Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
I'm still not sure. If the run was optimized for maximum score first (as it is) and then for minimal time second, how much shorter would the movie be, roughly?
It would be roughly 10 seconds shorter because, as Sand mentioned, it might be possible to find another RNG seed that would eliminate the last screen transition on level 3.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
I mean isn't this what fastest time as a secondary goal means? You maximize the primary criterion (score), and then whenever it won't sacrifice the score, you also optimize time, no?
True. I guess in end I didn’t want to have to deal with considerations like RNG or perfect menuing, when what entertainment value there is in the run is totally captured by the super-human typing that is covered in the primary goal. I really don’t see having Fastest Completion as a secondary goal adding any more entertainment value than is already there.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
c-square wrote:
That said, fastest completion time is not a secondary goal of this run, so I don't think it matters in this case.
Wait doesn't it aim for speed at all?
No, it does not. In this game, aiming for both speed and score is impossible since the more letters you type, the slower the run is. This is primarily due to the transition screens in level 3, but there's also lag on at least level 1, and perhaps the others too. While I did strive to optimize inputs in boot up and menuing, the only goal of the run to attain the highest score possible at the expense of fastest completion time.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Sand wrote:
I voted yes.
Is the order of sentences in level 3 fixed? Or is it random? If RNG manipulation is possible, then it may be possible to optimize for real time by getting sentences that are as long as possible. My reasoning is that the timer doesn't count down during the pipe transition cutscenes, each of which takes about 10 seconds. The 5:00 timer takes 8:42 of real time to count down. The last pipe transition cutscene happens when the timer reads 4:59, which means that if any of the earlier active typing segments had lasted even 1 second longer, it would have saved 10 seconds of real time with the same high score.
But if the order of sentences is fixed, then it really is an autoscroller and I don't see any way to improve it.
Thanks, Sand! If memory serves, level 3 chooses randomly from a set of pre-defined lists of sentences, and chooses randomly to start at one of the "start points" in the list. It then moves systematically through the list until it gets to the end and then jumps back to the beginning of the chosen list, repeating as many times as needed. So, while the answer to your question is "yes, the order of the sentences is fixed", one could manipulate which list is chosen or the starting point on the list, potentially resulting in a shorter time for level 3, which matches the spirit of what you were saying. That said, fastest completion time is not a secondary goal of this run, so I don't think it matters in this case.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
DJ Incendration wrote:
I have a question about the canceling message. Isn't maximum score acceptable category? I'm pretty sure there are maximum score publications. I really think this could be one of them.
Thanks, DJ! I've been away from the site for a long time and for some reason I didn't get notified you posted this. I've reopened the submission and I've got my fingers crossed this will get published this time (third time's a charm!).
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Samsara wrote:
Updating the movie rules is a separate beast. We're actively planning on doing so, but apart from some terminology updates we're holding off until we get these changes ironed out first. Ideally, we'd like to go through every rule with the community and figure out what's worth keeping and what needs to be changed/removed. It'll be a new thread, so look out for that in the near future.
Great! I'll definitely keep watch for that. Thanks, Samsara!
Last thing for now: I'm absolutely in favor of a Score Attack goal. It's an easily definable and measurable metric by which to judge a run, and is a standard way video game records have been historically captured.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Low-priority request:
In the forums, below the poster's avatar, is posted the date and time the user joined the site. Would you please remove the time (i.e. only show the date)? I don't think the time provides any value and it carries over to the next line, making it a little confusing to look at.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Memory wrote:
Those rules are in movie rules, not Standard rules. All movie rules still apply (tho they do need to be updated).
Thanks for the quick reply, Memory. I agree, there'll have to be some changes to the movie rules as they currently explicitly separate rules into 'vault' and 'moon' rules. My question then is, how are the movie rules changing in light of the new class system? If it's not determined yet, is there a way to be part of the conversation in determining the changes?
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Congratulations! I think this is an amazing step for opening up this site to all sorts of great content and growing the ranks of TAS content producers. Changing the status quo is never easy, so I applaud all the initiative and hard work that people have put into this to make it happen.
A couple thoughts/questions: I noticed that the new rules have no triviality restriction. I personally believe there shouldn't be one, however I wanted to double-check to make sure I'm understanding it correctly, especially given a recent conversation around the subject. Under the new rules, runs are acceptable as long as they exhibit sound TAS techniques throughout, appear optimal, and beat all known records for their chosen goal. Do I have that correct?
