 posts in this topic

Hello World,

like supposedly most people here, I've got accompanied by Pokemon my whole childhood through. Just a month ago I've started to actively play PokeMMO and at the same time begun to play competitively. And there we already have the first reason to warn you: Although I do know all Pokemon, their typings and specific tasks from Gen 1-3 by heart, I actually never played against human competitors until Jan 2017 due to lack of link partners.

On the other hand, I do know a bit about statistical analysis and other hardcore nerd stuff, so I guess that can help me to get better at figuring out what to do in PvP. Of course, throwing yourself into queue is by far the best way to improve, but I also like theory crafting. From the latter I have got a buttload incoming for you!

 

However, it should be noted, that statistics or maths in general are not representative for reality. Even Pokemon, although a game based on maths, is filled with such complexity and chaos, that it's very hard to predict the exact outcome of an encounter by numbers only. You either create a formula from a huge sample size and count in all imagineable variables, or handle the statistics with caution. I'd recommend you to do the latter with what I did. ^.^

 

This is based on information from the PokeMMO official usage statistics (as of February 8th 2017, but the data itself is from May 2016) and the Pokemon Database Pokedex. Google helped wording this ... I won't apologize for my English. :^)

I created the chart with OpenOffice Calc and deeply hope, that you are able to download and open the file I uploaded. More on this in a second ...

 

1. Motivations

 

  • First off, I wonder, if this kind of stuff is interesting at all or holds any value regarding competitive play...
  • Display various properties using a (hopefully) easy-to-read chart and by approach of statistical weights and averages figure out:
    • Best (and worst) attackers and move types
    • Best (and worst) defending Pkmn
    • Pkmn stats in comparison
  • By that:
    • Help creating a Team with good defense and offensive coverage
    • Grant easy access to several nice-to-know values, e.g. during battles

 

 

2. The chart of charts (for OU)

 

WIP-Note: At the moment I'm working on way more representative calculations counting in move sets and EV spreads (fow now, only a maximum of two spreads per Pkmn according to what I'd prefer), STABS and attack stats in addition to what I already considered. Problem is, this thing will be freaking huge and definitely not "easy to read" anymore. It could take some time, not only to pull this off, but to actually make it comprehendable...

 

Nevertheless I'd be glad if the following file, the "alpha version" so to speak, somehow finds its way to you, functioning, if not, please message me.

 

Download link to ODS version (recommended):

https://www.dropbox.com/s/4xuvcihrghm3hw5/OU battling statistics.ods?dl=0

 

Download link to XLS version (not tested with Excel yet):

https://www.dropbox.com/s/ea2zmet20ifuue1/OU battling statistics.xls?dl=0

 

This chart is ... well, a chart with lots of stuff in it. For descriptions of individual properties, you can mouse-over the column headers and read the tooltips.

The main feature is to be able to arrange values in ascending or descending order. That doesn't sound that cool, but it is, trust me. :^)

As of now, I only have the German version of Calc, so sadly I can't tell you what exact command there is necessary to pull this off with Excel. Either way you should be able to select the whole column you want to organize by clicking onto the header with the letter and execute the sort function. And don't forget to extend the selection when arranging the table, else the data gets messed up big time!

 

I should also note, that I only analyzed enough for it to be somewhat useful... I guess. The current chart is for the OverUsed tier only. A few OU Pkmn didn't even make it into the statistics this is based on, so please don't mind, if there is some (really) niche pick missing. Also, the weights I calculated for the arithmetic means could be waaay more refined, especially in the defense sector. Either way, in this state the approximations of weights are still quite lazy and I'm eager to count in a lot more than just play rate. But more on that later ...

 

In the following I will show you some of the results and try to explain briefly how I calculated that stuff. Suppose all stat values have 31 IVs by default, when there reads "minimum" I mean base stat + 31 IVs.

 

 

2.1. OU Offense

 

[WIP] 2.1.1. OU Attack Type Effectiveness

Spoiler

Refered to as "Offense Mean" in the "Offense Result" tab of the chart.

Roughly said, the number represents the effectiveness of the type in question against an "average typed OU Pkmn". But it is weighted in such a way, that the play rate of individual Pkmn is considered in the mean sum.

