Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
darkid wrote:
Well, I've spent a couple of weeks trying to get this to work, with a lot of help from Keylie. I'm having two main issues:
- Playback is inconsistent. Due to the way Unity loads content, stepping over a frame has different impact from just playing the movie outright. (Specifically, if you step and then wait, lag is incurred without any frame cost, but if you play, lag occurs and the framecount increments.)
- Savestates aren't working. Due to the way this game is written, it creates and discards threads fairly often (at the start and end of every level, as well as on every menu change). So, along with the other issue, I have to frame-by-frame step from the beginning of playback to work on this TAS.
I don't think this is a practical approach for this game -- is there any other option?
Does that mean watching the movie also has to be frame by frame? Well without savestates there's no point I guess.
However, your injection method doesn't seem to allow savestates either, is that correct?
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Well, I've spent a couple of weeks trying to get this to work, with a lot of help from Keylie. I'm having two main issues:
- Playback is inconsistent. Due to the way Unity loads content, stepping over a frame has different impact from just playing the movie outright. (Specifically, if you step and then wait, lag is incurred without any frame cost, but if you play, lag occurs and the framecount increments.)
- Savestates aren't working. Due to the way this game is written, it creates and discards threads fairly often (at the start and end of every level, as well as on every menu change). So, along with the other issue, I have to frame-by-frame step from the beginning of playback to work on this TAS.
I don't think this is a practical approach for this game -- is there any other option?
Does that mean watching the movie also has to be frame by frame? Well without savestates there's no point I guess.
However, your injection method doesn't seem to allow savestates either, is that correct?
I'm still working with Keylie, so the frame-by-frame might still get fixed.
You're correct, the injection I had didn't support savestates. It was slightly less necessary, because the hack would input all the keys on the first frame possible (including animation delays).
The game also has a built in "reset level" function / "undo" function, which I integrated into the tool I wrote, so it was pretty easy to go through inputs one at a time.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
Yeah without savestates I don't think we'd accept it as a TASing workflow.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
In Motocross Maniacs (Game Boy), there are 8 race tracks. After each track, the completion time is shown. Reaching the lowest possible completion time for each track should be the primary goal.
However, a trick/bug was discovered by RTA runners where pausing would extend a nitro boost very slightly, allowing time saves of about 1-2 frames per boost.
RTA runners have used this trick in track 1 and reached completion time 0:54.06.
A TAS without pausing can reach 0:54.08. I'm not sure if this is 1 or 2 frames behind. Also this time is speculated, not confirmed yet.
Is the TASer free to decide if he wants to include pausing, or is he required to do so to abide by movie rules: "The movie must beat all existing records"?
Smash Bros 64 Break the Targets players acknowledge that pausing messes up the completion time, so they ban pausing.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
Does pausing cost real time?
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
You can have this trick if the movie goes to Moons, but for Vault we disallow this kinda stuff:
Movie Rules wrote:
Movies that aim for in-game time instead of real-time are only allowed for Vault if that goal makes game-play shorter. If optimizing for the in-game time makes actual game-play longer, such a movie is not eligible for Vault.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
In the DOS version of Zeliard, if you input F9 you can adjust the speed of the game going from 0 to 9. I am unsure if this counts as a debug option or as a cheat. The manual only states it as a command key.
I found a segmented run which uses the different speeds, the SRC runs only seem to be set to one speed.
Link to video
EDIT: Added a scan to the manual due to a request: http://www.sierrahelp.com/Documents/Manuals/Zeliard_-_Manual.pdf
If it's mentioned in the manual along with the other controls, there's a really good argument that it's an in-game feature rather than a cheat or debug key. It's not unprecedented that such a feature would be used in a DOS run. It should be fine.
In-game settings and environment parameters that are explicitly supported and intended for normal play are allowed.
This means there are limited options the game was designed to work with, for example a few speed variants. Explicit support can be proven by in-game options, official instructions and PC spec recommendations, release notes, source code logic and comments, etc. Burden of proof is on the TAS author here. If this information is completely unavailable for a given game, rely on the environment specs that were common and popular in this game's era.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
DarkShamilKhan wrote:
Hello, I am a Sonic TASer and I have a question. More obsolete runs were found, so may I ask can me and Shiro take time to obsolete our run.
I don't understand your question.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Question here!
The Judge Guidelines currently state the following:
There are two basic opinions as to what kind of TAS ending is better:
- Minimize input length, the game should reach a stage where it completes itself, no matter how long it takes.
- Game must be brought to a certain completion state as soon as possible, after which no normal game input can prevent the game from completing, even when this lengthens the movie file length considerably.
The author can tweak their movie's ending as a stylistic choice, even after it was accepted (but before it is published). However, a run looking to obsolete a published run needs to do more than just tweak the movie's ending, as we aim to avoid meaningless publications. A run with real improvements is free to use a different ending than previous movies did.