I also noticed that there is no criteria stating that the movie must stand out from unassisted play. I do think there would be value to adding this rule back in. As I mentioned in my post above, I believe the value of this site is to offer movies showcasing superhuman play. The site's unique library would be watered down significantly if it also contained movies indistinguishable from those on non-TAS speedrunning sites.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
c-square wrote:
I ask that the question of triviality not be pre-judged here and now, but left open for discussion in the future. This way, the run is free to be adjudicated on its merits against the site rules at that time, and is not weighed down by any judgement of triviality made today.
That's a fair request! You can even cancel this until then, and uncancel whenever things change. Anything that would be put in the rejection note would basically repeat what is already said in the thread, and you won't be able to unreject it yourself if we reject it.
Thanks, that sounds great! Cancelling for now; to be brought out of stasis in the future!
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
UPDATE: My apologies for the copyrighted links. It was late and I was tired, but I still should have known better.
I did some thinking, and what it boils down to is this:
This TAS is destined to be rejected again because it doesn't have the entertainment value to make Moons, and Vault doesn't permit non-speed-based runs. And I fully accept that.
What I want is, at some point in time when Score Attack runs are permitted, to have the opportunity to reopen this for judging. I ask that the question of triviality not be pre-judged here and now, but left open for discussion in the future. This way, the run is free to be adjudicated on its merits against the site rules at that time, and is not weighed down by any judgement of triviality made today.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
The game seems to match what I meant. If the input rate is not constant, optimization is non-linear. Still borderline trivial IMO because you just look at the upcoming 6 chars, hold their buttons, and the game advances to the next few chars. You don't really have to ponder any strategies at all.
As letters must be entered in the proper order, just holding them down doesn't work. You'd chance having the wrong key picked up first. That said, it isn't that difficult a task to track where you've typed to so far, compare that to what keys have been accepted so far, and type some additional ones in order until you are six ahead of the last accepted letter.
For example: Say the string to type is "TASVideos is awesome!".
Frame 0
Since the keyboard buffer starts empty, you type six letters:
- Press Shift (doesn't count as a letter)
- Press and release 't'
- Press and release 'a'
- Press and release 's'
- Press and release 'v'
- Release Shift (doesn't count as a letter)
- Press and release 'i'
- Press and release 'd'
- Frame Advance
Frame 1
The game picks up three letters, 'TAS' from the keyboard buffer. The buffer still contains 'Vid'. You type three more letters to fill the buffer back up to six:
- Press and release 'e'
- Press and release 'o'
- Press and release 's'
- Frame Advance
Frame 2
The game picks up four letters, 'Vide' from the keyboard buffer. The buffer still contains 'os'. You type four more letters to fill the buffer back up to six:
Etc... I'm sure you get the point.
feos wrote:
c-square wrote:
Regarding Boxxle, do you mean that walking one direction takes a different number of frames than walking in another, or that pushing in one direction is slower than pushing in another? If so, I can certainly see a complicated function to solve; however, from watching the run it definitely appears that direction doesn't impact movement speed.
Even with constant speed, there's a lot of movement in different directions that you need to optimize. Proving it's trivial once you know the route, and that's the only possible route, would in itself be non-trivial.
Maybe I'm missing something, but it seems very similar to me.
For example: Say the known optimum solution to a level is "Up, Left, Up, Down, Right, Up".
Movement Phase 0
(Willy is stationary and hasn't started moving yet)
Press Up to start the first movement.
Movement Phase 1
When up movement starts release Up.
Press Left.
[Willy continues moving up until he has moved an entire square]
Movement Phase 2
When left movement starts release Left.
Press Up.
[Willy continues moving left until he has moved an entire square]
Movement Phase 3
When up movement starts release Up.
Press Down.
[Willy continues moving up until he has moved an entire square]
Etc..
In practice they are exactly the same. Both have are a pre-defined list of correct inputs and you just have to press the next correct button at the correct time. In both cases, you don't really have to ponder any strategies at all.
(If you haven't played Boxxle, I encourage you try it out. You can also try Mario Teaches Typing for comparison.)Mod edit: Do not link to copyrighted material.
