- a different board layout (it was even more different, including quad word scores, but they changed it to something closer to the original);
- fewer tiles, some of which score higher; and
- eight-tile racks, with a 40-point medium bonus for playing seven tiles and the regular 50-point bingo bonus for playing all eight.
I was looking around the internets to see if anyone had taken a crack at adapting computer programs to play the lexolous rules (Quackle allows different boards at runtime, but the tile sets and rack sizes are hardcoded), and I found the site of this crackpot whose stuff I've looked at before. He invented Octo Scrabble (scrabble with eight-tile racks, including a scaled bingo scoring system) a year ago, and swears by it.
(The reason I call him a "crackpot," though he is possibly more accurately merely an especially creative curmudgeon, is because he also has suggestions for how to reform basketball, the justice system, piano notation, units of measure, parade viewing, and ice water consumption. Among many, many other things. Also, it seems that 55 years ago one could buy leg padding not two blocks from my house.)
Back to the topic at hand, the other differences in the game don't seem as major as the rack enlengthening...you can sort of see some of Andrew's and my plays sticking to the new lines of multiplier squares in our first game the way they stick to their counterparts in regular corporate scrabble, but that stands to reason, and since the layout is pretty similar in qualitative (better multipliers as you get further from the center) and almost identical in quantitative terms (the official game has one more TLS per quadrant and they otherwise match), it doesn't seem to affect how the game plays very much.
What I'd really like to do is to run an AI against itself (nullus) for a couple tens of thousands of games under each rule set, and see how the size and variability of scores compare. Until then I did take the time to chart the differences in tile scores and frequencies:
Really, they seem to have mostly just scaled back the number of tiles (resulting in more games played, which translates to more "x started a new game of Lexulous" messages on Facebook), upping the tile values to keep scores up (Lexulous has 11 fewer tiles, but they average 2.3 points to Scrabble's 1.9).
So that's all I've got on that score. Realistically, I am unlikely to try to hack Quackle myself to get an idea about how the lexolous rules change the familiar game (it builds on Windows, but with MinGW, which I don't have installed, and Qt, which just kill me now).