The weights are calculated by play rate of defender divided by average play rate within the OU tier, the latter being the sum of all play rates divided by the number of Pkmn. The weight for each Pkmn and the effectiveness of the type in question against it are then used in the arithmetic mean.

 

  • 1.231 - Fire
  • 1.193 - Flying
  • 1.148 - Electric
  • 1.135 - Dark
  • 1.124 - Rock
  • 1.115 - Ice
  • 1.029 - Ghost
  • 1.020 - Fighting
  • 1.016 - Water
  • 1.014 - Psychic
  • 0.981 - Dragon
  • 0.941 - Ground
  • 0.921 - Bug
  • 0.882 - Grass
  • 0.868 - Normal
  • 0.786 - Poison
  • 0.753 - Steel

 

E.g. Chansey with its ~58% rate acts like a rich "resource of effectiveness". The Fighting type gets a huge boost by that, as does Fire get from the 4x damage against ~40% Forretress. Flying gets boosted thanks to the popularity of Breloom and Heracross. On the other hand the Ground type ends up rather underwhelming, because a lot of the prominent OU threats (>1000 raw usage) are immune to it. No surprise really, that Poison and Steel are the worst.

This could look a bit different for UU, NU or LC...

 

To-do: Gather data about move pools to take accessibility and strength of physical and special moves, STAB and offensive stats into consideration (kinda long-term)

2.1.2. OU Top & Flop Speed - Sweepers

Spoiler

From "MinSpeed" in the chart. Calculated with base speed + 31 IVs only. EVs and nature scale linearly, thus counting those in wouldn't make much of a difference.

 

TOP 5:

  • 150 (max. 200) - Aerodactyl, Jolteon
  • 140 (max. 189) - Sceptile, Alakazam
  • 135 (max. 183) - Starmie
  • 130 (max. 178) - Tauros, Espeon
  • 120 (max. 167) - Flygon, Charizard, Typhlosion, Linoone, (Slaking)

 

These are solid threats to look out for and, in my view, easy choices to wreck other sweepers. What else would they need that speed for? Of course, there also are the likes of salac berry, swift band, dragon dance, agility, swift swim, paralyze and what not, so it's obviously not that black and white. Their base speed however gives them quite an advantage, so I would say it's advised to either be able to wall them or have a way of outspeeding them, as their trainers usually don't bother with further boosting that speed other than training EVs and having a speed nature.

 

FLOP 5:

  • 60 - Rhydon
  • 65 - Marowak
  • 75 - Machamp, Ursaring
  • 80 - Swampert
  • 90 - Breloom

 

Interestingly enough, these are rather unpopular (or, in case of Breloom, hated) physical hitters, that can even get outsped by some walls. Still, the punch and/or bulkiness they pack is nothing to laugh at and parts of the reason why they are OU.

2.1.3. OU Speed - Walls

Spoiler

From "MinSpeed", too.

 

  • 101 - Milotic
  • 90 - Skarmory, Ludicolo
  • 85 - Umbreon
  • 80 - Porygon2, Weezing
  • 70 - Chansey
  • 60 - Forretress
  • 50 - Slowbro
  • 45 - Dusclops

 

The race of turtles, more or less. Can't hurt to know. Milotic with thunder wave or Ludicolo with giga drain can be quite annoying, when they're able to attack first.

I hope these are all OU Pkmn that are considered walls.

 

2.2. OU Defense

 

The following applies to all of the defense results. The weighted arithmetic mean here represents the average effectiveness of a specific distribution of attack types. It is then used as an attack modifier to determine the damage dealt to the pokemon in question. By this process the defense and HP stats, EVs, IVs and nature find their way into the result.

As of now, the weights I calculated are just taking into account, that almost every Pkmn uses one or two STAB moves, while unfortunately neglecting the bonus itself. E.g. a Water Pkmn will get a better defense result, when there are less Grass and Electric attackers. The more Pkmn there are, who can seriously harm the defender in question, the worse the latter's defense result. The calculation is still flawed, because its foundation is a lazy abstraction I made from the sample: Suppose all Pkmn only use their respective STAB moves. It's not entirely wrong, but I think I should be able to find a better approximation soon...

 

The numbers you'll see represent the fraction of total HP these Pkmn lose on getting hit by an "average typed attack" with 100 strength and a 150 attack stat. At this point, no attack modifier other than type effectiveness is considered.