In addition, unlike Mario Teaches Typing (MTT), Boxxle doesn't even require any frame precision! You can change your direction key anytime during the current movement phase and it will get picked up immediately after the current movement phase ends. Compared to the hundreds of thousands of sub-frame inputs required for MTT, TASing Boxxle is extremely trivial.
Furthermore, the MTT TAS qualifies more for publication than Boxxle because the latter fails the "The gameplay needs to stand out from unassisted play" rule. Compare Boxxle TAS gameplay vs unassisted gameplay. They are practically indistinguishable. Now contrast unassisted MTT vs. this TAS. There is absolutely no comparison between the two.
Regarding Salary Man and Bishi Bashi Special 3...
feos wrote:
When demoting Moons movies to Vault we don't do full rejudgement from scratch, because priority is too low. If those games run at fixed timer and all you do is just hit known-in-advance buttons as many times as you can, then that's trivial indeed.
Got it. I'll stop using them as an example of precedence. And a shout out to CasualPokePlayer for the response as well. That makes total sense why we wouldn't unpublish runs.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
No, I didn't think to mention it. To expand on what I wrote in the submission:
c-square wrote:
Levels 2 and 4 remove all limitations on the speed of input, meaning that multiple keys are accepted per frame. This means that the strategy used in Level 1 does not work here, and a completely different strategy is required to make sure the script is keeping the input flowing just ahead of what the system can accept.
Each frame, the system provides the game with between 1 and 4 inputs from the keyboard buffer. There doesn't appear to be any obvious pattern to the number of inputs accepted each frame. The script keeps the keyboard buffer primed with six characters beyond the last character accepted to ensure no frame is sub-optimal.
I'm not sure if that is what you meant by "makes input rate per level unpredictable throughout the whole level, and with due effort you could reach unexpected quantities of accepted input," though if it isn't, I'd like to know what was meant.
Regarding Boxxle, do you mean that walking one direction takes a different number of frames than walking in another, or that pushing in one direction is slower than pushing in another? If so, I can certainly see a complicated function to solve; however, from watching the run it definitely appears that direction doesn't impact movement speed.
The fact that Salary Man and Bishi Bashi Special 3 were once accepted to Moons is irrelevant. If they stayed in Moons, then there's no issue. But if these runs are being demoted out of moons and being considered for vault, then they should be subject to the same criteria as any other submission for vault. If they pass, they can enter vault, if they don't they should be unpublished. For example, Moons doesn't have a goal choice limitation either, but that doesn't mean a once-moons Playaround should get to be 'grandfathered' into vault just because.
(For the record, I think this highlights the inanity of our current tiered system, and I am really hoping discussions such as these go away soon with changes to how we publish runs.)
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
I'm just getting caught up on this now, and I absolutely love the idea. I know the conversation has died down, but I hope it's not too late to add my input.
To me, TASVideos has always strived to be the definitive collection of high-quality video game superplays. In that vein, I believe the criteria for publication on this site can be distilled down to one simple question:
Does it showcase both obvious and optimum superhuman play?
As long as a movie displays both, then there's value in adding it to our collection. And while both what is obvious and what is optimum are subjective in nature, for the most part both are easily demonstrable; for the times where they are not, that is precisely the reason we have expert judges.
feos wrote:
Question to everyone
How would you decide when there's too many branches if you had absolute freedom?
Our goal is to showcase superhuman play. There are several categories of superhuman play, including (ordered by increasing technical merit):
- Superhuman Speed
- Superhuman Precision
- Superhuman Tactics/Strategy
- Superhuman Luck
- ACE
These are the ones I can think of off the top of my head. I'm sure there are more.
As long as a submission showcases more obvious or optimal superhuman play along the same lines as an existing run, it should obsolete that run. If, however, a run introduces new categories of superhuman play, or if it introduces different aspects of the existing categories, then it deserves its own branch.
For example, there are often times when a game is played with more than one player, new strategies are introduced that are not available in the single player run. In this case, it deserves its own category because the play the movie presents is markedly different. In contrast, some games have little interaction between players, in which case a multi-player run may just showcase better superhuman speed or luck. As this second type of run would be providing a more obvious display of the same criteria, it would obsolete the single-player branch.
To sum up, I think what has hindered us in the past is that we have been too focused on the input to our process and not put enough focus on the output. While having clear technical criteria on what should and should not be published can give us some hard-and-fast decision-making advantages, it has proven inadequate because technical criteria does not equal quality.
