An absolute classic of the late 90's. I used to play this for hours against my siblings and friends, with some epic battles and nail-biting endings. I also had a lot of fun playing the single-player mode, featuring several modes: quick match against the CPU, and three campaign types: missions, skills training, and deathmatch.
This movie targets the missions campaign. It contains 33 missions, each with varying goals (e.g., kill AI worms, grab a weapon box, destroy weapon box) and difficulty. Beating a mission grants you access to the next one. You get a gold medal if you beat it in your first try. However, if you fail it, then some missions become easier (e.g., you get more weapons) but you can only get a silver or bronze medal. In some cases, it is faster to quickly fail a mission to get the silver/bronze handicaps. However, I personally dislike the idea of failing to get an easier path, so I decided early on to go for the "all golds" objective.
From the very beginning of this effort, the 2D Worms community helped and supported me greatly. Since this game allows you to create a new team and name the worms therein, I decided to give them the names of some of those community members, and then some of my favorite TASers (all of whom gave me permission to use their names). Unfortunately, the last two worms are never used in the missions, but nevertheless, this is an homage to them.
The game code, long abandoned by the original developers, has been given to Deadcode, a prolific community member with technical skills to maintain the code. Since then, he has implemented many quality of life improvements, including the ability of saving replays, and a TASing tool to create replays frame-by-frame.
I based my solutions GREATLY on well-established "round-time" community records. In the table you will see two categories: "Unassisted speedruns" and "Tool-Assisted speedruns". The former refers to replays from individual-level RTA records, and; the latter, refers to TAS-based replays. Naturally, the tool assisted times are faster than the RTA ones.
This movie places itself in the middle between the RTA and TAS records. The reason is that Deadcode's tool allows for precise 50fps inputs, including repeated actions that can't be replicated natively. Using Bizhawk+Win98, however, some frames are skipped for (probably) emulation reasons, where I cannot possibly get 50fps. I tried many configurations, but couldn't succeed. The image below shows the three approaches to speedrunning this game, and the current running times for each.
This effort took me many many many hours and re-records, with some moments where I didn't think it was even possible. Nevertheless, I ended up succeeding and even got three new TAS WR that improved even the TASes produced with Deadcode's tool!
Atlas Encode
The native encode suffers from some drawbacks: it's 1024x768, there are long loading times, there's screen tearing, and the camera sometimes misses the action (to follow the current worm actions). For somebody wanting to watch this in HQ, with full map view, and without any waiting, here's an encode.
I was able to do this by extracting the replays out of Bizhawk by exporting the HDD and opening it with 7z. I then took the replay files out of there and played them natively in my PC with a higher resolution, while recording it with OBS.
Acknowledgements
I'd like to thank the 2D worms speedrunning community for their encouragement, help and feedback. I wouldn't have been able to finish this without their help. Special shoutouts to: Mablak, RuffledBricks, TobyTrigger, Mr. Tophat, charles, korydex, Deadcode, FoxHound.
7) Load the XML file as ROM and run the installation movie. Export the hard disk at the end of the movie.
8) Download the movie's XML file, replace the paths as required, and load it as ROM
9) Download the submission movie and play it.
Q&A
Charles from the Worms 2D community asked me a series of questions. Maybe they will address some that may come from this community, so I'm including the Q&A here.