 

To do: Again, analyze move pools and create weights, which count in the distribution of strengths of attacks, STABs and attack stats of offenders.

 

[WIP] 2.2.1. OU Defenders - Walls

Spoiler

The creme de la creme of meat shielding and annoyance. See %dmg values next to "MaxPDef" and "MaxSDef". Everything except Chansey with 252 HP EVs, 252 Def EVs, matching nature and 31 IVs in all relevant stats.

 

TOP physical walls:

  • 15.4% - Skarmory
  • 19.8% - Forretress (Note: Actually 252 SpDef + Careful seems to be more of use, acc. to Smogon, changing soon)
  • 20.0% - Slowbro
  • 20.2% - Umbreon
  • 22.2% - Weezing

 

Nothing too surprising here, except that Slaking and Metagross end up tankier than Forretress on paper, if you run them with full physical defense...

 

TOP special walls:

  • 15.4% - Chansey (6 HP, 252 Def, 252 SpDef spread and Bold nature; Thanks to BurntZebra for mentioning)
  • 18.1% - Umbreon
  • 22.6% - Dusclops
  • 24.9% - Porygon2
  • 26.0% - Ludicolo

 

Milotic (18,7%) really is the third best special defender on paper, but its ability makes it more worth going for physical defense. Also, interestingly Skarmory (23.8%) and Slowbro (24.3%) would be even better at special walling than Porygon2, but that's just the effect of not taking the all-too popular BoltBeam into consideration according to the current lazy weighting.

 

There certainly are some outliers, that wouldn't get such a score if actual move pools would be taken into consideration.

[WIP] 2.2.2. OU Defenders - Walls in their "off-defense"

Spoiler

average %HP lost by popular walls against the attacks they aren't specialized in tanking, in case of unlucky switching, lures, mixed attackers or whatever... Shit happens. Taken from "%dmg_maxHP" next to "MinPDef" or "MinSDef". 252 HP EVs and 31 IVs included, but no defensive investment, except for Chansey.

 

TOP special walls being physical-attacked:

  • 27.3% - Umbreon
  • 29.7% - Dusclops
  • 35.9% - Porygon2
  • 37.9% - Chansey (6 HP, 252 Def, 252 SpDef spread and Bold nature; Thanks to BurntZebra for mentioning)
  • 47.2% - Ludicolo

 

TOP physical walls being special-attacked:

  • 23.8% - Umbreon
  • 34.8% - Slowbro
  • 35.0% - Skarmory
  • 45.3% - Weezing
  • 50.2% - Forretress (note: 252 SpDef + Careful seems to be of more use, acc. to Smogon)

 

Again, in the face of the ubiquitous BoltBeam Slowbro and Skarmory shouldn't look that good...

[WIP] 2.2.3. OU Top Defenders - Bulky Sweepers

Spoiler

Taken from "%dmg_maxHP" next to "MinPDef" or "MinSDef". 252 HP EVs and 31 IVs included, but no defensive investment. However, note that not all bulky sweepers run 100% HP EVs, some distribute HP and speed in a specific way to be able to ouspeed certain things, e.g. Agility Meta, Dragon Dance Gyara or Swift Swim Kingdra...

 

TOP physical bulk:

  • 22.7% - Metagross
  • 24.7% - Slaking
  • 31.0% - Feraligatr
  • 31.2% - Kingdra
  • 33.0% - Gyarados

 

TOP special bulk:

  • 27.4% - Gyarados
  • 30.6% - Metagross
  • 31.2% - Kingdra
  • 33.6% - Swampert
  • 34.5% - Slaking

 

Interestingly, some of them usually use speed steroids and, once boosted, transform into real monsters. Don't know what to think of Slaking though...

 

To do: Maybe make seperate lists for physical and special sweepers.

[WIP] 2.2.4. OU Top & Flop Defenders - (more or less) Speedy Sweepers

Spoiler

Taken from "%dmg_minHP" next to "MinPDef" or "MinSDef". Only 31 IVs in all relevant stats included, no defensive investment whatsoever. These are the Pkmn, you usually find running 252 Attack and 252 Speed EVs. The results are expected to be bad in respect to real defenders, but there are some who might survive a hit more than others. So it can't hurt to know...

