Monday, March 4, 2024

counting Hugo ballots

So here's another proposal to fix the Hugos. This one wouldn't hive the entire administration off to a permanent committee (which I think would be a mistake) but would create a continuing committee to watch over the software for counting ballots.

For it seems that "convention committees all seem to have at least one person on them, in a position of authority, who wants to be the one to invent the software suite to rule them all that will solve all future fannish endeavours henceforth," so they all reinvent the wheel, and this was done particularly badly at Chengdu, where McCarty wrote his own software which 1) had plenty of code errors in it 2) can't be corrected because the code is proprietary and he won't release it.

My, how different things are from when I co-administered the Hugos thirty years ago.

First off, in those days almost all ballots were on paper. (We got a few by e-mail. We printed them out, so they'd fit with all the others.)

Second, we only used software to count the final ballots. Nominating ballots and voter ID check were done by hand. The idea of creating software to count the nominating ballots seemed to me ill-advised. There were too many different nominations, too much irregularity in how they were identified. Maybe if there'd been 5 or 10 times as many ballots we'd have been forced into it, but a few hundred nominating ballots, most of them largely empty? Not a problem.

As for that final-ballot software, all three years we used the same program, which had been devised by the administrator from seven years before our first run. Why? Well, it was a reliable program, and its author was a friend of ours. I think others used it too, and I always presumed it (revised by the author as rule changes required) was the standard ballot-counting program, at least for a while.

I'm not computer programmer guy, so I have no idea what computer language it was written in, but I do know the code was simple and accessible. It could be filled out so that the names of the actual finalists would appear on the data entry screens, and then the end users just typed in the sequences of numbers from the ballot. Finish your batch, save the file, run the program for the complete results on that file if you're curious, then when the ballots have all come in, combine all the files and run the program for the final result. Then do recounts on categories where the results are tight. Us two administrators and a couple volunteer assistants did all the data entry work.

This program was only designed for manual data entry, so it couldn't count the electronic ballots used today without inserting an unnecessary and stupid manual step. And I suspect that the EPH rules of the far future would have been beyond it without a massive rewrite.

But it did everything we needed it to do back then. It even calculated adherence to the 5% rule which caused so much vexation in those days. The 1st ... nth place result cascades that we submitted to Locus and other news sources? Those were a direct cut-and-paste from the output of our wonderful little program.

No comments:

Post a Comment