setup - if you'd like to use the original soundbank, pick the "English" soundbank, or figure out how to change locale/region to United Kingdom, instead of United States (latter will also affect the enemy)
Lol, had no idae
setup - Worms 7 and 8 never appear in any of the missions, you can get away with not naming them, unless you're doing that for entertainment/shoutouts
Yeah, these are just shoutouts to some of my favourite TASers
setup - if file manipulating is allowed before the TAS, you can modify the team file WG.WGT to customize the team beforehand OR make a new language file where you can override worm names on setup. there's even an external program for editing the team file. might save some frames idk
I didn't know this was possible, but still prefer to have them typed, as it is part of the shoutout. TASers nowadays don't care much non-gameplay lost frames, if it brings some entertaining/meaning
general - try to use fadeskips, hit ALT+F4 during the fadeout to black to exit out of the game earlier
I tried this. It ALT+F4 doesn't have an effect if you don't play windowed, which is not the idea of the TAS
#3 - have you considered the frame perfect variant of the RTA strategy? there is one for silver, and one for gold. unless your AI manipulations are faster
I haven't tried another strat. Maybe this could have resulted in an improvement
#5 - use Skip Go right after you collect the crate, this'll start the ending sequence quicker. this might apply for every mission where collecting a crate is the win condition
Shame I missed this one. I did apply it on one of the last missions, after RuffledBricks told me about it
#6 - brilliant rope knock there, also done with RB, so poetic :D
his sacrifice won't be forgotten :)
#7 - yea, girder placement... refer to the venn diagram
Yep
#11 - THE 1ST TURN MINE MANIP LMAO THAT'S A NEW ONE
I took this idea from the tool-assisted movie, so not my idea :P but here I demonstrate it is humanly possible :)
#13 - is there any chance of making Worm 2 stay alive while not being targeted by the Enemy, skipping one drowning animation?
I am pretty sure there is (I did spend a lot of time trying) but RNG manip is really REALLY hard. To have a single shot at a possible different outcome, sometimes I needed to re-do a very complicated section all over again
#14 - what's with the second explosion sound after the enemy dies?
Lol, I wondered the same back then! I have NO idea. I thought maybe the weapon box, but checked and it's still there. We will never know...
#18 - goofy cursor movement on that 1st napalm there, unless there's some actual explanation behind this. also +style points for surviving, RTA usually makes a draw here
Yeah, this one could have been faster (less goofy), but positioning the mouse exactly where needs to be isn't easy because I needed to account for the game scrolling. Also there some waiting to ensure the flame's RNG goes as required.
#25 - i remember Deadcode has made a nasty TAS involving skipping to end timing early, but this is framecount we're talking here, including stuff between inputs, including the next mission etc., so different timing rules, a different solution
I only seen his framecount-targetting solution, it was outlandishly well executed. I opted to go for RTA-friendly one, but I wasn't able to replicate the 3s+3s grenade solution, instead going for 3s+4s.
#26 - ayy, finally someone used a cow in this mission, nice :D
Moo!! (shame they didn't give you a 2+ herd, otherwise the mission could have ended one turn ealier)
#32 - what's with the halt on the 2nd turn there? rng manipulation? also consider using Scroll Lock + mouse movement to manually move the camera to action
the wait is pure RNG manip, so that the clusters go exactly up, pushing the worm to the other. Couldn't make this one work any earlier. The camera was an issue: if you try to correct it after executing the inputs, you may introduce desyncs. So I had to choose between focusing the camera and seeing what my own worm is doing. Doing this blind is almost impossible :/
#33 - bruh that ending LMAO
Haha! I had something much MUCH crazier planned for this. Since in TASing we count the last input, I wanted to leave the worm just there and wait however many minutes until the turns passed and the box appeared, automatically grabbing it. But no, it seems you need to actively perform an action to grab a box, even if it fell on your head and it's your turn. So I opted simply for jumping into the abyss :P
ikuyo: Delaying to account for upcoming improvements.
making an account just so I can comment on this
Hi, I'm the "charles" mentioned in the runner notes. I've had ambitions to TAS this before, as my public notes seem to date all the way back to June 8th, 2025, but I quickly gave up due to running out of patience. I've tried using libTAS with PCem on WinXP.
Ever since the game released and replay support was eventually added, people in the past only considered IL RTA runs, until RuffledBricks started taking full RTA runs of it seriously around 2020, making multiple marathon appearances, including ESA and GDQ!
https://www.speedrun.com/worms_armageddon?h=All_Missions-Any&x=wkpwrvdr-p85r6mvn.z19kmj0qhttps://www.youtube.com/watch?v=3jVqQ2N8aGEhttps://www.youtube.com/watch?v=6NFwlcoLlYw
This game is a really strange case, as it respects wine (it even has a beginning dialog, whether you want to set some registry stuff), but there's no native Linux build of it, meaning you need to emulate a Windows machine to run it. Thankfully, this being an old enough game, it runs on very old operating systems, where even the official updates get accepted, too.