I'm currently working on a fastest completion TAS where the ending cutscene for the game lasts for about 20 minutes. I have the choice between 2 completely different routes. Route 1 has a shorter input time and immediately reaches the beginning of the ending cutscene afterwards. Route 2 takes about 10 additional minutes of input time, but immediately wrong warps to the END screen at the very end of the ending cutscene, saving about 10 minutes to reach that state.
In other words, route 1 reaches a game completion state earlier, while route 2 reaches a later game completion state earlier.
Am I correct to say that both routes are valid for Vault based on the stylistic choice I want, as described in the Judge Guidelines?
I'm asking because it's not clear whether the stylistic choice is limited between input time vs game completion state, or also includes which game completion state the author prefers to reach first.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
It's up to you, as long as the game properly reaches the final state eventually. Both ways are fine for Vault, because when comparing optimality of 2 movies we compare actual gameplay, which in your case would be identical.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Thanks feos, but your answer appear contradictory to me. I'm not sure if it's because the term "actual gameplay" means something different than what I expect, or because I was unclear in the first place, so allow me to clarify the 2 routes.
Route 1: Warp to the last level, then beat the final boss in 5 minutes. The ending cutscene cannot be interacted with and is 20 minutes long. Total AVI time after the initial warp: 25 minutes,
Route 2: Warp to the last level as before, but then leave, take 15 minutes to make a massive detour halfway around the world to acquire a power-up, then glitch the game using said power-up to wrong warp to the final scene of the ending cutscene without defeating the final boss. The ending cutscene now ends almost immediately afterwards. Total AVI time after the initial warp: 15 minutes.
As such, isn't route 1 containing less actual gameplay than route 2, thus invalidating route 2 for Vault even though the latter's AVI time is faster?
I understand that input time vs interactive state time can both be considered "actual gameplay" and is a matter of opinion, but that logic doesn't apply in this case so I'm still confused here.
Joined: 11/13/2006
Posts: 2822
Location: Northern California
The way I see it, route 1 would be preferred in every case. Even if post-input is shortened by route 2, the ending itself is still triggered far earlier in route 1, so route 2 would just be "suboptimal" in that regard as it requires 10+ minutes of extra gameplay.
TASvideos Admin and acting Senior Judge 💙 Currently unable to dedicate a lot of time to the site, taking care of family.
Now infrequently posting on BlueskywarmCabin wrote:
You shouldn't need a degree in computer science to get into this hobby.
Thanks Samsara for the clear answer!
That makes me a bit sad though as I personally view end of post-input more meaningful than end of gameplay. Not sure how others feel about this, but it may be worth reviewing this case at some point in the future.
That said, I guess I shouldn't complain as it saves me a lot of extra work for now!
Hi, I have a few questions about a potential future submission for Super Smash Bros Melee. The submission would be for completion of every character's Target Test stage.
1) Almost half of Melee's 25 characters must be unlocked, sometimes requiring up to several hours of playtime. Could a valid DTM submission be started with a non-empty memory card that has unlocked all characters?
2) Several characters take advantage of a glitch that is activated by playing through part of the first level of a different game mode. The total time spent playing through this level is larger than the time saved by the glitch, but Target Test is a collection of 25 individual levels which are individually timed using the in-game timer. Could a submission primarily targeting minimum in-game time (with a secondary goal of minimizing submission length in total frames) be accepted?
3) Similarly, 5 of the 25 characters depend on RNG for their individual level world record times, and RNG would similarly be needed to minimize the time spent menuing (as the fastest possible way for the player to select any character is with random select - random select is closer to the character selection cursor's starting point than every character is). If a strategy forces a tradeoff between minimum in-game time and minimum submission length (say if iterating the RNG seed to get to a seed that enables the optimal in-game time costs more frames than the optimal in-game time saves over a strategy that doesn't depend on RNG), does tasvideos have a preference on which strategy is to be used?
4) Most individual levels do not have accurate re-record counts, as a) players improving on others' records do not iterate on the previous record holder's DTMs, and b) newer records are generally created on a hacked version of Melee with (among other things) debug mode access, then the final inputs are applied to a vanilla version of the game, and we have not kept records of re-record counts on the hacked version. Would the lack of an accurate re-record count negatively impact the submission?
Joined: 11/13/2006
Posts: 2822
Location: Northern California
PracticalTAS wrote:
Hi, I have a few questions about a potential future submission for Super Smash Bros Melee. The submission would be for completion of every character's Target Test stage.
1) Almost half of Melee's 25 characters must be unlocked, sometimes requiring up to several hours of playtime. Could a valid DTM submission be started with a non-empty memory card that has unlocked all characters?
The SSB64 run does this, so it's acceptable as long as there is a verification file for it. Given the nature of Melee and character unlocking, though, the nature of such a file can be discussed further. I highly recommend talking to Mothrayas about it personally, that would be the easiest path of figuring out the best balance between being site-acceptable and leaving your Gamecube on overnight to unlock Mewtwo.
