TASVideos

Posted: 8/30/2026 1:03 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

What games do these timing nuances matter for? I know there's the SMB confusion (although I don't see why both runs can't exist). For other games, can't the TASers just figure out what timing method makes the most sense (probably standard TAS timing in 99% of cases anyway)?

Posted: 8/26/2026 3:18 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

HappyLee wrote:

At the same time, I hope that more people will also respect TAS rules. We do not need several nearly identical versions of every TAS category, such as a “no L+R” Super Mario World TAS or a “no L+R & no U+D” Mega Man TAS. I wouldn't want to call the Super Mario World TASes here the "L+R & U+D" TAS; I simply call them TASes.

Why not though? I'm currently working on a Yoshi's Island "no L+R" run, although in my case the run is significantly different due to the absence of a major glitch. But in general, if people want to make the runs, why not? I think all these runs are potentially interesting and useful. Besides that I basically agree with the others. I don't really see what people at tasvideos can or should do. The SMB problem seems to be a problem with the RTA community's use of various terms. You're probably better off talking to them and their content creators. Personally, I don't really see why it all matters lol. They're still on the verge of a TAS tie. They're just trying to tie the TAS that is most relevant to them. If "the true any% TAS" is not something they ever need to worry about, then yeah, I wouldn't expect them to talk about it. It might just be the case that the L+R TAS is of less general interest than the one they're trying to tie and never gets the same level of recognition. It's still a fine piece of work of course.

Posted: 8/15/2026 9:49 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

World 1 is now complete! https://youtu.be/iX-O21Kn0vg?is=AHHKz6xJVnaMXZIA

Posted: 7/28/2026 3:40 AM
(Edited: 7/28/2026 3:41 AM)

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Thanks for this. I think the Igor thread was sad. Putting the religious stuff aside, I don’t think the onus should be on an author to morally justify their use of AI (replace AI with LLM/Generative AI/whatever as required). An author also shouldn't be shamed or embarrassed in their submission. Either the site allows AI or it doesn’t. I don’t think the discussion needs to go much beyond that. I read the policy draft and wanted to share a few thoughts. Generally, I think the content could be streamlined. I appreciate that many people are interested in the site’s values, but, as an author, I mainly want to know what is and isn’t allowed. Can I use AI to write my submission text? What if I only use it to fix my spelling and grammar? Can I use AI to write Lua scripts or other supporting tools for me? (admittedly I have no idea how you’d ever check this). I’m not saying these questions haven’t been answered, but it would be nice if I could find things like this at a glance in a more clear-cut manner. I think whatever rules are finalized for text in particular should also be presented to authors at the moment of submission (just so there are no excuses!). I agree with FitterSpace that the draft feels too opinionated. This is most obvious in the TASVideos AI Policy section, which feels a bit bloated and may invite unnecessary challenges. Another example is the paragraph beginning with ‘Just remember, you have an infinite amount of time to make a TAS.’ While I appreciate the sentiment, I think statements like this are unnecessary and potentially provocative. For example, maybe I don’t have as much time as I’d like to make a TAS, or to learn how to write Lua scripts, or to learn Assembly to understand a complicated glitch (I was actually faced with this problem recently). These are cases where AI could be useful to people. In general, I don’t think so many opinions/value judgements are needed. You could outline at the beginning that LLMs are not supported due to various ethical concerns (perhaps with some links/references to sources that describe the issues in more detail) and have a brief statement of the site’s values. The rest of the policy could then focus more on practical questions about what is and isn't accepted, accessibility considerations etc. This may or may not matter if the policy is streamlined in the way I’m imagining, but it seems to me that there is not a sufficiently clear distinction between AI and 'the AI industry'. I think this is important because most of the moral issues seem to concern the latter. The sentence ‘TASVideos does not support LLM usage in its current state.’ seems clear enough (although what is meant by state? state of the technology or the industry?), but later we have ‘This policy will not change until the AI industry does. For as long as it remains unchallenged, unregulated, and unethical, TASVideos staff and developers will remain opposed to using it.‘ Here it seems ‘it’ refers to the industry. Admittedly I don’t know much about this, so what I’m saying might just be rubbish, but, does this for example mean that the site is not in principle opposed to self-hosted, open-source, AI models?

Posted: 7/23/2026 5:27 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Obviously a brilliant run. Perhaps one of the very best we'll see. I'm also super impressed by the scale of the collaboration that went into it, including all the people who made the various tools that were used. It's a bit hard to fathom for those of us who TAS 'smaller' games in comparison.

Posted: 7/18/2026 2:34 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Does the site have an official stance/policy on LLM usage in preparing TASes? I had a quick search, but I'm not really sure where to look. This thread gives the impression that any mention of an LLM makes a submission unacceptable.

