Serious proof of recent cheating at marp

greenspun.com : LUSENET : MARP Editors : One Thread

The game soukoban Deluxe (a very lovable game that was never meant to be played timed) has coerced Phil Lamat to cheat to complete all the levels on one credit. There is proof when watching "him" attempt to complete level 5-d that he is using an external keyboard programmed level completer that could be used for any of these puzzle games that don't have variations.

You see "him" or his external program nearly compelte World 5 level d, and then something goes wrong, the input goes out of sync a few characters of input (perhaps he accidentally hit the keyboard while he was piping the input into mame.) after this point you can watch the program attempt to compelte the level out of sync, once the input is finished (the level isn't complete) so phil resets and then probably starts of the input driver to compelte the level with out going out of synch.

He is a great sokoban player (i've been playing sokoban for more than 10 years and i only found 3 levels i had done better than all 55 levels in this game). But as a "soukoban deluxe" mame player this is very seedy indeed. What should be the course of action? How will be able to tell if phil will use this method of mame inputing in the future? what about in the past?

This has the very unlikely possibility of being a -playback -record cheat but the fact that the input goes out of synch at one point and you can watch the player move to compelte the level (but just a few squares out of position) leads me to believe it's not, but some sort of external input driving mechanism where phill can play a level and record the keystrokes, then play mame -record and pass the optimal keyboard input program for each level when it comes up.

oh boy, this is going to be fun, not. i'm glad this board is here though...

-- Anonymous, September 07, 2000

Answers

btw, i've been suspicious of bj learning to play soukoban just a week before he recording a compelte 55 level sweep. this to my knowledge could have only been done with -playback -record or pausing. I'm not the greatest sokoban player in the world but i'm pretty good, and it is very hard to remember the optimal moves on one credit unless you've had some serious study of the problem(s).

And i was obviously suspcisious of phil because "he" (or his program) makes NO mistakes at all except for the 5-d out of synch problem, this is impossible to NOT pause for a few seconds to see where you are in the game once in a while. at least BeeJay made some mistakes, which still could have been -playback -recorded... but acusing beejay is far from what i'd do unless he had a track record like phil.

-- Anonymous, September 07, 2000


I would recommend gathering more evidence on Mr. Lamat and this .inp, and if your suspicions are confirmed, Chad, putting a general vote of expulsion before the MARP community. The vote would state what the charges are (including all evidence gathered against Lamat), and would be a simple yes/no for players at the site, and would also be signed by all 5 editors (not necessarily Zwaxy)

If I'm not mistaken, this is the third major cheating charge leveled at Lamat, including:

1) Splicing League Bowling for a perfect game

2) Deliberately slowing down ccastles despite the rules in T2

3) Preprogramming moves on Soukoban Deluxe

I've had it. I don't want to fight with him any more, I don't want to post anything more about him on the msg board, and I don't want to have to e-mail him about slowdown problems or repetitive pausing. It's time to put the A back in MARP. Phil Lamat is one player we can do without, and a good example to use to set a precedent for cheaters at MARP.

Sincerely, Q.T.Quazar

-- Anonymous, September 07, 2000


Not to mention phil's

4. engineering an autofire mechanism (when he claimed he didn't know what autofiring was) with track and field.

I'm glad you agree with gather more evidence and discussion options before action is taken, i want to tread lightly on this subject at first because it could be just me that is seeing things. maybe someone else has another explanation for the strange inp he submitted, but since i know the game well it's easy for me to come to conclusions.

Perhaps we should have a three strikes rule? and if this is the third strike take a larger action than just dismissing the inp that qualified the strike. I really don't like the idea of banishing someone even if they continue to cheat.

If noone thinks any cheating inp at all is good to have, i'll argue that this particular inp submited by phil even though it was cheating was very pleasing to see someone else play sokoban as well as me (beejay didn't do it justice) (in sokoban there are no real timing issues but in mame soukoban there is) I think there is a place for this inp and others like it, but not to score points and not to go unnoticed in a big batch of "zany"/"cheeky" inps that are in the miscelanous bin. more 2 cents.

-- Anonymous, September 07, 2000


more mistakes, i can accept the idea of banishment but i'd rather try to find a better form of punishment than banishment forever...