Sadly, I gave feedback to this run too late, as editing inputs in the middle of the run is borderline impossible - desyncs and frames not lining up correctly (despite the engine running in 50fps), to name some of the problems.
Not voting on this because I'm biased.
Welcome, Charles! Thank you for your comments!
Now we have deterministic tools to TAS DOS and Windows 98 games, which the emulator's hardware can lift. If you have other games on your mind, you may give them a new try!
TASing is like making a film: only the best takes are shown in the premier.
https://xkcd.com/3246/
omg hi spikestuff big fan 😳
The Q&A written in the notes is basically my feedback, though applying it is a different problem. This includes some things that might save multiple frames across multiple missions. With that said, it's not an "unoptimized" TAS, as some Deadcode records were clearly sniped.
The Q&A written in the notes is basically my feedback, though applying it is a different problem. This includes some things that might save multiple frames across multiple missions. With that said, it's not an "unoptimized" TAS, as some Deadcode records were clearly sniped.
You're right. You know, TASing is an iterative process. "Version 1" is often dirty at the beginning, but gets better afterwards.
TASing is like making a film: only the best takes are shown in the premier.
https://xkcd.com/3246/
I've always wanted to see a TAS of this game and this one did not disappoint, voting "yes"! No idea what is optimal here, I just had a lot of fun watching this.
Current project: Gex 3 any%
Paused: Gex 64 any%
There are no N64 emulators. Just SM64 emulators with hacky support for all the other games.
This was great! Easy yes vote. My friends and I used to play this on recess periods when we were all ... well, suffice to say we had a lot fewer back problems and mortgages back then.
Excellent work! Some incredibly fun trick shots in there.
including repeated actions that can't be replicated natively
I'd like to clarify this. The TA build allows the following things that can't yet be replicated natively:
Selecting any weapon (including those that would require multiple presses of a function key) and using it in the same frame
Note that with native input, you can't select a weapon and use it in the same frame if there is currently no weapon selected (which can only happen if the team's inventory has no default-selected weapons). This happens for example in Mission #4 "Cool as Ice", where with native input, selecting and using the Jet Pack takes two frames, and selecting and using the Uzi also takes two frames.
Pressing Tab multiple times in the same frame, followed by any other action (including those in this list) also in the same frame
Pressing and releasing Space to detach Ninja Rope and then immediately pressing it again on the next frame to shoot the next rope. I'm pretty sure this is possible in native keyboard input, but BizHawk does not allow it (it requires a key to be released for one full frame before being pressed again) unless the framerate is set to >=100 fps.
Pressing Backspace twice in one frame to do a backflip (e.g. with only 1 frame of retreat remaining)
Note that this is already possible natively in the current alpha, via adding the backslash key as a second Backspace-equivalent binding
Firing a powered-up weapon with zero power, whereas with keyboard input the minimum is currently 1 out of 75 (and the maximum is "76" out of 75).
It was partially Lex's encouragement that resulted in the above features.
I plan making all of those things possible natively by making the input handling that's done between frames more flexible.
Also in the TA build, the mouse cursor can be moved anywhere instantly. It seems to me that this is a gray area in terms of how possible it is "natively".
I doubt use of the above features has much impact on the total time of the tool-assisted Mission replays. I'd estimate that it saved a few seconds at most.
For completeness, I should note that there are additionally some things that the TA build allows in "unrestricted control" mode that are not used in any posted tool-assisted replays – unless specifically noted, but I think I've only shared examples of that privately – and in particular, they're not used in any of the posted Mission replays:
Moving the worm and moving the cursor for a mouse-aimed weapon simultaneously
Moving the cursor for a mouse-aimed weapon outside the normal range. If a girder or teleport is attempted it will fail, but strikes and homing weapons will succeed
Logic version
The current set of tool-assisted Mission record replays cannot represent a contiguous run because they aren't all done in the same game logic version. In this regard, eien86's TAS clearly wins, for now.
I plan on redoing them all in the same logic version, but not yet... because some past game logic changes significantly altered the behavior of certain missions from how they were in WA v1.0-3.0:
The crate drop algorithm was changed, altering the position of the crate drop and wind direction/strength on the first of Missions #18 "Rescue Agent Dennis" and #27 "Bazooka on the Rocks".
