credits to Jonagold for coming up with this idea in the finalmouse thread
http://www.overclock.net/t/1531877/finalmouse-2015/1850#post_24118688
http://www.overclock.net/t/1531877/finalmouse-2015/1870#post_24121092
AUG 2015 UPDATE: microe (creator of mousetester) releases
http://www.overclock.net/t/1570462/mousecomparator
use this instead. it's much easier and the sums plot are easier to interpret
basically, the procedure is to plug both mice in (use the adjacent ports so that polling is in sync), flip one mice 180, press two mice tightly against each other, and move them together horizontally without slipping.
originally Jonagold and some others were basically looking at which mouse responded first with the first count.
i've found it more useful and consistent to start from rest and push the two mice fairly hard for ~15cm, compare which mouse reaches 10 counts of motion first, and count how many counts the other mouse reported in that interval
disclaimer: i am human and so this applies to me. however i am aware of this and did my best to not cherry-pick plots which show one mice particularly favorably. i repeated each comparison several times to get a feel for the general differences, and below i will show plots that i found to be representative of the typical difference
g502 (positive) vs g303 (negative), 400dpi, 1000hz

almost perfect match, not too surprisingly.
at time = 648ms, both g502 and g303 have moved 10 counts
g502 (positive) vs wmo (negative), 400dpi, 1000hz

very close
at time = 2133ms, wmo has moved 10counts , g502 has moved 9 counts
g502 (positive) vs finalmouse (negative), 400dpi, 500hz (because finalmouse only does 500)

at time=1878ms, g502 has moved 11 counts, finalmouse has moved 5 counts
g502 (positive) vs torq x5 (negative), 400dpi, 1000hz
at time = 573ms, g502 has moved 10 counts, x5 has moved 6 counts.
g502 (positive) vs g100s (negative), 1000dpi, 500hz
both reach 35 counts at time = 344ms
for all of these, i started from rest and moved the two mice as fast as i could for ~15cm or so. the motion of the mice was about as fast as when doing a flick in cs.
i've also done like wmo (positive) vs g502 (negative) and it gives the same results
i've found this to be quite repeatable but you should try with your own mice so that we have more data and can compare.
http://www.overclock.net/t/1531877/finalmouse-2015/1850#post_24118688
http://www.overclock.net/t/1531877/finalmouse-2015/1870#post_24121092
AUG 2015 UPDATE: microe (creator of mousetester) releases
http://www.overclock.net/t/1570462/mousecomparator
use this instead. it's much easier and the sums plot are easier to interpret
basically, the procedure is to plug both mice in (use the adjacent ports so that polling is in sync), flip one mice 180, press two mice tightly against each other, and move them together horizontally without slipping.
originally Jonagold and some others were basically looking at which mouse responded first with the first count.
i've found it more useful and consistent to start from rest and push the two mice fairly hard for ~15cm, compare which mouse reaches 10 counts of motion first, and count how many counts the other mouse reported in that interval
disclaimer: i am human and so this applies to me. however i am aware of this and did my best to not cherry-pick plots which show one mice particularly favorably. i repeated each comparison several times to get a feel for the general differences, and below i will show plots that i found to be representative of the typical difference
g502 (positive) vs g303 (negative), 400dpi, 1000hz
almost perfect match, not too surprisingly.
at time = 648ms, both g502 and g303 have moved 10 counts
g502 (positive) vs wmo (negative), 400dpi, 1000hz
very close
at time = 2133ms, wmo has moved 10counts , g502 has moved 9 counts
g502 (positive) vs finalmouse (negative), 400dpi, 500hz (because finalmouse only does 500)
at time=1878ms, g502 has moved 11 counts, finalmouse has moved 5 counts
g502 (positive) vs torq x5 (negative), 400dpi, 1000hz

at time = 573ms, g502 has moved 10 counts, x5 has moved 6 counts.
g502 (positive) vs g100s (negative), 1000dpi, 500hz

both reach 35 counts at time = 344ms
for all of these, i started from rest and moved the two mice as fast as i could for ~15cm or so. the motion of the mice was about as fast as when doing a flick in cs.
i've also done like wmo (positive) vs g502 (negative) and it gives the same results
i've found this to be quite repeatable but you should try with your own mice so that we have more data and can compare.