2) Several characters take advantage of a glitch that is activated by playing through part of the first level of a different game mode. The total time spent playing through this level is larger than the time saved by the glitch, but Target Test is a collection of 25 individual levels which are individually timed using the in-game timer. Could a submission primarily targeting minimum in-game time (with a secondary goal of minimizing submission length in total frames) be accepted?
In-game time is acceptable, though it is explicitly a Moons goal. Tiering shouldn't be a problem with Melee, though, so no worries there.
3) Similarly, 5 of the 25 characters depend on RNG for their individual level world record times, and RNG would similarly be needed to minimize the time spent menuing (as the fastest possible way for the player to select any character is with random select - random select is closer to the character selection cursor's starting point than every character is). If a strategy forces a tradeoff between minimum in-game time and minimum submission length (say if iterating the RNG seed to get to a seed that enables the optimal in-game time costs more frames than the optimal in-game time saves over a strategy that doesn't depend on RNG), does tasvideos have a preference on which strategy is to be used?
If you're aiming for in-game time, in-game time takes precedence over real time, though real time should still be minimized wherever possible. If it's only a matter of frames, it's no trouble at all, so feel free to sacrifice those frames to get the optimal RNG. It shouldn't even be a problem at all unless the manipulation takes several seconds or more per stage.
4) Most individual levels do not have accurate re-record counts, as a) players improving on others' records do not iterate on the previous record holder's DTMs, and b) newer records are generally created on a hacked version of Melee with (among other things) debug mode access, then the final inputs are applied to a vanilla version of the game, and we have not kept records of re-record counts on the hacked version. Would the lack of an accurate re-record count negatively impact the submission?
Not at all. Just 0 out the count from the submission DTM and the parser will mark it as "unknown". Alternately, you can put in an estimation. Or don't bother doing either and just give that explanation upfront with an unedited final file, nobody's going to call you out on it.
TASvideos Admin and acting Senior Judge 💙 Currently unable to dedicate a lot of time to the site, taking care of family.
Now infrequently posting on BlueskywarmCabin wrote:
You shouldn't need a degree in computer science to get into this hobby.
Thank you for the very quick reply! I have a fifth question that I tried to edit in, but it looks like you got to the first four before I could add it, so I'll move it down here:
5) Controller polling in Melee is decoupled from the game engine, which greatly complicates the act of writing a TAS if a fix in the form of a gecko code is not applied. The completed DTM can be modified to reverse the effect of the fix (and has been console verified to sync). Would tasvideos have an issue with a submission that has been written with a gecko code active, then modified afterward to sync without requiring the gecko code?
Joined: 11/13/2006
Posts: 2822
Location: Northern California
It's perfectly fine as long as the DTM syncs on Dolphin for a Judge and a Publisher.
TASvideos Admin and acting Senior Judge 💙 Currently unable to dedicate a lot of time to the site, taking care of family.
Now infrequently posting on BlueskywarmCabin wrote:
You shouldn't need a degree in computer science to get into this hobby.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
SmashManiac wrote:
Thanks feos, but your answer appear contradictory to me. I'm not sure if it's because the term "actual gameplay" means something different than what I expect, or because I was unclear in the first place, so allow me to clarify the 2 routes.
Route 1: Warp to the last level, then beat the final boss in 5 minutes. The ending cutscene cannot be interacted with and is 20 minutes long. Total AVI time after the initial warp: 25 minutes,
Route 2: Warp to the last level as before, but then leave, take 15 minutes to make a massive detour halfway around the world to acquire a power-up, then glitch the game using said power-up to wrong warp to the final scene of the ending cutscene without defeating the final boss. The ending cutscene now ends almost immediately afterwards. Total AVI time after the initial warp: 15 minutes.
As such, isn't route 1 containing less actual gameplay than route 2, thus invalidating route 2 for Vault even though the latter's AVI time is faster?
I understand that input time vs interactive state time can both be considered "actual gameplay" and is a matter of opinion, but that logic doesn't apply in this case so I'm still confused here.
If the ending itself is just that long, it doesn't matter if you make the game jump to the beginning of that ending or to its end. What matters is when you trigger the ending itself, and actual gameplay is what happens before this ending, it's the thing that makes the ending happen. So we measure optimality of that, and movies end when it ends. If route 2 triggers the ending later, then it's less optimal in terms of Vault.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Joined: 4/17/2010
Posts: 11477
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
Samsara wrote:
In-game time is acceptable, though it is explicitly a Moons goal. Tiering shouldn't be a problem with Melee, though, so no worries there.
In-game time is usually eligible for Vault.
Vault rules wrote:
Movies that aim for in-game time instead of real-time are only allowed for Vault if that goal makes game-play shorter. If optimizing for the in-game time makes actual game-play longer, such a movie is not eligible for Vault.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
