Hi TheBlazingmo,
Thanks for your answer, but I am still not convinced that character selection is enough to account for the difference in lap times.
First, Spikestuff showed that your movie can be significantly improved without much effort, and brought the solution much closer to the RTA one, even with the difference in character speed.
Second, as pointed out above, your time for 'Studio Amigo' is 22 seconds slower against RTA in this 90 seconds-long race. This is ~25% of the total running time, which implies that Shadow is 25% slower than Big overall. I'm not savvy on this game to tell whether this actually true, but I think some hard evidence towards this should be presented to justify the difference in time. Furthermore, this being a TAS, one would expect that some of the 'character' disadvantage can be offset by the ability to polish every turn much better than an RTA player can.
I also still have concerns about the difficulty selection. I don't think a strong enough point has been made to justify the choice of 'easy' mode against 'hard'. Does rubberbanding slow you down or boosts the CPU players? If the latter, then why is it a problem still?
Bottom line is, although this submission is vastly superior to the previous one, it still falls short to the expectations of a relatively well-polished TAS. I would suggest you to do another optimization pass, and provide a time table showing your times for all races.
Watching the movie and reading the posts above makes it clear to me that the '100%' does not apply, as least given the jurisprudence in the site. One might expect in such category that all (possible) lemmings are saved, which is clearly not this case.
I believe that the category "all levels", or similarly "no skips" apply much better here, since it is clearly the goal stated and accomplished by the authors.
Awesome movie, btw. I vote yes.
Hi TheBlazingmo
I am taking a look at the speedrun.com's individual level record page:
https://www.speedrun.com/ssasr_ds/levels?h=Glitched&x=9d8g537k
It seems that many of these records are faster than the ones in your submission. Looking for example at 'Studio Amigo' (@2:45 in your encode), their time is 1:31.566, while in your movie this is 1:52.41.
Am I missing something here? Am I looking at the wrong speedrun.com page?
Hi Technickle,
Seems like you missed an optimization at the end. If you don't press any buttons, the credits will roll on their own. This means you can finish the last movie much earlier by releasing the gas before the end. Remember in tasvideos.org, we consider the last input as movie end (although exceptions might be considered).
Here is the .bk2 file I produced to show this. I'm pretty sure you can improve it further.
https://tasvideos.org/UserFiles/Info/638455090636489203
I will not claim co-authorship for this improvement.
We have tried to sync this movie unsuccessfully, even after trying different configurations.
SJ, when possible, please provide a .zip file with the contents of your 'User' folder in Dolphin. We need this to make sure we have the exact same settings as you had when you encoded the movie.
I think the game choice is good but have some questions.
1) Where is the hard difficulty option in the menu? Could not find it in game or videos for a Hard mode.
2) Why are the axis inverted and auto Leveling settings changed?
3) Is there significance to have 89% sensitivity instead of 100%?
4) Looks like a lot was played with a controller. How much of the game uses tas inputs?
Noticed that part with the van in the first level, how you go up pretty close to the enemies to shoot them, then slowly back up towards the door while shooting them. I tested sniping headshots all the way near the door that opens after you kill all of them and that works. I would expect activating the trigger that causes the van, then running towards the door, quickly getting the headshots from right next to it would be best. Already being where you need to be for the next trigger can help a lot.
A lot of movement also seems to be not quite taking the shortest path from point to point, but instead using curving arcs. What might help is to immediately after taking a turn is to then find the right angle immediately towards the next goal and use that. Try not to adjust the angle on the way.
Hi Lars_Hendrick, could you please address the points made by CoolHandMike here?
Hi SJ,
I was enjoying the movie when suddenly it started desyncing. It happened around input ~680000 at one of the roller skate challenges. The situation at input 756008 looks like this:
I am using Dolphin 5.0-16793 and this ROM whose hash coincides with that of the movie. Any hints of what could be going on? Did I miss any configuration?
Edit: adding portable.txt in Dolphin's folder and disabling Dual Core emulation didnt help. The desync happens in the same place nevertheless
I made a mistake, frame 5 is when it should appear, sometimes it takes 4 frames for me though. I'm not sure why. If it's still desyncing and you have the right version then I'm not sure what's going on.
We need to delve deeper into what's happening. Can you please tell me the following?
- Executable MD5 hash
+ For this please run the 'md5sum Mobility' command in the same folder where the game is
- The contents of the 'assets' folder. Let's make sure we got the same config
- Operating System version
+ For this please run the 'uname -a' command in the command line
- Are you using a virtual machine (e.g., VMWare)?
- Screen resolution
No need to do all this anymore, feos was able to sync your movie.
I made a mistake, frame 5 is when it should appear, sometimes it takes 4 frames for me though. I'm not sure why. If it's still desyncing and you have the right version then I'm not sure what's going on.
We need to delve deeper into what's happening. Can you please tell me the following?
