Eloise and I have been working through things to do in the evenings since moving to Scotland. So far, that's included several films and the first series of Yes Minister on DVD. We also have a pile of old board games to get through though, including
Blood Bowl - a Games Workshop game that's basically a more violent version of American football set in the Warhammer world. It must be well over ten years since I played and the first time for Eloise, but it was good fun.
For the record, I was orcs and she was humans. I was receiving the ball (and thus attacking) in the first half, but got so stuck into pummelling her team that I forgot to do anything with the ball until too late. The strategy paid dividends though as by half-time I'd ko-d three of her lineman. We'd both had a man sent off for fouls, but I was able to start the second period with a numerical advantage of 11 to 8, which gave me a greater chance to stop her scoring.
As it is, Eloise came pretty close to the opening touchdown, managing to run a catcher into my end zone and get a thrower with the ball dangerously close - only for him to fall over attempting to dodge by blitzer and open up a good scoring opportunity. At this point, one of my orcs was able to scoop up the ball, while I inflicted further damage by pushing two more of her players into the crowd (where they were also ko-d). By the end of the match, she only had four men left on the pitch - who were regularly getting knocked down - though it still required my lineman to 'go for it' (and use my final team re-roll in doing so) in order to score the game's only touchdown in the final turn. A fine example, I think, of the orc strategy (i.e. concentrate on roughing up the opponents first and then use numerical superiority to score a running touchdown).
I think I agree with Bill Rapaport, that the R&R system is not necessarily a bad one, it's just that it doesn't always work as well as it should. I certainly know that I've had plenty of papers improve substantially as a result of a R&R verdict, whether or not subsequently accepted.
On the other hand, journal editors need to exercise discretion. I think that there can be value in sending the revised paper to a totally new referee (ideally along with the original ones), to get a fresh look at it. Problems arise when there's a long time lag involved (though unavoidable at times), when the referees give conflicting recommendations/advice, and when some of their comments are idiosyncratic. A good editor should of course be willing to accept a paper even if the author doesn't do everything the referees suggest.
There is one other problem with the system, which probably affects me more as a referee than an author, and that is that I probably do tend to give to many R&R verdicts. It's a safe 'middle ground' for the many papers that show some promise but also have some flaws. Again, though, the editor's role is important here, since there's a difference between the verdict given by each of two or three referees and the overall verdict given by the journal. The editor needs to make the call when, say, an accept and an R&R verdict ought to be an acceptance (or conditional acceptance), rather than R&R.
That the defects aren't necessary to the system doesn't, however, mean that the system ought not to be abolished. It might be that the chances of having a well functioning R&R process are so slim that we would be better off abolishing it. But I don't think that's so. I think that - while there will occasionally be cases where things go wrong - most journals can manage the system reasonably well (and it is of course open for any given journal to decide how many R&Rs, if any, to give).
One possible improvement would be to have more distinction within R&Rs. Some journals/editors do seem to distinguish minor and major revisions, but it seems to me that R&R can sometimes basically be a conditional acceptance and other times is no more than a vague willingness to look at the paper again if you completely re-write it (rather than permanently rejecting it). Perhaps it would be useful if these were always clearly distinguished.