-- Anonymous, September 07, 2000

Don't get me wrong--I'm not that adamant about giving the boot either, but I'm seriously out of ideas with Phil. He just keeps showing a flagrant disrespect for the other players at MARP. He cheated in two tournies at MARP, and would have continue to do so if we had not caught him. He cheated on track n' field, and when BenJos caught him, he played Editor against Editor and whined to Pat to get his score reinstated (and the score was indeed bad). Days after 'someone' posted how to splice recordings at MARP, he posts a perfect 300 on League Bowling. Don't even get me started bout his Gun Dealer recording, or the dozens of games we had to delete due to autofire or slowdown, particularly the latter. I'm not sure what we CAN do to make sure Phil plays in a manner fair to everyone else at MARP. We could try imposing a tg3-only limitation on him, but then he'd probably just spend his time working on hacking that. And I don't even think tg3 would catch a preprogrammed keyboard.

Even if skito's charge doesn't hold up, we are going to have to seriously think about what to do with players who insist on not just not following the rules, but not giving a f$%^ about the other players at MARP. I can understand the frustration some of the players here feel at having a score wiped due to speed or NVRam problems--skito does that to me all the time :) But the problem with Phil is a blatant and chronic one. Sooner or later we're going to need a solution.

How many of you trust his scores now as it is? How many of you thought, when he uploaded his Phantom 2 .inp a couple of months ago, "Gee, I wonder how he scammed that?"

Q.T.Quazar, MARP Rules Coordinator

-- Anonymous, September 07, 2000



I am planning on watching the recording tonight. I haven't watched it yet, so I might be missing a point here, but a recording being slightly out of sync does not prove anything, nor does it indicate cheating.

I have seen lots of recordings go out of sync when played back. For instance, this will happen with Bubble Bobble if you record with - ym3812opl and then try to play it back with -noym3812opl. The reason it doesn't look "suspicious" then, is because when a playback is out of sync in a game like this, you won't survive very long. Then you restart the level and you are not anywhere near the position you were when you died, so it doesn't look like you're slightly off anymore.

What I am trying to say that a lot of recordings that don't play back correctly because they get slightly out of sync, will very quickly no longer be slightly off because the player dies or can't get up a ladder or something.

Think of an imaginary game where you can only move left and right and where things are dropping from above all the time. Your task is to catch them, but other than the game keeping track of how many you caught, not catching them does not have any consequences. Now imagine you record a game like this and during that game, you catch everything. Then you go on to play it back, but, for some reason, the playback goes slightly out of sync early on. What you will see then is a playback where the player still makes all the correct moves, but slightly off, so you'll see a lot of near misses.

And going out of sync isn't all that uncommon. It happens quite a lot when trying to play back something using different sound settings than the ones that were used when recording.

Although I do not quite understand what you are talking about, Chad, I have the impression that there is more to it than what I just talked about. I'll know for sure once I've watched the recording. So all I'm trying to say right now, is that a playback being slightly out of sync does not mean the player cheated, and that it could very well be an incompatibility between some of the settings used by the player and those used by the observer.

As for banning him... the first thing I was reminded of is that, some time ago, there were plans and votes and such to ban Stig. Whatever happened to that? Also, if you are going to make it a public vote, then a lot of people probably won't vote unless it's anonymous. And while I like the idea of a "3 strikes and you're out" rule, it needs a lot more clarification and discussion on what exactly would be considered a strike. Sure, there are some clear-cut cases like RLH, and (to a lesser extent) autofire, but outside those, how do you prove that someone did something intentionally? Anyway, I guess this makes another agenda point for next Sunday.

Cheers, Ben Jos.

-- Anonymous, September 07, 2000


The thing about this one is that is it out of sync AND THEN goes back into synch. WHen you play soukoban deluxe you have a key that will reset your level (not reset the clock) but reset your level so you can start fresh in case you made a terrible mistake or in case you want to leach moving points to do a level a couple of times before completing it.

from what i can tell, level 5-d starts out fine, gets out of synch, phil hit's the reset button then procedes to play the level again in the same way. i do not believe this can be explained by a playbackability issue but i could be wrong.

