1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
  2. ATTENTION: For our 2023 season all of our tournaments will be hosted via the Pokémon Perfect Discord server rather than the forums. Please join us there and continue to enjoy our tournaments! https://discord.gg/2CsWWnan2A
  3. Tournaments

    Check out the 2024 Tournament Calendar and join our discord server to participate in our tournaments!

ADV OU ADV Archetypes: Data Analysis and Organized replays of 355 Top Level ADV Matches

Discussion in 'ADV Discussion' started by deluks917, Mar 17, 2018.

  1. deluks917

    deluks917 Host Emeritus

    Joined:
    Jan 21, 2016
    Messages:
    238
    Likes Received:
    250
    I was very interested in which archetypes under and over-perform in the ADV metagame. My original goal was to find ways to exploit the metagame. However over-all ADV is quite balanced. I orignally published this analysis week1 of SPL in the sharks chat. However I have since updated the numbers to include the regular season SPL9 Games. The data is from SPL7, SPL8, WCOP11, WCOP12, The Callous Invitational and SPL9.

    Results:
    Cloyster_Spikes: 54% - Games: 57
    Jolteon_Skarm: 54% - Games: 44
    Skarm_Spikes: 51% - Games: 137 (18 mirrors)
    Skarm_Mag: 51% - Games: 55 (6 mirrors)
    Mixed_Offense: 50% - Games: 92 (12 mirrors)
    Spikeless_Balance: 49% - Games: 81 (12 mirrors)
    Physical_Mag: 48% - Games: 78 (14 mirrors)
    Forre_Spikes: 48% - Games: 74 (4 mirrors)
    Special_Offense: 47% - Games: 36
    Physical_Offense: 46% - Games: 45
    Misc: 14% - Games: 7

    (mirror matches are excluded)

    Most strategies win close to 50%. The main over-performers were Cloyster and Jolteon. The main under-performer was magneton-less physical teams. Given that skarm_spikes was the most played archetype and skarm_jolt/skarm_mag also saw significant play this is perhaps unsurprising.

    MU Replays:

    I am also including a document that organizes the replays by mu. If one is interested in seeing how special offense vs skarm spikes actually plays out there are several replays to look at. The same is true for almost all mus. I also include the winrate per mu but the samples sizes are ussually very small and not significant. Since this document is quite large I am linking it as a pastebin: Replays Grouped By MU

    Full Data:

    I am including the full data in case someone else wants to take a look and maybe correct any errors. The data consists of two JSON files. One JSON file is an array of replays: the replay url, the players, the pokemon used, etc. This file is too big for a single pastebin so its split in two. The second JSON file contains the tags.

    Replays Part 1
    Replays Part 2
    Tag File
     
    Last edited: Apr 9, 2018
  2. undisputed

    undisputed Member

    Joined:
    Apr 30, 2017
    Messages:
    53
    Likes Received:
    84
    This is interesting analysis. Thank you for doing this Deluks.

    I have a quick question though. You said Cloyster_Spikes was "already 58% before [you] added the SPL9 games." However, in SPL 9, Cloyster is 1 win and 6 losses. How did Cloyster Spikes increase in win percentage to 59%?
     
    deluks917 and Isa like this.
  3. deluks917

    deluks917 Host Emeritus

    Joined:
    Jan 21, 2016
    Messages:
    238
    Likes Received:
    250
    Yeah I made a mistake entering the SPL9 data. Will fix soon.
     
    undisputed likes this.
  4. deluks917

    deluks917 Host Emeritus

    Joined:
    Jan 21, 2016
    Messages:
    238
    Likes Received:
    250
    Errors have been corrected. I also added links to the raw data. Thanks undisputed for catching the previous mistake.
     
    undisputed likes this.
  5. Disaster Area

    Disaster Area Little Ball of Furr and Power Member

    Joined:
    May 4, 2014
    Messages:
    5,617
    Likes Received:
    2,520
    the results seem statistically insignificant to me.

    which I guess is a significant result in and of itself, if it appears that no playstyle has a statistically significant advantage versus the metagame as a whole.


    Maybe seeing how individual styles of team match up against other styles could tell us more.
     
  6. deluks917

    deluks917 Host Emeritus

    Joined:
    Jan 21, 2016
    Messages:
    238
    Likes Received:
    250
    Yes the main result I see is that the playstyle winrates are close to 50%. This was non-obvious to me before looking at the data. Something having a 48% or 51% winrate in the data is not really that different from 50%.
     
    Disaster Area likes this.

Share This Page