 

TOP physical defense for sweepers:

  • 43.9% - Tauros
  • 44.4% - Flygon
  • 46.7% - Marowak (speedy variant)
  • 47.5% - Charizard
  • 48.0% - Ursaring (speedy variant)

 

TOP special defense for sweepers:

  • 38.2% - Gardevoir
  • 44.2% - Jolteon
  • 44.4% - Flygon, Charizard
  • 45.5% - Espeon
  • 48.0% - Ursaring (speedy variant)

 

Surprisingly bulky at times. Of course, they are still far from safe when facing super effective moves. But if you happen to switch in on a not so effective attack, they should be capable of taking the hit with relative ease.

 

FLOP physical defense for sweepers:

  • 85.4% - Alakazam
  • 78.5% - Breloom
  • 77.5% - Haunter
  • 71.5% - Sceptile
  • 69.1% - Houndoom

 

FLOP special defense for sweepers:

  • 97.6% - Breloom
  • 88.8% - Rhydon (speedy variant)
  • 67.4% - Haunter
  • 63.6% - Blaziken
  • 63.3% - Zangoose

 

The glass cannons of OU. In my view, this is especially useful to know, as you can wreck havoc among them without even using super effective moves. Because my sample attack is actually a bit weak compared to the damage real sweepers can dish out, a well placed STAB attack essentially is a death sentence for them.

 

To do: Maybe make seperate lists for physical and special sweepers. Exclude speedy variants of slow sweepers later, when builds are considered in the equation.

 

Thanks for reading! :)

Edited by Schwubbeldiwubb

Share this post


Link to post
10 hours ago, SpartacusGD said:

very nerdy but this is very nice of you

I'm studying Physics and am kinda used to this kind of stuff. Also, doing this is almost as fun for me as playing PokeMMO, so I'll take that as a compliment. :^)

Edited by Schwubbeldiwubb

Share this post


Link to post

Just one quick nitpick on the thread

2.2.2. OU Defenders - Walls in their "off-defense"

For this section, you gave all the pokemon 252 hp and 0 defense, but this is a bit inaccurate, at least in actual competitive play. EV's only boost stats fairly consistently, regardless of the base stat. So 252 hp evs on chansey changes chansey's hp from 325 to 357, slightly less than a 10% boost, while 252 defense evs on chansey changes chansey's defense from 27 to 62 (using a bold nature), a whole 130% boost in defense, which somewhat changes chansey's bulk on the physical side.

From this

252 Atk Choice Band Heracross Low Kick (60 BP) vs. 252 HP / 0+ Def Chansey: 662-782 (185.4 - 219%) -- guaranteed OHKO

to

252 Atk Choice Band Heracross Low Kick (60 BP) vs. 252 HP / 252+ Def Chansey: 288-342 (80.6 - 95.7%) -- guaranteed 2HKO after Leftovers recovery

or

252 Atk Choice Band Heracross Low Kick (60 BP) vs. 4 HP / 252+ Def Chansey: 288-342 (88.3 - 104.9%) -- 31.3% chance to OHKO

 

I do like the overall concept of the thread, but there are some intricacies in certain aspects of pokemon that you can't just apply some blanket rule to all pokemon, when there are unique cases. 

Zehkar and Schwubbeldiwubb like this

Share this post


Link to post

Thanks for mentioning that. Absolutely makes sense, Chansey will be corrected in the guide. :D (the chart will get updated later)

 

I wondered myself and that's part of the reason why atm I'm trying to work in the main move sets and EV spreads, gathered from Gbush's guide and Smogon (in which unfortunately some builds don't make sense for PokeMMO because of the new gen physical/special spread). As I said, the current chart and results are to be taken with a grain of salt, but it will get a lot better approximated soontm. You'll still need salt for that, but not as much... :^)

When I'm done with that, then there are still various builds, items, actual effect of speed stats in a battle, status attacks, healing and "switching pressure", that lack in the statistical consideration. I'd need waaay more detailed play rate statistics (distribution of different builds of Pkmn) and observations of countless battles over time, e.g. for effective damage calculation of moves like Will-O-Wisp and Toxic or Dragon Dance... I don't think I'll be doing this much. xD

 

But when I'm done with WIP, the numbers should give you quite detailed "stationary" insights in respect to several specific averages.

 

Edited by Schwubbeldiwubb

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now