-- Anonymous, September 07, 2000


I can agree with what's wrong with Phil - but as much as I want to get rid of him - I WILL do this in a Robert's Rules fashion.

This means: Any suspension or expulsion(which equals the taking away of a member's rights) will require 2/3 vote.

There will be a fair trial on the message board(which means Phil has the right to defend himself).

This is the only fair way to do it. Trust me: A 50.01 % vote for expulsion(and him getting expelled with that percentage) would just lead to major controversy. At least it would reduce the controversy a little.

I would like to discuss these incidents further on the editor's chat meeting Sunday Night at 10pm if nobody would mind...

Thanks for reading. GB9

-- Anonymous, September 07, 2000


OK, I think I'm slowly beginning to understand what you are saying. I still have to watch the recording, though.

Let's assume you are right about Lamat having used some external keystroke recording program. Could he really have been so stupid to have done what he supposedly did? I mean, he saw that it was out of synch, reset the level and restarted his external program to play back the keys again. Surely he must have known all that was being recorded into the .inp. It's possible to be that stupid, yes. We've seen MvdV do it with his Super Sprint screenshot in T2. And since he had the keystrokes recorded, why couldn't he just have done it ALL over again, but correct this time? Just wondering.

Could it be, instead, that he had all the moves written down on a piece of paper, and was reading from that piece of paper while making the moves, and not looking at the screen? Then, when he finally looked, he saw that things went wrong, hit the game's reset and did it again?

I do not doubt for a minute that Lamat cheated on a lot of games, and I am not trying to defend him here, but I am still not sure what (if anything) all this proves in this case.

You see, today I've been thinking a lot about what it would mean to use a keystroke recorder. And I don't think it would work. There is just too much going on in the computer all the time to make two successive games be identical. All the time, there are tiny variations in speed (regardless of whether autoframeskip or some fixed frameskip is used), and those variations are highly unlikely to occur at the exact same moments the second time. An external program is just that: external. A keystroke produced by it is not in any way connected to a particular frame in the game (unlike the keystrokes recorded in an .inp), and there is no way to force the entire to be in the exact same state as it was when the keystrokes were recorded. Not even under plain vanilla DOS. Somehow, I think a keystroke recorder will not give the "desired" results with MAME. That's just my guess, though. I could be wrong.

Also, the keystroke recorder would be pretty useless if you didn't edit the keystrokes. Because unless you do everything perfect the first time (in which case you wouldn't even have needed it), you'd need something like the re-recording cheat to make it feasible. So you'd want it to play keystrokes back to a certain point and, when at that point, immediately have it switch to recording keystrokes instead. Or, alternatively, you'd have to have two instances running, one playing back, the other recording. In either case, you still have to have a way to cut off some of the end of such a keystroke recording, so that the next time you use it, it stops BEFORE the trouble began the last time, and those final, wrong moves, aren't repeated.

Well, that's it for now. Like I said, I don't think an external keystroke recorder will work, but I could be wrong. I'll watch the recording tonight and then I'll probably have some more to say.

Cheers, Ben Jos.

-- Anonymous, September 07, 2000


yesm i agree it would have been pretty stupid to let that remain in the inp, AND i agree having an external keystroke recording would be very difficult to configure for most games. Especially since we have the analinp program to detect chronic keystrokes (which i tried to see if i could find any chornic 2 times a second keystrokes and didn't find any.) However, Soukoban moves are the same time all the time, you move 2 times every 3s game seconds which translates to a real time amount per move. It would be very easy to configure a macroizer that used direct input to a dos window to send a sequence of keys (developed and honed by a windows non-mame sokoban game).

I do not think for an instant that he looked away from the computer for at least 50 moves (which are done as quick as you can make them) while his soukoban character was playing blind robot. I do think - playback -record is more of a possibility and he just forgot to edit one game out that happened to end on a reset...

I welcome everyone to doubt my findings, and see the recording for yourself, unfortunatley it takes a while to playback, but it can be playedback with out sound to speed up the throttling till you get to the crucile level 5-d. i can't believe i can't spell crucial. no wait, i can beleive it.