It used to be possible to press Space to cut the parachute of any crate drop. This would have been an extra avenue of randomness manipulation, which in the above two missions would even allow manipulating the first turn wind.
Originally, Double Damage did not increase the explosion crater size of a 100-damage weapon such as Holy Hand Grenade. This significantly alters Mission #15 "Hot Stuff".
The number of frames that pass at the start of Sudden Death changed, altering the wind direction/strength on the first turn of Missions #4 "Cool as Ice" and #20 "Rumble in the Farmyard".
A bug in the way CPU worms plan their walking/jumping movement was "fixed", but the fix introduced a worse bug, which heavily biases CPU worms to only walk in one direction.
The amount of time that passes after the last active object stops moving changed in v3.6.31.1, altering the resulting round time in most missions.
I plan on rectifying this by reverting to the original behavior in Missions. (In the case of the CPU movement bias, that will just be reverted altogether.) To redo all of the runs before then just wouldn't make sense.
eien86's TAS
My perspective that while eien86's TAS represents a impressive effort, the emulator used imposes arbitrary restrictions, some of them quite unrealistic. In particular, any modern PC (even systems 20+ years old) should be able to handle input processing at the full 50 fps. The fact that the TAS had to fight against that unrealistic limitation, and that going back to edit an earlier mission would require redoing every subsequent one, and that it couldn't take advantage of TA build features and be truly competitive against previous WA TASing, makes me feel wistful about this TAS. But I like that it brings attention to WA TASing – which I hope will lead to TASing being done with the WA TA build rather than an emulator.
With realistic "natively possible input" limitations, and the power of the TA build (trajectory lookahead, info display, etc.), I think a full "native possible" TAS run should come much closer to the TA build total, not falling 4.6+ minutes short of it like this one does.
So, I have mixed feelings and biases, and I'll abstain from voting.
eien86 wrote:
and even got three new TAS WR that improved even the TASes produced with Deadcode's tool!
It's nice to have some competition, even if for now it is fleeting. I've beaten all three of those records now (Missions 06, 18, and 21 – see round time records). This brings the total down to 2408.94 seconds (measured as >=v3.6.31.1 equivalent), which is 34.78 seconds less than it was before eien86's TAS.
The fact that WA TASing can now be done with BizHawk will likely result in me releasing the WA TA build publicly very soon. What stopped me from doing this for so many years (which I have regretted enormously) was that I was afraid the tool would be used to fraudulently pass off TAS replays as being RTA.
The fact that WA TASing can now be done with BizHawk will likely result in me releasing the WA TA build publicly very soon. What stopped me from doing this for so many years (which I have regretted enormously) was that I was afraid the tool would be used to fraudulently pass off TAS replays as being RTA.
In Doom speedrunning community, the mainstream source port (DSDA-Doom) which provides TAS tools, provides header markers to replay files, indicating that TAS tools were used. You can apply something like that to your build, maybe by introducing exclusive compatibility for a clear distinction.
On the other hand, Doom is a much more complex game than WA. The superior results are deeply examined by other people, so the act of cheating is disclosed one way or the other.
TASing is like making a film: only the best takes are shown in the premier.
https://xkcd.com/3246/
In Doom speedrunning community, the mainstream source port (DSDA-Doom) which provides TAS tools, provides header markers to replay files, indicating that TAS tools were used. You can apply something like that to your build, maybe by introducing exclusive compatibility for a clear distinction.
The TA build already does something like that (though the earliest versions didn't). But it would probably not be too hard to forge, especially with the power of today's reverse-engineering tools and AI.
But I'm not going to let that fear and caution block a TA build release anymore... I feel it's mostly moot anyway now, with there being ways of TASing WA that do not require a special build and thus cannot put any special markers in the header.
Dimon12321 wrote:
On the other hand, Doom is a much more complex game than WA. The superior results are deeply examined by other people, so the act of cheating is disclosed one way or the other.