Our acceptance criteria should be instead be focused on the output of each run, and drive maximizing what makes TASses wow both TAS and non-TAS communities alike. If we strive to host the most obvious and optimum superhuman runs, then everything else should take care of itself.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
And just like with Sharp Shot, this movie is not a speedrun record, it's a score attack. While in the future I would really like to add score attacks for timed games as an acceptable goal (looks like the only competitive goal for them), the current Vault rules are worded in a way that doesn't allow non-speedruns.
Yes, I fully agree with your reasoning above.
feos wrote:
If some glitch is discovered that makes input rate per level unpredictable throughout the whole level, and with due effort you could reach unexpected quantities of accepted input, then there's again a complicated function to solve. Until then, it looks absolutely linear and pre-determined.
I don't quite understand what you mean here. Both levels 2 and 4 have unpredictable input rates, varying between 1 and 4 letters per frame.
I also don't feel my point about other equally trivial vault publications has been addressed. The solutions for each level in [3480] GB Boxxle by Jigwally in 4:58:50.20 are predetermined and the input for each is absolutely linear. Similarly, the mini-games in [2377] PSX Salary Man Champ: Tatakau Salary Man "2 players" by Spikestuff in 16:52.47 have equally linear and predetermined inputs. How is each of the above publications distinctively different and non-trivial compared with this submission?
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
feos wrote:
Does the game ever punish you for not doing anything? For mistakes?
Yes. There are actually four scoring metrics for each level:
- Keys Typed
- Errors
- Words per Minute
- Accuracy
Any time you input an incorrect key, you are given an Error which is counted against you at the end of each level. Errors also negatively impact your Accuracy score. In addition, typing slowly will result in lower Keys Typed and Words per Minute scores.
This TAS maximizes all four scoring metrics each round (with zero Errors being considered optimal). The final scoring pages of rounds 2 and 4 show zero for both Words per Minute and Accuracy due to the game not being able to handle the sheer number of characters typed in each round, however an accurate value for each can be easily calculated.
feos wrote:
Also
feos wrote:
But why it is absolutely impossible to TAS this game manually?
Oops, I missed this. It is certainly possible to TAS the game manually, however as I mentioned in the updated submission notes, it is rather unfeasible to do so in real life. The TAS requires over 224,000 frame perfect inputs (updated, since I had forgotten that level 4 drops the hundred thousands' digit). Assuming each input averages approximately 15 seconds to author (which I consider very conservative given the time it takes to load the save state, switch to the virtual keyboard, click twice on the correct input key(s), switch back to the main window, frame advance, confirm the result and create a new save state), this TAS would take over 933 hours to TAS manually. Someone who did this professionally for 40 hours a week could finish in roughly 5.5 months. However, more realistically, someone who has 4-6 hours per week to work on TASsing would require 3.5 years of boring, repetitive work to get this done. And that is also under the assumption that all inputs were perfectly entered the first time, and no mistakes were made that would require the TASer to restart from the error point.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Thanks for the quick response. I can certainly see the similarities between this run and the Sharp Shot TAS. For what it's worth, there are a couple differences between the two.
I was wrong, in fact, to say that it is impossible to complete the levels sooner. On the third level, the clock is stopped on each transition screen. Thus, the more transition screens you traverse, the longer the level will last. This makes it impossible to both achieve fastest time and highest score on that level. As this TAS aims for maximum points, it takes 7 minutes and 42 seconds to run out the 5 minute clock. A TAS aiming for fastest time would complete every letter except the final period on the first screen and then sit around doing nothing to end the level at the minimum 5 minutes.
The second difference, (which in my mind is insignificant but I'll mention it for completeness-sake), is that for the first three levels, the times are only set minimums. There are "bonus" letters you can type that allow the player to see an animation of Mario collecting coins (these are not tallied in the game), which stops the clock and provides the player a chance to rest and refocus. Of course the TAS foregoes these, and thus ensures the minimum time is achieved for levels 1 and 2. As I said, I don't really consider this a significant difference as it takes literally no effort to ignore the "bonus" letters.