-- Anonymous, September 07, 2000



One more piece of info, the keystroke recorder i'm guessing phil used wouldn't record the whole game, i think you could have level macros, with a sequence of keystrokes for each level (this is why 5-d is played twice once with the error in the sequence macro and once with out the error.) still a -playback -record sounds more feasible but that makes errors much easier to circumvent. with a keystroke programmer, you might make many errors and you wouldn't be able to fix them unless you restarted the recording, so perhaps he made only one error and didn't want to redo the whole thing and forgot about it after playing the remainder of games... which would mean he didn't know or want to use the -playback -record. it is difficult to think of a scenario in which phil would have cheated on this recording. but the inp is suspect.

-- Anonymous, September 07, 2000

OK. I've watched it. First of all, if you say that almost all his solutions are as good as, or even better than, yours, that must mean that he always uses those solutions. No improvising. I have trouble believing, though, that anybody has all those solutions memorised. So that could mean he uses a cheat sheet; a piece of paper on which he has written down all the solutions. Although using a cheat sheet is not really "clean", there is no rule against it, and even if there were, there is no way of detecting it.

What I noticed next is that his movements are discrete, not continuous. When he has to do three steps to the right, for instance, you see him do three separate steps (with brief pauses in between), and not one continuous movement stopping when he has moved the correct distance (although the game allows you to do this). At first, this made me think of him looking at a piece of paper and not the screen while doing his moves. Just about all his movements in the entire game are like this. If I were playing, and I had to move in the same direction for a while, I'd just keep my joystick in that direction constantly and not release it until I were close to my intended destination.

But I noticed something else: The lengths of his movements in all four directions as measured in frames by InpHisto is very close to constant: In around 70% of the cases (or even more) this is 9 frames long. 9 frames is 0.15 seconds at 60 fps, which is a nice round figure... . And remember that this not only applies to how long a button is pressed, but also how long it is released in between presses. Can a human be so consistent? I have my doubts. Pressing at high speed and getting a lot of lengths around the 3, 4, and 5 frames lengths is a lot easier than having 70% of your presses and times within presses last between, say, 0.14 and 0.16 seconds. Because shorter than 0.14 would make the length 8, and longer than 0.16 would make it 10. At first, a keystroke recorder didn't fit into this, but neither did a cheat sheet.

Finally a figured out a way that accounts for all the observed facts. I am not saying that he did it this way, but it is the best explanation for the facts (ALL the facts) I can come up with:

First of all, this game is very "forgiving" as far as accuracy is concerned. Although I haven't tested it, I bet there is a relatively large range of how long a press can last to make it still perform one step. For instance, any press that lasts between, say, 5 and 13 frames will make you move one step. So if you have some way of generating the average of that (9 frames) pretty consistently, then it becomes easy. Even an external program, with all its inaccuracies relative to MAME, should be able to do well here. After all, a "huge" (in terms of how fast computers are) error in the time a keystroke lasts will still produce the desired result.

Now, imagine a keystroke recorder that has macros like you suggested, Chad, so that by hitting a few keys followed by a number (or something similar) it will generate a sequence of keystrokes from that particular macro. So you'd have 55 (is that how many levels there are?) different macros that can each be started individually and easily by means of some hotkey followed by a number. This would account for the repetition of that fateful level, and the easy way in which it was restarted. He must have not been watching the screen all the time, or if he was, not have a way to restart the macro before it had played back completely.

This still doesn't account for the almost constant length of all the presses. But now imagine that this recorder has another nice function: The ability to write your own macros. So not recording one, but the ability to play back keystrokes put into some script that was written by the user. It's easy to imagine a script that simply has something like "R R R U U L D L L" etc. in it. And if a utility like that supported scripts like that, then surely it must also have a way to tell it how long each keystroke will last and how long the pauses in between will last. So specify 0.15 seconds for both, and voil`, an easy and convenient way (after finding all solutions of course) to make a "perfect" recording.

The reason you didn't detect how constant the moves were, Chad, is because the presses and releases last too long. Although autofire detection relies on consistency that is assumed to be much more constant than a human can do, it only measures the consistency for "bursts", so when there are lots of presses within a short period of time.

Remember, though, that although my theory accounts for all the facts regarding Lamat's recording, I think it can hardly be considered proof. I think the proof we had to convict Stig was a lot more solid and that (for reasons still unknown to me) didn't happen.