What relevance does how complex the game is even have to this? But if you are going to compare their complexity (for whatever reason), isn't WA a hugely more complex game than Doom? I would think it has far more weapons and game mechanics, and schemes result in huge possibilities for playing so many very different types of games. And new glitches, as well as qualitatively new game scenarios, are still being found 25+ years later.
On the other hand, Doom is a much more complex game than WA. The superior results are deeply examined by other people, so the act of cheating is disclosed one way or the other.
What relevance does how complex the game is even have to this? But if you are going to compare their complexity (for whatever reason), isn't WA a hugely more complex game than Doom? I would think it has far more weapons and game mechanics, and schemes result in huge possibilities for playing so many very different types of games. And new glitches, as well as qualitatively new game scenarios, are still being found 25+ years later.
I described a flow a suspiciously good movie goes through. Cheater movies are filtered out. You fear (or used to fear) that disguised RTA movies may go through, and in the meantime, you acknowledge a higher complexity of WA. So what's the problem then? Trust the community then. I think it won't let you down
TASing is like making a film: only the best takes are shown in the premier.
https://xkcd.com/3246/
Admin Assistant, Senior Judge, Experienced player
(624)
🇨🇱 Chile
Joined: 7/8/2021
Posts: 205
Location: 🇨🇱 Chile
Deadcode wrote:
The fact that WA TASing can now be done with BizHawk will likely result in me releasing the WA TA build publicly very soon. What stopped me from doing this for so many years (which I have regretted enormously) was that I was afraid the tool would be used to fraudulently pass off TAS replays as being RTA.
Have you looked into the ways that other similar tools have attempted to handle this issue? Both Celeste and UFO 50 have mods that function as TAS tools that have implemented their own safeguards. Any modded Celeste run includes an impossible to remove image in the Chapter complete screen to flag the run as done in a modded build. CoffeeTools forces an input viewer on running any UFO 50 game to signal the runqwas done with the tool on.
This is an interesting case where the necessitiea of RTA communities (signaling the usage of tas tools to prevent abuse) and the goals of TASVideos (making TAS tools that are as close to the original games as possible) often collide, but we acknowledge that given how close together those tools are to their RTA bretheren, it would kot be feasible for us to ask for changes.
And regarding support for WA TA, well, we are right in the middle of discussing support for ILs, so I think we could make some way to have the tool supported on our end.
GuanlongX wrote:
auuuugh my face when games are creative and cool and don't fit arbitrary rulesets
And regarding support for WA TA, well, we are right in the middle of discussing support for ILs, so I think we could make some way to have the tool supported on our end.
It'd be so FIRE to have WA IL TASes being published here.
They are hand-maintaining their own leaderboard where you can download the IL replays and replay them locally using the game itself. But I reckon a wider audience would love to see those incredible solutions as encodes in TV's youtube.
Some amazing moves.
Has anyone tried to find the fastest random seed in Worms 1? If I remember correctly you can insert a number as a seed and get a random level. Some levels are completely flat so we can defeat all opponents fast, but I'm not sure what the best seed is, and the best weapon to use on flat ground?
The fact that WA TASing can now be done with BizHawk will likely result in me releasing the WA TA build publicly very soon. What stopped me from doing this for so many years (which I have regretted enormously) was that I was afraid the tool would be used to fraudulently pass off TAS replays as being RTA.
Have you looked into the ways that other similar tools have attempted to handle this issue? Both Celeste and UFO 50 have mods that function as TAS tools that have implemented their own safeguards. Any modded Celeste run includes an impossible to remove image in the Chapter complete screen to flag the run as done in a modded build. CoffeeTools forces an input viewer on running any UFO 50 game to signal the runqwas done with the tool on.
The problem is, I feel that adding such safeguards is just an invitation for someone to crack the protection, and from there it'd become and arms race. Reverse-engineering is easier these days than ever before. Furthermore, such protection won't mean much unless the replay files are encrypted. And if they're encrypted, there needs to be a layer of compression beneath the encryption, otherwise the files will waste more space than they need to, with no possibility of ever being compressed as much as they should be able to.
Such implementations need extensive debugging to make sure they don't break things. And I also feel that there's a conflict of goals with such implementations; if I do it in a clean way that can be easily maintained (which it is my strong inclination to do), it'll be easier to reverse-engineer and crack.
