[WoW] Battle for Azeroth - Page 261

The basis for that tool comes from this MMO-C thread and the spreadsheet therein. The killpoint system itself is very interesting and only has a few minor flaws (mostly related to the difficulty of obtaining useful and accurate information from the Armory). The major problem is that the spreadsheet's calculation for the number of killpoints required per legendary is simply wrong.

The obvious, intuitive way to make the calculations would be to average the killpoints of everyone who's gone from N to N+1 legendaries. That is, sum up all the killpoints each player acquired between N and N+1, then divide by the number of said players. If you do that with available data you'll find that outside of the first legendary (which is substantially easier to obtain thanks to a hotfix by Blizzard a while back) all subsequent legendaries take roughly the same number of killpoints (the 250-300 KP I mentioned ealier).

The spreadsheet is doing two things beyond a simple average which make no mathematical sense:

• They're including players who still only have N legendaries.
  
• They're diving the resulting number by the % of all players who have reached N+1 legendaries

Including players who haven't yet reached N+1 legendaries skews the numbers. The flaw in the methodology of adding them becomes obvious if you consider potential data extremes. Imagine you have 1000 people with N legendaries, only one of whom has N+1. If the other 999 all stopped doing anything the instant they got N legendaries (i.e. the number of killpoints they've collected since N is 0) the calculation would erroneously suggest that obtaining the N+1 legendary should only take 1/1000th of the time it took the one player who has one. Simply put, only players who have actually obtained the N+1 legendary should be considered.

Dividing the resulting number by the % of all players who have reached N+1 legendaries is similarly fallacious. If this was a calculation of half-life it would make sense, but because players do not start progressing at the same time, do not progress at the same rate, and completion of progress is not the same due to RNG, it's an utterly incorrect model to use. Imagine using that calculation to try to predict how long it takes to finish an Olympic race after 1/10th of the runners had finished. If the 1/10th of the runners took 10 minutes to finish, dividing by the % of all runners would suggest that it normally takes 100 minutes. That's blatantly absurd, but it's how the calculation is being done.

Unfortunately despite the fact that the flaws in the math have been pointed out numerous times in both the MMO-C thread and on related reddit threads the author of the spreadsheet has never corrected their math, and while the work done on killpoints is novel the breakpoints provided by the spreadsheet should not be taken seriously.