Yes, I believe that Phil cheated, and, yes, I think that my theory might very well be the way he did it, but the theory doesn't prove anything. It might even be seen as an attempt to find a way to incriminate him, despite the facts that I don't think I left out any facts I can't explain.

And even if he didn't cheat, this all shows of a new way of cheating that we can't really catch yet. Although it probably only applies to a small set of games (only games that are very "forgiving" as far as inaccuracies are concerned), it definitely is something to think about. I don't think there is any way in ANSI C to list all the processes that are running. If there is, though, then a version of MAME could be written that stores the names of all the currently running programs in the .inp. Now THAT, I'd say, would be sufficient proof. I can do this in Windows, so I could write a MAME32 version that does this, but I don't know of any way of doing it in ANSI C / DJGPP / DOS MAME, other, perhaps, than issuing a "MEM /c" command from within MAME and capturing the output...

Cheers, Ben Jos.

-- Anonymous, September 07, 2000


I notcied the skip step as well, there is absolutley NO reason to do this for EVERY contiguous directional move unless you are using a program to do it. I looked at beejays recording and there is no place at all where beejay taps in the same direction once each move, that is ridiculous. (I had thought this was a sideeffect of throttling and using -soundcard 0 but now i'm sure it's another sideeffect of phils magical program(s))

I looked at inpHist as well and couldn't find any constant frame bits, which bits did you notice were =9?

These are two more affirmations that phil is using a macro program. AND that we can tell he's doing it by the constant 9 frame presses. Now i'm much more convinced we need to do something.

The mem /C (ps augx) option would be beautiful to put in the new mametg! the inp files are already megs a pop, so another 10k if that of processies running output is no big deal at all as long as it's encrypted, well hmm encrypting that would be very difficult unless the encryption is seeded with a timestamp like the rest of the inp should be. we could tell mark this for him to ignore, but we should tell him anyway. this could also catch slowdown programs that might be runnig while mame is playing, but i guess if it's tgmame slow down should be caught by the frame rate counter already but more info the better.

-- Anonymous, September 07, 2000


All the direction bits were length 9 70% of the time. However, this percentage only applies to lengths 10 and less. I do not let lengths of 11 and more count in this percentage (and the same goes for average lengths), because if I let all the lengths count, then this biases the percentage way too much. For instance, when someone inserts a coin, it starts with a long run of the same value for the coin bit, then a short run while the button is pressed, then a very long run while the button is released, probably lasting for the rest of the recording. So you'd get a run of, say, length 50,000 and you don't really want to count that.

Here are all the relative bits for Soukoban Deluxe: Bit 8: P1 start, 0 means pressed. Bit 11: P1 button 1, 0 means pressed. Bit 12: P1 up, 0 means pressed. Bit 13: P1 down, 0 means pressed. Bit 14: P1 left, 0 means pressed. Bit 15: P1 right, 0 means pressed. Bit 59: Coin 1, 0 means pressed.

Now since a large number of lengths were more than 10 in Phil's recording, it painted a picture that was a bit biased. Lengths 9 accounted for 70% of the lengths 1-10, and lengths 9 and 10 combined for over 90%. But since a large number was over 10, I decided to have a further look at those as well. Then, it doesn't look nearly as suspicious as before.

In addition, I also made InpHisto output two more histograms for each bit, viz. for when that bit is 0, and for when that bit is 1. For instance, it turns out that in Phil's recording, of the 639 times that a bit-13-run has a length of 9, 638 times this bit is 0, and only 1 time it is 1. Since 0 means pressed for this bit, that means that he had 638 PRESSES of length 9, but only one time that the time in BETWEEN presses lasted 9 frames. I'd show some of the output here, but I don't think it would show up too nicely on this board. Anyway, it turns out that most (usually 90% or more) of his PRESSES (so when the bit is 0) are within the 9 through 12 framelengths length, with 9 usually occurring more than 10, 11, and 12, but not nearly as pronounced as it looked before.

Since this case shows that lengths over 10 can be important too, I have now changed it so it displays (and takes into account when doing percentages and averages) lengths 11 through 14 as well. While at it, I also made the autofire detection part of Analinp a bit faster. I can either send you the latest version now, or a little later, because I still see room for significant improvement in speed of the autodetection part of Analinp.