That pretty much explains right there why I've delayed releasing the TA build for so long – now 22 years. The impediments to giving it proper protection are too daunting.
That is already too long. I would like to release it as soon as possible, to encourage the TAS community to TAS WA in a way that's competitive with previous work. Clearly my previous model of giving a TA build to people who ask (and seem trustworthy, and intend to use it to make TASes for public sharing) has not worked, because for the past several years I've been the only one TASing WA (and even before that, there was virtually no serious competition). It seems pretty clear that the TAS community insists that the tool be public if serious competitive TASing is to occur.
It's a Catch 22. I've been feeling very demotivated for the past several years, from many factors. It would help greatly toward motivating me to work on future releases of WA, if it had a competitive TASing scene. But if before even releasing a TA build, I need to implement layers of protection on the replay format strong enough that they won't be easily crackable, and debug that extensively to show it to be sufficiently reliable, then that's work that'd have to be done in a demotivated state. And that's on top of all the other non-TAS-related things I've already determined are necessary to implement before releasing the next normal WA build.
In the meantime while working on all that, interest in TASing WA continue to wane, and/or there will be more emulator-TASing of WA (which isn't competitive with TA build TASing), and/or someone else will reverse-engineer WA sufficiently to make a TA module for it.
And if TASing can be done with BizHawk anyway, which results in replay files cannot flag that they were TASed, what point is there in even putting all the effort into implementing reliable strong protection, when that time could be more productively spent on other things?
You can never create an environment that makes cheating 100% impossible for RTA speedruns, both on a software level and in leaderboard rulesets, it's just not feasible. If someone is determined enough to cheat, they will find a way. This is why communities always have to be built on trust. Protections against cheating are necessary up to a level that is realistic, enough to deter the vast majority of people from even attempting it. But the most important part is having a collaborative and friendly culture in your community, so that no one would even want to cheat in the first place.
Current project: Gex 3 any%
Paused: Gex 64 any%
There are no N64 emulators. Just SM64 emulators with hacky support for all the other games.
I voted yes. Worms Armageddon is one of my biggest nostalgic games, and I've always loved seeing speedruns and TASes of the missions. There's some incredible strategies and crazy movies all around. Without comparing it to the demos made with the TA tool, the TAS also looked quite solid and well executed.
That said, yeah, the large time difference is hard to overlook. With TASvideos starting to accept game-specific tools nowadays, I honestly think that we would be better served by getting the IL records on the site in some good way.
That said, yeah, the large time difference is hard to overlook. With TASvideos starting to accept game-specific tools nowadays, I honestly think that we would be better served by getting the IL records on the site in some good way.
You have to consider that ILs with the TA tool and a fullgame run with Bizhawk are not comparable. Doing it all in one playthrough is a completely different challenge. Unless the TA tool gets modified to the point where it can also produce a fullgame run that can be played back, this particular TAS here justifies a publication regardless of the existing ILs.
Current project: Gex 3 any%
Paused: Gex 64 any%
There are no N64 emulators. Just SM64 emulators with hacky support for all the other games.
scrimpeh wrote:You have to consider that ILs with the TA tool and a fullgame run with Bizhawk are not comparable. Doing it all in one playthrough is a completely different challenge. Unless the TA tool gets modified to the point where it can also produce a fullgame run that can be played back, this particular TAS here justifies a publication regardless of the existing ILs.
I was mainly going off this statement in Deadcode's post:
With realistic "natively possible input" limitations, and the power of the TA build (trajectory lookahead, info display, etc.), I think a full "native possible" TAS run should come much closer to the TA build total, not falling 4.6+ minutes short of it like this one does.
This made me think that there is more time to save without relying on "impossible" TA-build only input, even in a full-game run. Other than that, I'm not sure what differences there are between ILs and a fullgame run, gameplay-wise, other than the medal system maybe. But maybe I am missing something.
eien86's movie used "cycles=fixed 600000", and had a huge proportion of lagged/skipped frames – I measured that only about 56% of the game logic frames were rendered (and polled for changed input).