Posted: 7/17/2026 7:21 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Disclaimer: I haven't watched the whole thing. This is a wild run. Already within the first 5-10 minutes I have no idea what is going on lol. I admire your passion for the game, especially given that you've made other similar videos. I can see why the religious stuff is problematic, but I guess that's more for the site admins to figure out, so I won't say anything more on it. I guess I had two main thoughts/concerns with what I did watch. 1) I think personally I would enjoy more an abridged run that features the best selection of these "glitches". Is it possible, for example, to follow the any% route through the game (or something close to it... can you make your own route using ACE that only features the best levels?) and have something that is e.g. in the 30-60 minute range? Maybe I'm just spoiled at this point, or maybe it's an attention span issue, but to be completely honest there's only so much wackiness I can take before I start to lose interest. It's all technically very impressive, but ~3 hours is a very long movie, especially when there isn't a clear goal. 2) As I was watching, I found myself wondering how many of the "glitches" being shown were "genuine" and how many were only possible because of ACE. I appreciate that you've explained exactly what you've modified in your submission text, but it's obviously a lot to take in and keep track of when watching the run. There is a lingering feeling I have while watching of "yeah, that's crazy, but it's ACE, so you can do anything, right?". I know that isn't fair, and I'm not really sure what I'm suggesting. Maybe it's better if it's presented as a "playaround" rather than a "glitchfest", but maybe it doesn't really matter in the end.

unwright wrote:

You didn't do this TAS, you let ChatGPT do the TAS... it ran the TAS for you.

If I understand the author correctly, then this seems unfair. They did all the inputs themselves.

Posted: 7/9/2026 4:52 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I've started working on a new magical journey TAS and here's the first WIP! https://youtu.be/-Xz_oZFtoQ4?is=aMzeHT3tNthX3MzY Fairly standard stuff so far, but the next few levels should have some pretty big surprises 😉 The RTA time for this category is around 40 minutes, so I guess it'll be 30-40 mins in the end.

Posted: 6/23/2026 3:23 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I'm sorry if I came across as offensive. I didn't mean to imply you did no research. I'm just passionate about this game and want to see the best runs possible. Just to be clear, I didn't vote no. I was thinking more about future runs and what new considerations would need to be made. I appreciate that the resources for this game aren't the best. If you or anybody else is interested, I'd recommend joining one of our discord servers. You can find some links here https://yoshi.science/index.php/Main_Page

Posted: 6/23/2026 2:28 AM
(Edited: 6/23/2026 2:33 AM)