Cheers, Ben Jos.

-- Anonymous, September 08, 2000


I think 90% of frames between 3-6 is not suspcisous but 90% of frames between 9-12 is. but do you think it's suspicious?

Great work on speeding up analinp, the damn thing is so fast already compared to earlier versions... send me a copy when you're done no rush. i trust you more than me analizing beta inphist output, so the more you can refine it the better.

-- Anonymous, September 08, 2000



Yes, I think it's suspicious. And a lot more so, coming from Lamat. But I don't think it's enough to convict someone. You know... during T3, which was the first tournament we used TG3, I analysed every single upload Lamat made, and not just with "Checker". I have a lot more knowledge than what is displayed by Checker, and even then, I couldn't catch Lamat. Maybe our main problem is that we don't think like he does. We observe things and try to find irregularities and then prove them. Or we try to hack some version of MAME. But we are not into cheating. We don't try to come up with new ways to cheat. Like they say, it takes a thief to catch a thief... . So far it's been the neverending battle between the locksmith and the person who picks locks. Or alarm systems and burglars finding ways around it. Maybe we should start specialising on cheating instead. :-)

Cheers, Ben Jos.

-- Anonymous, September 09, 2000


If you don't think it's suspcious enough to convict, what is the "reasonable doubt" that you have that the recording does NOT display cheating techniques? I can't explain it any other way, he couldn't have been looking away for 50 moves... i.e. i don't have a reasonable doubt. I definitley am not going to push anything if i'm the only one though and i'd like to be convinced otherwise.

-- Anonymous, September 09, 2000

This is a very interesting technique and 100% feasible. With Lamat being a Math Professor, I'm sure he is more than capable of obtaining a program or writing one himself to perform this task. You guys keep talking about proof, proof, proof--we need proof...

Well here it is: this is what we do. We analyse the INP carefully by looking at the level where he screwed up and then corrected himself. We isolate the keystrokes involved by defining a begin and end point for each of the two sequences. We dump the keystrokes (forget about timing issues) into seperate text files. We then run a comparision program between the two files. If they're identical---BAM he's nailed.

-- Anonymous, September 11, 2000


good idea, they don't have to be identical, since i first suspected he accidentally hit the keyboard while the macro program was running, so there could be several "additional" strokes to the optimal solution that failed to compelte the level compared to macro that completes the level. Let's call who is going to get the keystrokes on this inp, Ben Jos are you interested in laying out the keystrokes for the offending level? If not i'll do it. This will take a ton of time trying to isolate because the recording playsback so slowly... however you can get the midpoint of the level with the "B" key as it was probably pressed once during the whole game [ after the offending macro messed up and right before the "good" macro completes the level].

-- Anonymous, September 11, 2000

Actually, it's not THAT hard to isolate them. But even if the two sequences are identical, does that prove he did what we think he did? I think it only proves he always uses the same solution for that level, because I still am not too sure why he couldn't have been looking at his cheat sheet all the time and played blindly.

Anyway, I guess the more we analyse his recording, the more answers we'll get.

Cheers, Ben Jos.

-- Anonymous, September 12, 2000


I guess i'll handle the analyzing of this inp. I agree that identical sequences does not prove anything, but identical sequences (with a mistake thrown in) plus the fact that the frame distances between moves is similar, that is enough to convict him of a macro program.

If the frames are not the same in the two levels then it still could be a look away scheme, but to play this game in look away mode for 55 levels and MAKE ONE MISTAKE is to have sold your soul to the devil for hand coordination.

-- Anonymous, September 12, 2000


I've isolated the keystrokes for the attempted 5-d completion and the reset and then the final 5-d completion.

