Just a quick update. A run has now been performed for the Fun Levels on SNES, that is the fastest recorded run of the first 30 levels of the game thus far, beating out the record set by Usumgallu on MS-DOS 5 months ago. It's pretty impressive given it was also done on original SNES hardware. The new time for the Fun Levels (especially the ones that feature the full 30 levels) across platforms is 41:32, only 47 seconds faster than the previous MS-DOS run.

Also the first run for the Genesis version of the game (the Japanese release) was submitted a couple weeks ago as well. So for now the fastest run on Genesis is exactly 10 minutes slower than the new SNES run.

https://www.speedrun.com/lemmings/run/z033wloz

Quote from: 607 on April 15, 2018, 12:41:19 PM
... You were using a glitch for your runs!!
I mean, that's no problem, but I never realised! Now I get why some of your times seemed unbeatable to me... Wow.

Ah, I thought you were too, but I see now you were just pausing.

Yes, basically, when the level starts, the IGT starts and the pre-open music starts. However, if you pause immediately, the IGT stops, but the pre-open music continues. Once you hear the door opening sound (but the door hasn't opened yet), you can unpause the game. The door immediately then opens, and IGT resumes. So it saves 2-3 seconds on every level (if you are playing on MS-DOS). It's unclear at the time what other ports have something similar, but SNES definitely has the option to leave the level early, and as long as you get the necessary lemmings, the next level will continue.

Quote from: ccexplore on April 11, 2018, 12:58:30 AM
I haven't looked at this in ages.  Are the current MS-DOS version solutions submitted for this thing even anywhere close to optimized?  I suspect many probably aren't, despite efforts that have already been made a long time ago on the challenge threads here (but never yet translated into the form of a video or similar that conforms with the submission requirements used here).

Oh, definitely not even close, I wouldn't think. Although I'm sure Fun in of itself is probably not too far from an the actual optimized time.

QuoteAlso, are we measuring based on game time or real time?

For the purposes of running multiple levels, real-time takes precedence, so that levels have to be done in quick secession. For the individual levels, the IGT timer is more important. This is sort of important for MS-DOS because things like pausing and pause glitches and such for are useful for stopping the IGT, whereas it would be particularly slow over a series of long levels to have to do a pause glitch every level (and going slower in real time) in order to make the IGT work better. I haven't tested all the consoles yet, so it's not yet clear what sort of things like this can or cannot be exploitable.

QuoteEither way, there are difference in the two ports in terms of how many physics update occur in a game second, and also how long in real time it takes between physics updates.  As a result, the exact same solution involving the exact same pixel-level movement of the lemmings, will still take slightly different amount of time in each port, whether or not you are measuring in game time or real time.  As a result that will make timings from the two ports not directly comparable, until you adjust for those differences accordingly.

Oh definitely. The leader-boards are separately by platform because the number of ports and their minute differences are so numerous and often tiny but still important. That's actually kind of why I was shocked the SNES time for Fun could be done that quickly, given that I assume the console version runs slower.

When emulating MS-DOS, increasing the processing speed allotted to the program doesn't really affect in-game play. The internal timer for the levels keeps the refresh rate locked. However, it does speed up loading screens so that the transition between levels is quicker. This however is limited, so there is a limit to how quickly loading screens can move, which is naturally important for keeping the playing field equal for MS-DOS users. I suspect this allows one to get through loading screens faster than a SNES would, and that sort of stuff can add up.

If more runs get added soon on different consoles, could be fun to take the current records and put them in the same video for comparison.

Interest is starting to pick-up.

One player got a 1:13:54 on the Fun Levels from a SNES emulated run 8 days ago.

Today a Japanese player was able to finish the Fun Levels on the SNES 44:04, which 1:19 away from being faster than the MS-DOS version.

Quote from: Forestidia86 on March 02, 2018, 06:27:14 PM
One can actually beat Fun 2: "Only floaters can survive this" (Dos) in 16 s IGT if you use pause glitch and "High Perfomance PC" with Esc early enough.
Just hit Esc when the timer is still at 4:44 (provided you have used pause glitch optimally).

Got to beat 607 to it when I get home.

I'm thinking Level 1 can be done in 16 seconds too honestly.

I imagine in most instances shaving off that last second will always come down to perfect execution of ESC and pause glitch.


Any tips on level 4? I got :36, he's done to :33. As far as I can determine, you got to get the lemmings that aren't digging facing the right direction at the right time. I was thinking about at work, maybe possibly setting the speed to 1, using the first lemming to dig, and seeing if he completes it fast enough for the next lemming not to get stuck in the top part.

So, after messing around with the backend of the leaderboard enough; it just made sense to allow DOS players to use whatever functions is necessary to get a better time. While means, basically allowing ESC.


Interesting thing 607 noticed on the runs. If you set the DOS settings to "High Performance PC" than the fade out of the level post hitting ESC takes longer.

Basically, whenever you hit ESC, the time immediately stops, but level will continue. Basically, if you need one lemming to beat the level, you can hit ESC a couple seconds before the lemming hits the exit. This freezes the IGT, and the fade out for the level begins. If the lemming makes it to the goal during the fade out, then the game will count it. So using ESC not only goes faster for any given level for obvious reasons, but it also goes slightly faster on all levels with 100% needed completion rate, because you hit ESC a couple seconds before the last lemmings enter, giving them just enough time to get in before the level ends. DOSBox cycles don't appear to affect the level fade out.

Curious if anyone had an idea why "High Performance PC" would cause the fade out to last longer?


Also, would y'all like or mind if put a link the discord on the speedrun forum?

Thx.

