kills: the api total KILLS per world are 10-50% lower than they are ingame in the wvw window. This way the total api kills seem to be 1-7 skirmishes too old (2-14 hours too old).
deaths: the api total DEATHS per world are close to the wvw window values. During the first skirmish right after the match reset they can be a bit lower. Later the api deaths will be HIGHER than the wvw window deaths. (strange)
when: It is not something that suddenly happens at a certain point during the match up. The kills and deaths are wrong right from the start of the first skirmish of a new match after the reset.
k/d: because the api kills are much lower and the api deaths are higher than the ingame values, the k/d ratio calculated from api kills/deaths is typcially too low by about 0.1-0.3 for every world (e.g. ingame red 1.0 is api red 0.8)
has been tested for 2 wvw match ups on dates between 15.08.21-21.08.21 at different times.
screenshots:
left ingame wvw window. right tool that polls the api (kills, deaths, k/d)
at reset 20.08.21:
previous match 15.08.21
Expected Behavior
kills and deaths total and per map should be close to the ingame wvw window values. A delay of 5-15 minutes is okay. But not several hours and the k/d should not be distorted that much
Steps To Reproduce
open wvw window ingame.
Select 2nd tab "Match Overview".
select the "Match History" tab
4 select "Kill/Death (Total)" tab
Maybe related, but probably not really: https://github.com/arenanet/api-cdi/issues/417#issuecomment-268643700
Just an interesting note about how lye implemented the tracking bit. Given that he mentioned that it was kind of a legacy diagnostic already back in 2016, might mean that the value is no longer properly tracked almost 5 years later.
Just a guess and for @dsnider-anet to check, but maybe it's the outnumbered buff preventing counting the kills/deaths correctly because that buff prevents war score from being granted to the killing color?
Current Behavior
wvw/matches/stats returns wrong kills and deaths. example match (millersund world, EU): https://api.guildwars2.com/v2/wvw/matches/stats?world=2206
has been tested for 2 wvw match ups on dates between 15.08.21-21.08.21 at different times. screenshots: left ingame wvw window. right tool that polls the api (kills, deaths, k/d)
at reset 20.08.21:
previous match 15.08.21
Expected Behavior
kills and deaths total and per map should be close to the ingame wvw window values. A delay of 5-15 minutes is okay. But not several hours and the k/d should not be distorted that much
Steps To Reproduce
Happening since
not sure. at least some months (forum thread if from november 2020), probably much longer
Anything else?
impact: this affects a lot of wvw websites and tools. examples:
websites
tools
script used by twitch streamers
Maybe related, but probably not really: https://github.com/arenanet/api-cdi/issues/417#issuecomment-268643700 Just an interesting note about how lye implemented the tracking bit. Given that he mentioned that it was kind of a legacy diagnostic already back in 2016, might mean that the value is no longer properly tracked almost 5 years later.