- Executable MD5 hash
+ For this please run the 'md5sum Mobility' command in the same folder where the game is
- The contents of the 'assets' folder. Let's make sure we got the same config
- Operating System version
+ For this please run the 'uname -a' command in the command line
- Are you using a virtual machine (e.g., VMWare)?
- Screen resolution
When I was making the TAS, I noticed that sometimes the game would take 2 frames to start up instead of 3 frames, and that the TAS would always desync if it took 2 frames to start. Try making sure that the main menu screen appears on the third frame of the movie, and then see if it syncs.
I keep trying and the movie desyncs every time. Perhaps it's taking more frames to load on my side. For me, the first time the main menu screen is visible is frame 5, exactly where you press 'space' for the first time (see picture). Is it the same for you?
Hi Merl_
I am afraid I'm having problems syncing the movie. First and foremost, I'd like to see whether we have the same version of the game. Can you please get me the MD5 hash of the executable on your side? Mine is:
Hi SFan,
Seems like CoolHandMike found some improvements that could perhaps be resync'ed to the end with further effort sync to the end. Would you like to keep trying to improve it further or shall his file be the new submission file? In the latter case, he should be credited as co-author.
Hi Merl_
I am puzzled by your choice of game. You chose to TAS Cyclo 8 version 0.5 (an early beta) whereas a later (release, apparently) version not only exists but also a movie of it ([5728] PICO8 Cyclo 8 by Merl_, rythin & hiii in 01:35.87) by you and others has been published
TASVideos.org rules state that release version submissions obsolete that of earlier versions. Given the only track in this version is highly similar to that of track 2 in the released version, there is also not enough differences between the two versions to consider them different games altogether.
Is there anything I am missing here?
Done, I made a movie of the Hard route. Like I say in the description, I'm using a hack that fixes the dpad controls to be in line with Solstice/Snake Rattle n Roll/other isometric games, I just literally can't play it with default controls, it breaks my brain. Just shoot me a DM if you want the patch for the hack.
Hey, many thanks! I believe the patch will be needed to reproduce the movie. Could you please upload it somewhere or where it is linked?
Airball, an unreleased isometric platformer, kinda like Solstice but as a ball. There are a couple of prototype ROMs but I ran on the the Retrozone version ROM, a 2007 reproduction, but it seems to be functionally the same as Proto 1. Overall goal is to collect 7 items and bring them back to the starting room. Get the Spellbook and then you can start turning in the other items. You can only hold 1 item at a time, so the route is travel to the items and deathwarp back, avoiding any airpumps as they act as checkpoints. For whatever reason after a game over, the Cross item will move from it's set location, to the room you game over in. So that is used to avoid having to go pick it up.
Three difficulties, the items all in different spots in each one. Medium and Hard you can die falling from too high and from overfilling the ball with air, Hard you die in 1 hit to damaging obstacles. Hitting Select in Medium or Hard toggles on super speed, I would assume a TAS keeps it on the whole time. Happy to share my maps if anyone plans to make a TAS.
RTA runs
Easy - https://www.youtube.com/watch?v=a0lgxf5Oie0
Medium - https://www.youtube.com/watch?v=LIA2upiawqY
Hard - https://www.youtube.com/watch?v=rIlEJNVKrP8
Hey, thanks for bringing this up. Looks like a real cool game to TAS so I'm gonna put it in my bucket list. One thing you could do as RTA runner is to record a movie of a complete Hard run and upload it as user file. It doesn't have to perfect; the goal is to have it as a base line to build the TAS upon. It's much better than reverse engineering the route from the YouTube video.
Is there any accuracy that gets sacrificed from QuickNES? Or is this simply an improvement to the performance?
There is no accuracy sacrificed, at least as far as these tests show. I tried my best to keep the emulation intact and only improve performance. Adding more games and tests should help in making sure these optimizations don't diverge from the original emulator.
A small caveat for the benchmark is that I use partial save states for QuickerNES (e.g. for some games I do not preserve the nametable and sprite memory). This helps performance a bit in re-recording for botting purposes but wouldn't be useful for casual gameplay.
QuickerNES project is an attempt at making QuickNES even faster on modern x86-64 systems for the purposes of botting. Nevertheless, it preserves all functionality of QuickNES so it could be adopted into other projects. The main features are:
Core re-record performance improvement of 20~68% across 20 games (see benchmark results)
Implements all 54 mappers in the libretro fork (+mapper 70, which is a copy of mapper 154)
Test set of 28 movies that replicate QuickNES final state hash
Hi alex
Just checked the video. Seems like eXoDOS is a game collection and not an emulator per se. Instead it uses DOSBox to run the games.
If anything, any project that adds TAS capabilities to DOSbox would also benefit eXoDOS. However, I don't believe any such project exists. If you need load/save state functionality, I can recommend DOSbox-x:
https://dosbox-x.com/
For actual DOS tasing, the libtas+pcem toolkit is much more mature and readily available:
https://tasvideos.org/EmulatorResources/PCem