Added the Windows '95 Levels. Those can be ran now (at least the lemmings levels, not the oh no! more levels).


Btw, is there a Lemmings discord?

Quote from: Forestidia86 on February 25, 2018, 06:25:45 PM

Quote from: Nihilist Comedy Hour on February 25, 2018, 05:39:41 PM
Basically to get that one extra second, you got to start digging with the right lemming in the middle. But then you got to start digging with lemmings in front of that one. It prevents any of the lemmings towards to back to end behind the original digger once they get to the bottom.

Cool, this strategy works out really well and seems reasonably safe. But unfortunately you lose one second with the pause glitch. You can hold the timer when it is still 3:00.

Interesting. My noise was being wonky for some reason. (I think the increased cycle speeds interferes with the audio, but need to look more into it). But it's hard to pull out the glitch with the audio cue to know the earliest spot to do it (and if you've done successfully. It seemed like I was able to pause until immediately once the igt hit 2:59, but I wasn't sure.

Quote from: Forestidia86 on February 23, 2018, 03:03:15 PM
For safety Proxima's method to just assign from left to right seems to be best from what I encountered in my attempts. It still should generate a time of having 2:11 left. It's possible to solve the level faster as ccexplore indicated but I still have no reliable method to do that. Maybe someone else can do it/knows how to.

Edit: With pause glitch (pausing shortly after the level starts before the hatch opens) and increasing spawn interval while pause the Proxima method even has 2:14 left. The question is whether one should really take the ingame time.

(With Pause Glitch) I was able to complete with 2:15 left.


Basically to get that one extra second, you got to start digging with the right lemming in the middle. But then you got to start digging with lemmings in front of that one. It prevents any of the lemmings towards to back to end behind the original digger once they get to the bottom.

https://youtu.be/aOGb_l2ltoY

Did a little testing this morning and it does appear that DOSBox cycle speed does affect the loading times. On the bright side, it has an upper limit. I'm not actually sure where it is, but it appears around 12000 cycles the loading speed tops out. This also affects the intro screen credits scrolling, which is a good way to measure if you have enough cycles to max out the loading speed.

Quote from: Proxima on February 25, 2018, 04:36:06 AM
The pause glitch is that (on DOS Lemmings; don't know about other versions off-hand) if you pause just when the level starts, then unpause, the hatch opens immediately, saving two seconds of in-game time.

Oh, gotcha. That's uh... pretty interesting.

It is interesting question as to igt/rta I haven't quite determined yet. I'm tempted to say igt is more important on levels, and rta more important on multiple levels. But I intend on taking into account input from runners.

Quote from: Forestidia86 on February 23, 2018, 03:03:15 PM
For safety Proxima's method to just assign from left to right seems to be best from what I encountered in my attempts. It still should generate a time of having 2:11 left. It's possible to solve the level faster as ccexplore indicated but I still have no reliable method to do that. Maybe someone else can do it/knows how to.

Edit: With pause glitch (pausing shortly after the level starts before the hatch opens) and increasing spawn interval while pause the Proxima method even has 2:14 left. The question is whether one should really take the ingame time.

Is that a really a glitch though? I assumed that was an innate function; pausing and then being able to use the F1-F12 functions while paused?

Quote from: ccexplore on February 23, 2018, 10:29:48 AM

Quote from: Nihilist Comedy Hour on February 22, 2018, 02:21:28 PMYes, that's a good point I had forgot about. I'm not sure it would actually allow you complete the level faster

It totally would.  If the last lemming out has to dig, he would definitely be the last one to exit, and would always be slower to exit than any solution where he doesn't have to dig, since digging down the full thickness of the floor is slower than falling onto a partially-dug floor and then falling off from that.

So the ideal fastest solution would have the last lemming out survive falling off a partially dug floor without digging himself, while also not overtaking the last digger used (otherwise that last digger becomes the last one to exit, behind the last lemming out).  So to avoid the overtaking part, you want to find the lemming farthest ahead of the last lemming out, whose digging would not completely dig through before the last lemming out catches up.  (For ports where dig-canceling works, of course you would just exploit that way ahead in the front of the line and be done with it.)

Makes perfect sense to me. Just got to find out how many lemmings from the last one that is exactly; do that first, and then basically spend the rest of the level saving the rest, and that's pretty much... optimized, essentially.

Quote from: ccexplore on February 22, 2018, 08:30:30 AM

Quote from: Nihilist Comedy Hour on February 21, 2018, 06:33:04 PMFor example "We All Fall Down" is pretty straightforward. Basically, the only way of optimizing the level is basically setting release to 99 as fast as possible. That pushes the last necessary Lemming out the fastest. However, to turn all those lemmings into diggers would basically require to click on every one of them while their window for clicking will basically be reduced down to a single pixel. You'd pretty much have to move the cursor slightly to the left by one pixel, and then timed each lemming movings along the cursor, turning them into diggers on their exact pixel.

Obviously, a human probably can't do this under any conditions absent the use of tools.

Yes, that's a good point I had forgot about. I'm not sure it would actually allow you complete the level faster, but it would probably be easier to say, click on every other lemming than every single one. And it's four whole pixels? It seems so smaller.

I think though, that if you were to use a digger say before, the last 2 lemmings, that would allow the 2 last lemmings to pass the digger, making the third the last lemming the furthest from the exit. I'd have to think about it for a moment, but seems like that would cause level to end a second earlier (assuming no escape).

I didn't know about dig cancel for other ports though. That's pretty cool. Seems like that might make a few levels easier, seeing how you wouldn't have to expend as many builders to cancel out a dig.