Futures Rewritten

Please note: This is a naive implementation which I intend to improve over time. Information on the methodology can be found at the bottom of the page.

The Balance - Melee

Opener:

Job Average Max Total
Ninja 1021843 1191178 11
Samurai 972279 1202359 12
Dragoon 941716 1063314 39
Monk 851554 957263 18
Viper 820427 937050 30
DarkKnight 748000 961437 53
Reaper 743183 1097406 11
Warrior 588919 687196 9

2M Burst:

Job Average Max Total
Dragoon 915126 1514544 316
Reaper 875046 1381701 88
Samurai 852637 1535223 102
Ninja 808805 1436259 95
Viper 755852 1326331 246
Monk 743330 1431036 155
DarkKnight 600054 1327451 439
Paladin 532392 901435 396

The Spear - Ranged

Opener:

Job Average Max Total
Pictomancer 1087031 1411486 48
Machinist 783260 879386 3
RedMage 754836 926179 8
Summoner 730123 966816 4
BlackMage 642200 761398 14
Dancer 626807 771953 25
Ninja 512910 570549 11
Bard 511261 774371 37

2M Burst:

Job Average Max Total
Pictomancer 924055 1814782 384
Summoner 817596 1205713 32
RedMage 743179 1445464 64
Machinist 728175 1163964 24
Bard 721707 1231650 296
Dancer 641198 1300598 200
BlackMage 593759 938798 112
Reaper 410642 665470 88

Methodology:

Windows in a fight are only counted towards the total tracked average if the boss is killed.

If a card is played within thirty seconds of the fight starting, the window is considered a part of the opener.

Every time a card is played, the card calculator checks the damage done during the duration of the card by all players. This is compiled into an average (mean) based on the total number of windows checked. Maximum and averages are currently both being calculated at the moment the card is played. In the future, I will include data from the 'ideal' card uses that the card calculator finds.

Every time the average is updated, it's re-calculated using the following:

new_average = old_average + (damage_from_damage_window - old_average) / (number_of_data_points + 1)

Changes I'd Like To Make:

  • Ignore outlier data, specifically if the damage is exceptionally low compared to the existing average. This would help with any windows being ended early beacuse of killtime, or any deaths on clear runs.
  • Incorporate the 'true' optimal target which the calculator calculates.
  • Consider whether a window has a pot active or not.