Precomputed Tic-Tac-Toe

You can play tic-tac-toe against a set of precomputed data. For any move you make, a response has already been calculated; no scripts will be run, no moves will be analyzed. Your move will simply receive the precomputed response. (In fact, you can figure out what the response will be before clicking; it's encoded in the URL of the link, which is probably displayed at the bottom of your browser when you move the mouse over a link.)

(You can't win.)

I tested these pages with Lynx 2.8.1rel.2, Netscape 4.6, Mozilla M12, MSIE 3.0, and kfm 1.85. They didn't look very good with MSIE 3.0, which doesn't support PNG, but they were playable. They were unpleasantly slow in Mozilla; I'm not sure how to fix that. The images and their alts were not displayed in kfm, rendering the game completely unplayable. With Netscape, I tested in 640x480x256, 1600x1200x16777216, and 1280x1024x16777216. All three resolutions were good.

If you're a geek, you might want to

• peruse the Perl program that does the precomputing;
• download the 40-kilobyte package to make a carbon-copy of this web page (or laugh at how awful my code is),
• know that the correct strategy is explained at a web page at UC Riverside, which lists board patterns to look for. This program doesn't recognize patterns #15 and #16, but I don't understand those; I suspect they aren't really relevant.
• know that this set of precomputed web pages (including 239 board positions, the images, this page, and the above package) takes up 383 kilobytes;
• know that I drew the background in the GIMP by doing Solid Noise on a 200x200 grayscale image, then adjusting the response curve so only a narrow band was bright, then converting to RGB, then running Emboss, then adjusting the color balance. (I reduced the images in size after I drew the X and O.)

  (The images thus produced were unsatisfactory; Emboss doesn't create a nicely tilable output image. I wrote my own 53-line emboss filter in Perl that produces tilable output a while back; I used it to produce the new images.)

  For the page background, I increased the brightness until the result was washed-out enough to not interfere unduly with reading.

Kragen Sitaker <kragen@pobox.com>