Even with the first difference mentioned above, I can see the spirit of the judgement of #6837: Winslinator's INTV Sharp Shot "maximum score" in 04:05.24 applying here. I'd like to request an update to the vault movie rules to make it clearer that the fixed time restriction doesn’t only apply to sports games. Something like changing:
Sports running with a fixed time, such as football, soccer, or basketball are not eligible.
Games featuring multiple successive sports, such as track & field and Olympics, are eligible as long as one of the sports fits the criteria.
to:
Games with a fixed time are not eligible. These include sports such as football, soccer, or basketball, but also extend to any game with a fixed time.
Games featuring multiple segments where some segments have a fixed time are eligible as long as one of the segments does not.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Hi feos,
To answer your second question first: Yes, the levels are timed no matter what and the times are fixed. No, it is impossible to complete them sooner. This means, of the two goal choices allowed for vault, fastest completion is not an option. Full-completion is the only valid goal choice for this TAS; which takes us to your first question.
The only reasonable definition of full-completion here is maximum points. To qualify for maximum points, "max score" must be reached in every level and the movie must complete the game. In this TAS, maximum score is obtained in every level by having programmatically ensured the maximum number of point-gaining inputs are entered in every frame. The game is completed by playing through each level until all unique content is exhausted.
In essence, this game is an auto-scroller that aims for maximum score instead of shortest time, similar to the goals of #2065: Highness's NES Gumshoe in 14:43.98. Alternatively, you could look at it as the furthest distance travelled on the letter/word line within the time limit (which is equivalent to maximum score in this case). Regardless, with the Gumshoe submission, as would be the case with this TAS, any submission that obtains a higher score would qualify to obsolete the existing publication.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Thanks CasualPokePlayer and feos for your responses. I would appreciate the opportunity make my case on the submission and have it be judged on its merits given the new ruleset.
EDIT: Saw the submission was reopened. Thanks again, feos. I've added my reasoning for publication in the submission comments, and await the judges' feedback/decision.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
Hello judges!
Based on the recent rules change, I'd like to request a rejudgement of #5893: c-square's DOS Mario Teaches Typing "maximum score" in 25:57.78. I can attest that developing and tuning the lua scripts to tas it optimally was certainly not a trivial task.
Also, I was very (happily) surprised to see the rule change. If you're able to share, what was the catalyst that prompted a second look at the blanket 'educational'-title restriction?
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
I'm so happy to see this submitted. What an amazing work of art, completely optimizing and breaking this game. Yes vote and I'd love to see this get stars.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
I found this entertaining enough to vote yes. I’m with Radiant; “only smallest jumps” feels too arbitrary. However, I also agree that if “fewest jumps” is a valid category, then “least jumping” is as well.
EDIT: I did some thinking and realized why this feels arbitrary. "Only smallest jumps" cannot stand as a category on its own. By itself, "only smallest jumps" could mean 15 smallest jumps, it could mean 1 million smallest jumps. It's not a metric, but is instead a self-imposed gameplay limitation, similar to foregoing glitches. It requires the "shortest time" goal as well to make it a meaningful run, and give a metric by which a future run could obsolete it.
In contrast, "fewest jumps" can stand on its own. If you took the time component away, a "fewest jumps" run would still give a well-defined metric by which the run can be measured, and subsequently obsoleted. Same would be true for "fewest frames jumping".
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
GMP wrote:
I used the cracked version and thought it would be exactly for the reasons eien86 stated and additionally I would like to add that this game is open source now as the creator of this game, Jordan Mechner uploaded the source code of the game to GitHub. But if it still fails to clear the site guidelines, I completely respect that.
Is the version you used the one that’s on GitHub? A strong argument could be made for using that one, as it comes directly from the creator.
Experienced Forum User, Published Author, Active player
(378)
Joined: 9/25/2011
Posts: 652
slamo wrote:
I think we have a problem with the game files used in this one. So this run uses a cracked version of the 1.3 PRINCE.EXE file that skips a copy protection sequence. The copy protection is a mini-level that appears after level 1 where you have to look up a letter in the manual and drink the corresponding potion.
I was really sad to read this. slamo is right; per the guidelines, the non-cracked version should be used. I'm hoping there might be a chance we could accept this anyway and say any subsequent run has to use the proper version. At the same time, I can see the judge not wanting to set a precedent of allowing it. :(
GMP, if you can, please reach out to the team doing the "with glitches" any% and tell them about this before they get too far on the wrong version.