With "cycles=fixed 4000000", there's virtually no frame lags or skips, and the TA build replays can in many cases be replicated exactly. In this proof of concept, all of Basic Training and Mission #1 match the TA build round times (with one training stage actually being improved upon when I was forced to depart from the exact TA build replay input sequence). It took 5 hours to make this.
(I don't think DOSBox's cycle count setting is a realistic representation of computer speed. It strikes me as being similar to BogoMIPS.)
Link to videoBizHawk movieXML settings fileDOSBox configuration file
Uses the same .hdd generated by eien86's WA installation movie.
The resulting replay files
In Mission #1, the TA build replay suddenly places a girder on the very first frame after a rope attach. To do this with native input in the BizHawk run, I selected the girder ahead of time and moved it into place while the worm was in the Rope Roll state, and then clicked to do a "buffered" placing of it before the rope attached. Then pressed Enter to actually place it, on the first frame after the rope attached, replicating the timing of the TA build replay.
I accidentally generated a bunch of Pause and S key pressings in this run, due to assuming that conflicting bindings weren't possible. The Pause presses caused some screenshots to be saved, adding a little lag to the run; so, it could perhaps be even a bit faster than it is in this demonstration. The S key presses also had to be kept in the run (I tried removing them, but even that eventually caused a desync). I've fixed my bindings, but will be posting this run as is.
Before trying the increased cycle count, I tried enabling "Timer Workaround" in Advanced Options; it did not help. I also tried the DirectDraw 8-bit hardware renderer, which should in theory be faster, but it caused glitches in BizHawk/DOSBox-X and didn't give a dramatic speed-up, so I stuck with the default DirectDraw 32-bit renderer.
eien86's movie used "cycles=fixed 600000", and had a huge proportion of lagged/skipped frames – I measured that only about 56% of the game logic frames were rendered (and polled for changed input).
With "cycles=fixed 4000000", there's virtually no frame lags or skips, and the TA build replays can in many cases be replicated exactly.
That's some really good observation! I thought cycles alignment to render/input is required for some DOS games only. Like, Windows ones would probably not fit this principle.
TASing is like making a film: only the best takes are shown in the premier.
https://xkcd.com/3246/
i've been meaning to write a wall of bullet points talking about some things that got mentioned in the Q&A, so here goes:
There's a TAS-specific method of walking called "flipwalking", which is a variation of skipwalking, where skipping a few frames of walking makes you walk faster. more about that here
the TAS uses the American soundbank, whereas the de facto default soundbank should be the English/British one. With my limited attempts to TAS this, the only explanation I have is that the region is set to the United States by default. It should be set to the United Kingdom to get the desired soundbank. This could be done during the setup movie.
Is making a custom team necessary? In the game files, in \User\Teams there's WG.WGT file, which can be edited externally. is editing the game files before the TAS starts allowed? One could also make a custom language file that fills out the Team and Worm Names, too. Where is the line drawn?
In Operation Market Garden there is an RTA strategy that is frame perfect because of holding down key sets before the turn starts. Is that strategy faster than the current TAS?
Rope knocking... oh dear, rope knocking... Before v3.8 had that as a scheme setting, it used to be an online-room setting, or a toggle in settings. But in the case of single-player content, it's always disabled. However, it is not disabled after the turn ends (after turn time and retreat time). For example, turn 3 in Sand in Your Eye (around frame 44000), losing control of the turn is used to rope-knock an enemy worm to water.
The biggest concern I have with the difference between Deadcode's TA build and emulating Windows is mouse movement - there are multiple weapons used in the run that require mouse movement to pick a location or a target to function (girders, air strikes, homing missiles, etc.), Deadcode's TA build can do that in 1-ish frame, whereas in emulating Windows, you have to move the mouse, which isn't always a given to do this quickly, especially when something is very well timed.
Is making a custom team necessary? In the game files, in \User\Teams there's WG.WGT file, which can be edited externally. is editing the game files before the TAS starts allowed? One could also make a custom language file that fills out the Team and Worm Names, too. Where is the line drawn?
Making a custom team is fine since it's a small speed/entertainment tradeoff for a personal touch in a TAS. Something small like custom names usually isn't counted against the author in judgment for optimization as long as it's noted in the submission notes.
