More Info out Phils recordinggreenspun.com : LUSENET : MARP Editors : One Thread
Alright I've complied some stats on phil's possibly illegal recording. Hopefully everyone understands the current situation: phil made a puzzle recording for soukoban deluxe that plays very unusually compared to other players. But there hasn't been concrete evidence untill now. I'll footnote the suspicious facts.
Every solution is solved in phils recording in a tapping manner, as if he was counting the moves before hand and reading them off a cheat sheet (all the other recordings play like you are playing pacman (holding the joystick in the direction you want to go then changing direction, phil goes a direction waits 1/5th of a second then goes in the same or another direction and waits for every single move.) This can be done in the real game, so it doesn't look strange untill i noticed level 5-d. Phil starts to play the level, looks good untill move #210. Then he commits what appears to be the only error (1) in the 5500+ move recording; he misses a "tap" to go up and continues tapping the remaining 122 moves to attempt to complete the level. If you believe it could be (2) possible that he was looking away from the keyboard for the entire level and then said oops i'll restart the level (something you can do in this game legally), here's what happened. after his 232nd tap move is depressed (the last that he makes during the incomplete "error" attempt at this level) he waits 2/3rds of a second (3) and then presses the reset button to start the level over again and this time finish it completely.
I have a few stats to report using the "analinp -H -k phil.inp" option. This option reports all key strokes of the recording (even if multiple keys were held down) and for how long each input state lasts, even states where no input is given in the recording. I cut and pasted phils two level 5-d attempts and found out they are the same moves except for one missed up move in the first attempt. On phils side, The frame lengths of the presses and releases differ in that presses as if a human had made them, this aspect doesn't rule out the RP cheat.
(1) 1 error in 5500+ moves. do the math, can anyone be that perfect? (this can't be explained)
(2) knowing the moves of a level and looking away from the monitor is impossible for me to do for one 100 move level (this can't be explained)
(3) notices that he's been doing 122 moves wrong in less than a second after finishing them, suposedly blindly from a cheat sheet? (this could be explained by the fact that the reset button is pressed after each level to speed the level change along, he could have.)
Don't read this paragraph if you want to maintain your sanity :) Adding another mystery to the brew, the level is optimally solved with 336 moves not 333 moves in which both of his attempts try to solve it in (the first one 332 because of the missed Up move). I have played the level with phils tap moves and none of them are double moves, each moves one square. The three missing moves are at the end, and phil does not hold down the last move in either 5-d case. But he does solve the level the second time and he gets credit for 336 moves for the level. This could indicate an error in the analinp output, but i've checked other levels and all are tap moves and solve the problem with out needed to hold down the last move to move more than once.
-- Anonymous, September 19, 2000