The moves in each of the attempts is identical save for one "less" keystroke in the failed attempt. I suspected the timings between each of the moves would be identical, they are not. However this recording is still very suspcicious because the timeings for nearly all moves are 20~ frames held down 10~ frames released. this is something a human would never do. To prove that we'd have to get Ben Jos to determine a human couldn't keep that kind of timing. The second thing still suspcious about the recording is that if a human was doing this you would have to look away from nearly half of level play. I've just tried this (wrote down an optimal level solution, turned off the monitor and attempted to complete the level), i was able to make all the moves correctly (no missing, no doubles) but the soukoban man didn't complete the level because i didn't use the right timing of my moves. I seriously doubt anyone can play a complexe soukoband level with the monitor turned off or looking away for more than a few moves. But even with this there is something else i noticed.

The after the level is reset (phil notices the level is wrong then attempts to complete it) only a few seconds pass before phil hits the reset button and starts chuging away at 20frames down 10 frames up. If only seconds pass after he notices the level was wrong, I find it again hard to believe he finds the mistake he made adds the extra move to the right place on his cheat sheet and continues to play another level with out any hitch all with in a few seconds. It is possible that he simply should have hit an extra move and it was already on the cheat sheet and he just decided to play again, but then why was there only 1 mistake in 55 levels, ~5500 moves, one mistake.

the new version of analinp will allow everyone to study keystrokes in games like this. analinp -k -h inp.inp for keyhistory.

-- Anonymous, September 14, 2000


Very interesting thread to read.

Chad - I can safely assure you that I learnt all the levels of this game in the 1-2 weeks it took me to get all 55 levels completed. While I had never played Soukoban previously, I had played some similar games previously and it was a lot like the YamYam/Wiseguy games with respect to logic so that helped - perhaps these games would be worthy of further analinp analyses as well as they have similar timing allowances etc.

The main reason why I didn't do the game "justice" by "maximum optimising" of all the levels was that I only optimised them as far as necessary to get the "best steps" bonus. Further optimisation was pointless from the point of view of maximising the final score.

Cheers, BeeJay.

-- Anonymous, April 14, 2001


btw, This issue never really got around to anything. Since we couldn't really figure out what happened with phils recording. I think the option to record the synchronous key states is still in analinp, but even with that analinp couldn't prove in this case if a non-human mechanical means pressed the keys. I'm wondering what your take on this is beejay? Have you seen Phils 5-d recording yet? Does it seem strange to you that the entire level was completed with a single error in the middle of input, then the reset button was pressed and the level was done again with the error fixed?

-- Anonymous, April 14, 2001

I have now taken the time to watch this recording and I definitely agree that Phil was obviously not playing this game in the usual manner or he would have picked up his mistake sooner. He most certainly was not watching the screen while playing or playing it himself or he would have reset the level almost immediately upon realising his mistake.

I further think it is likely that he was using some form of keyboard macro(s) for his input as even using a cheat sheet he would perform more regular checking of his position on screen. I also believe he would have had to have allowed longer between his moves if he was doing this blind to ensure he didn't input the next direction prior to the Soukoban character completing his previous move and we would therefore have seen some more noticeable pauses throughout all the levels of his game.

That said, there is no reason why someone could not remember all the moves required for this game and be able to play it at near full speed without pauses or mistakes in the game. I probably would have played a flawless game, although not necessarily absolute optimal level completion technique, given a few more games as the levels while they seem complex when watching are not that hard to remember at playtime - well not to me anyway.

I had played Boxy Boy on the PC Engine sometime prior to this and as the clone of this game recently added is called Boxy Boy I guess I had really played this game previously so I wasn't as quick as I thought to learn the levels. :(

No wonder it seemed similar to other games I'd played previously. ;-

As to what to do with this.. well I'm a new Editor but this score definitely needs to be DQ'd if nothing else - but perhaps wait until Ben Jos has MameTE completed such that he can list (encrypted) all the other processes that were running that could possibly be a keyboard macro type program.

Cheers, BeeJay.

PS: I did download the NBA Jam recordings to see if I could get them to playback but unfortunately I don't have the Roms at home currently. I wish the MAME Team would hurry up and release 037Final so I can cut the roms onto CD and look at some of the new release games.... like Do Don Pachi eh BBH ?! ;-

-- Anonymous, April 17, 2001


Yeah, I agree with you, something is strange about that recording, it's hard to say anything yet though... I think the best thing to do is wait for TE and see if phil will record again and we can check out what other programs were running.

-- Anonymous, April 17, 2001

Moderation questions? read the FAQ