User:Praise/Armor Penetration Notes

These are some simple notes that will help you make more accurate predictions. I did not investigate the subject exhaustively.

• Armor penetration stacks additively, unlike other percentages in Guild Wars
• The effective armor value is finalised using various rules, before being used in other calculations:
  • If the armor value after armor penetration is not an integer, it is rounded down to an integer

Armor penetration with rounding down example Hammer @12mastery vs 80 armor target

strength attribute expected armor (no rounding) expected dmg (no rounding) expected armor (with rounding down) expected dmg (with armor rounding) observed dmg 0 80 35.00 80 35.00 35 1 79.2 35.49 79 35.61 36 2 78.4 35.98 78 36.23 36 3 77.6 36.49 77 36.87 37 4 76.8 37.00 76 37.51 38 5 76 37.51 76 37.51 38 6 75.2 38.04 75 38.17 38 7 74.4 38.57 74 38.83 39 8 73.6 39.11 73 39.51 40 9 72.8 39.65 72 40.20 40 10 72 40.20 72 40.20 40 11 71.2 40.77 71 40.91 41 12 70.4 41.33 70 41.62 42 13 69.6 41.91 69 42.35 42 14 68.8 42.50 68 43.09 43 15 68 43.09 68 43.09 43 16 67.2 43.69 67 43.84 44

• • If the armor value after armor penetration is an integer, and:
    • if it is an even number, it is used as the finalised value
    • if it is an odd number, 1 is subtracted

Armor penetration with subtracting from odds example Hammer @12mastery vs 100 armor target

strength expected armor expected dmg observed dmg 0 100 24.75 25 1 99 25.18 26 2 98 25.62 26 3 97 26.07 27 4 96 26.53 27 5 95 26.99 27 6 94 27.46 27 7 93 27.94 28 8 92 28.43 28 9 91 28.93 29 10 90 29.43 29 11 89 29.95 30 12 88 30.47 30 13 87 31.00 32 14 86 31.54 32 15 85 32.10 33 16 84 32.66 33

Please note that these examples (there are many more not shown here) are simple cases with only one source of armor and one source of penetration. It probably gets more complicated when stacking effects like bonus armor, special armor, cracked armor, flat damage reduction, percentual damage increase, and multiple sources of armor penetration. Rounding or subtraction might be done in between the process of stacking, giving unexpected results. If you have data elucidating (part of) this process, I would be interested to read it.