win rate of first team building a nuke VS numbers of players in the game

[yaniv-aknin]

win rate of first team building a nuke VS numbers of players in the game

[yaniv-aknin]

Heya, thanks for this idea! I started looking at it, here are a few observations I made. Note I rarely play multiplayer/team games, so maybe I'm a bit clueless about them :)

• It's not easy to know when a nuke was built, but easy to know when a nuke was launched; so I'll go with that
• I assume we'll look at all VALID games with more than two players
• I'll look at all games where one team has at least one player with VICTORY result, and the other teams do not have a victory result
• Not sure what's a good visualization plot; maybe a set of violin charts, one per number of players, each mark would represent a victory or loss for the team and the height would be 'time to nuke'

[yaniv-aknin]

Hmm, we'll also look at the first nuke launched by any given team.

[yaniv-aknin]

I got this far (looking at 10% of nuke launches on 1st January 2022 in valid >2 player games; for the real answer I'll look at all nuke launches in 2022Q1) -

WITH
  nuke_launches AS (
    SELECT id, player, offset_ms,
    FROM `fafalytics.misc.cmds_q122`
    TABLESAMPLE SYSTEM (10 PERCENT)
    WHERE `fafalytics.faf_ladder_1v1.COMMAND_NAME`(CAST(JSON_EXTRACT_SCALAR(payload, "$.cmd_data.command_type") AS INT64)) = "Nuke"
      AND dt = "2022-01-01"
  ),

  q122_valid_multiplayer_games AS (
    SELECT *, ARRAY_LENGTH(g.playerStats_gamePlayerStats) total_players
    FROM `fafalytics.faf.games_denorm` g
    WHERE ARRAY_LENGTH(g.playerStats_gamePlayerStats) > 2
      AND g.validity = "VALID"
      AND DATE(startTime) = "2022-01-01"
  )

SELECT g.id, ARRAY_AGG(ps.player_player.login), ps.team, "VICTORY" IN UNNEST(ARRAY_AGG(ps.result)) is_victory, MIN(n.offset_ms) first_nuke_ms,
FROM q122_valid_multiplayer_games g
JOIN UNNEST(g.playerStats_gamePlayerStats) ps
INNER JOIN `fafalytics.misc.meta_q122` m ON (g.id, ps.player_player.id) = (m.id, m.faf_player_id)
LEFT JOIN nuke_launches n ON (m.id, m.replay_player_id) = (n.id, CAST(n.player AS STRING))
GROUP BY id, team
ORDER BY id

Which yields a table like so:

id	f0_	team	is_victory	first_nuke_ms
16026235	[JTuceHok,Arler]	3	false
16026235	[S_Monstr,The_Lupes]	2	true
16026241	[Asgaria,TheHardModer,iosh,AsteroidComa128,crustaceanMODE,brack8]	3	false
16026245	[Bigwebmaster,YourMom7,YourMom7]	2	true	2555000
16026245	[Insulner,Lidake]	3	false
16026246	[gonec,Pablo-Escobar,XenocideX,Zickfried]	3	false
16026246	[KapToqpKa,Dadkys,Cetuk,MIHAI]	2	true
16026256	[PaintThinner,SealForce,FFF,AgeOfEpicness,Paradox_of_War]	2	false

Which kinda looks plausible. :)

But I ran out of time, so I'll revisit this issue when I'm next free.

[yaniv-aknin]

I think this is beginning to look interesting.

Here's a line plot showing "average time to first nuke, by team size, and did the team win or lose":

Now a violin plot that gives you an idea of density:

And perhaps most interesting, but not very easy to read, is one where we consider "no nukes built" to be "0"; this is useful to show how many games are decided without a nuke at all, and especially the difference between win/loss when not building nukes (more teams win if they launched at least one nuke):

Would love to hear your thoughts, when I have a moment I'll clean this up and submit a proper analysis.

[yaniv-aknin]

I've asked the pros (reddit, kaggle) - let's see what we get :)

[IndexLibrorumProhibitorum]

So, it looks like nukes are more frequently used as the number of players goes up, but using a nuke might not lead to a statistically significant increase in winchance? Is that about right? I'd like to see this cleaned up a bit and with p values included :P