jim said:
>   I just checked my previous claims about the problems with P3
>   (for example) against a pretty straight-forward text.  
>   I reviewed 200 pages from that text.  
>   P3's made 38 changes on those pages.  
>   Of these changes 7 change represented a positive
>   contribution towards making the txt correct.
>   Of those positive changes about a half could
>   easily be found by a simple tool like guiguts.
>   10 of the changes introduced by the P3's were negative changes
>   -- changes that moved the text to a less perfect state.

jim, and others, if you're going to continue to discuss the
problems in the current system at distributed proofreaders,
please do it in your own thread, and not in my threads, ok?

as for your findings, jim, it helps to report the actual lines.

but it's probably not necessary.

in the past, i have documented the same things you report,
in great detail, in book after book after book, with evidence.

in comparison, your anecdotal reports are relatively flimsy.

i'm not saying you're wrong.  indeed, you're absolutely correct.
i'm just saying your reports are not going to convince anyone.
heck, there are people here who refused to believe what i said,
in spite of the fact i piled up enough evidence to choke a horse.

(nor did i _create_ the evidence; i used data taken directly from
various experiments, performed over at d.p. by other people...
the truth is out there, and easy to find, if you just care to look.
this is what is so silly about all these "experiments".  i've told
everyone here the simple correct answers, so all that's needed
is to test these simple hypotheses and see they _are_ correct.
but instead people are testing overly complicated stuff in ways
that are not definitive, leading them to become more confused.)

***

perhaps the most impressive findings of my results were these:

1.  the best way to know a page is "finished" is when proofers
stop making changes to it...  up to that point, it's not finished!
it's the "best" way to know because a no-diff is easy to measure.

2.  3 rounds of p1 were as effective as a series of p1-p2-p3.
(simple solution to queue problems?  run the text through
p1 until every page comes out repeatedly as a no-diff page.)

3.  the third round of p1 found as many additional errors
as the p3 round, but _neither_ route found _all_ the errors.
the p1(3) proofers found errors that the p3 proofers missed,
and the p3 proofers found errors the p1(3) proofers missed,
plus there were other errors that both p1(3) and p3 missed.
(takeaway: the p1 proofers are not inferior to the p3 proofers,
and a "make the page better" philosophy will eventually work
to "create a perfect page", without all the attendant pressure.)

4.  "parallel" p1 was _not_ useful at turning up any more errors,
but it might have value to determine a page is "done", although
more research would need to be done to test that hypothesis...

-bowerbird