[e2e] 150ms - tolerable latency for quakeIII

grenville armitage gja at ureach.com
Thu May 24 13:06:25 PDT 2001

"Thomas H. Ptacek" wrote:
> On 23 May 2001 13:35:25 -0700, grenville armitage wrote:
> > than estimating based on human factors, I think the histograms
> > reveal players voting with their feet. Based on the distributions,
> I think you need to provide more of your methodology.

Fairly simplistic right now. Run two public QIII servers for
approx. 65 days, and log the end-of-game ping times reported by
QIII server (roughly every 15min).
> Is the 150ms number the result of people "voting with their feet" due to
> intolerable latency, or due to the fact that Quake3 scores (and sorts?)
> servers by their latency, and users will tend to choose the "best
> scored" server, regardless of whether many "poorer scored" servers might
> have acceptable lag?

In my experience, neither the in-client game launcher, nor Gamespy3D,
tend to 'score' QuakeIII servers per se. Players are able to sort the
lists of public servers according to ping (ascending or descending, at
least in Gamespy), number of other players, game name, etc... so I
surmise there's nothing forcing players towards lower ping time
servers aside from player preference.

> In which case, is it reasonable to then say that a statistically
> significant number of users have connectivity putting them at 150ms
> distance from a typical Quake3 server?

It is certainly an interesting question what size my sample pool
is of players within 150ms of my servers vs the pool of players
outside this range. I'm not even sure yet how to estimate the
potential pool of players at any given temporal radius around
a given point on the internet.
> Or are you measuring the latency to players leaving mid-game, in which
> case you could claim to be measuring the latency at which players tend
> to become disgruntled due to lag?

Actually, this first set of results was somewhat opportunistic.
I noticed that my logfiles contained end-of-game ping times, and
decided to chart them. Thus, there's no insight into the ping times
of people who left, only people who stayed. I've since re-instrumented
my server to collect in-game ping estimates, we'll see what that


More information about the end2end-interest mailing list