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Well done on completing the TAS. Yoshi's Island is not an easy game and it's always nice to see new people taking it on. I have quite a lot of comments about the run, most of which is critique. I hope the authors (or future authors) will take some of these points on board. First of all, I don't see how this is ACE. I'm not an expert on the 2-7 glitch, but as far as I understand it's basically having too many sprites in memory to cause some sort of overflow that leaks into level data to unlock 6-8. If it really were ACE, wouldn't you be able to execute arbitrary code? And if you could execute arbitrary code, it would obviously make more sense just to jump to the end of the game. I think this run is simply a good old "any%" run that uses every available glitch to beat the game as fast as possible. Wall jumps are missing throughout the run. The authors claim "Unlike the SNES TAS, Yoshi cannot wall-jump", but this is simply false. Searching for "super mario advance 3 tas" on YouTube brings up this https://www.youtube.com/watch?v=JM76W1X1leA which is 13 years old and uses wall jumps in the very first level. This suggests to me that the authors probably did not do a lot of research before starting the run, which is unfortunate because the game has quite a long history. I also agree with Tompa that the run generally feels quite boring. Yoshi's Island gives opportunities to juggle enemies, throw eggs etc. all over the place. None of it is mandatory, of course, but, in my opinion, it is one of the things that makes the game special and it is a shame not to showcase it. I also think Walgrey did not actually answer Tompa's question. The point is that a lot of things (egg shots, juggling eggs etc.) can be done without losing time. This is again demonstrated in the YouTube link above, which I think is probably the best reference for what a good future GBA YI TAS should look like. I would be interested to know if the left+right running trick is possible on GBA, as it would obviously save time everywhere, especially if the runners do not want to add in all the egg antics. I also noticed some additional things on individual levels. 1-1: You are aware that running on slopes loses time. Why then do you land on one just after the first piranha plant? There are some other small examples of this in other places in the run, but I won't point them all out. You are also aware of the slope trick to increase Yoshi's running speed. It's not clear then why, for example, you jump over the first few slopes in the level, which is the ideal place to increase your speed. 1-2: A wall jump would save time climbing the wall. It's also not clear why you used the 2nd pole for the extended flutter rather than the first one. 1-6: The first cave in this level looks strange. Why do you bounce twice on the first arrow lift? Also, is it really not possible to avoid the flutter after the big bounce? I understand that you want to avoid red coins, but this looks suspicious. Wall jumps might allow you to reach the warp faster like in 100%. The final cave would also obviously benefit from a wall jump. The other thing to note is that 1-6 actually has a 1-1 warp on SNES. See here: https://www.youtube.com/watch?v=gNft0w9BFQA . It would be interesting to know if it's possible on GBA. 1-7: Are melon bug boosts not possible on GBA? (See 4-5 on the SNES 100% TAS). They can be used to get large boosts in 1-7 on SNES. Waiting for the flower at the end also seems strange. I would imagine that just collecting a few more flowers and getting the bonus on a more favorable goal ring position would save some frames. 1-8: The key pot mechanics look a bit different on GBA (eggs seem to affect it less?), but the obviously question is why didn't you hit the pot with more eggs? These could have been brought in from earlier levels. It seems weird to ride the 2nd arrow lift in the final room. Why not just spit it out and jump on it? The boss fight also looks quite slow. Is a 1 cycle not possible like it is on SNES? It looks as though you could have benefited from having more eggs. 2-1: A wall jump would probably help finish the big climb faster. 2-2: Wall jumps would be useful throughout this level, for example to climb the wall near the spikes in the first room directly rather than having to go over the pole at the top, which is more of an RTA route. They could also be used in the 2nd room to avoid the 2nd big spring. 2-4: A wall jump would avoid the need for extended fluttering in the 2nd room. Also, there is a trick with the red switch that doesn't require the extended flutter, which you can see in this RTA run https://www.speedrun.com/sma3/runs/m38powqy . The final room can probably be done faster either with chaining wall jumps, or at least turning around faster (e.g. by holding down). 2-5: You should check if using a shell to boost over the walls near the start is faster than taking the lower path or not. 2-6: You could use an egg to clear some foam near the start. This lets Yoshi fall a bit before initiating the ground pound. An upward shot on the door feels like a weird choice. Surely a neutral egg shot is faster there? 2-7: The room with the falling blocks looks slow. I'm not sure whether it's possible to get underneath both or not like it is on SNES, but at the very least it's preferable to skip the 2nd one rather than the first. See the RTA I linked for 2-4. Their route looks better than what you did. 6-8: The Bowser fight just feels very slow. I dunno how different it is to SNES, but a few basic things I would look into are 1) can the 2nd rock sequence be skipped with successive shots like on SNES? 2) In general, we found it preferable to chain shots together. This reduces the time Bowser spends stunned, which means he runs back to you sooner and is therefore accessible again to egg aiming angles. Overall, I think this TAS is not a bad effort, but could probably be substantially improved both in terms of speed and entertainment.

Posted: 6/13/2026 12:28 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Time for a bump! I got back into TASing again recently and feel inspired to put together a new Yoshi TAS. It won't be as ambitious as the 100% run, but there are still a bunch of categories for this game that haven't really been touched by TASers. I think the any% category landscape (in my vision at least) looks something like this: ACE = Masterjun's current credits warp TAS No ACE = The older left+right warp TAS. This was obsoleted by the credits warp, but the old route still has some interesting potential improvements involving the null egg glitch. It could be an interesting run, but it doesn't exist atm. No ACE, No left+right warp = null egg categories. There are actually two variants of this run. One involves save corruption to unlock 6-8 early and one goes through every level (this one is usually called magical journey). Neither of these two currently exist. No null egg = Warps. This run is what Carl Sagan's existing run attempted to be. At this point in time, however, it is massively improvable, in particular due to the recent discovery of a plethora of new warps involving tongue glitch + level corruption. Carl's TAS is now beatable by a good RTA run, although nobody has achieved all the new warps yet in a single run. This TAS would be super interesting, but is also very complex. A good project for a motivated TASer! Here's a demo run (real time with savestates): https://youtu.be/gQUeuSv6vcE?is=YdQov552-Dg7lnIL No warps = Warpless. This is the most popular speedrunning category and most resembles vanilla gameplay. It would be similar in spirit to spezzafers old run, but with more modern optimizations (e.g. wall jumps). This run actually has some interest, but the person involved would welcome collaborators ;) I'm mostly interested in the null egg categories as they're relatively short and show off some nice glitches. There is also interest in putting together some sort of playaround/glitch showcase run, as there are a load of ridiculous glitches that don't really get showcased in the mainstream categories. Anyway, if anybody is still interested in this game, either as a TASer or just a fan, feel free to reach out on discord. The community has had a bit of a revival this year.

Posted: 1/26/2023 12:34 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I think when it comes to ROM hacks, there's usually some expectation that the run will involve gameplay that is substantively different to the original game. I haven't watched the other TAS of this game, but going by this run alone, I didn't see anything that hasn't been done before. It looked like basically just holding right + jumping through a couple of levels that look more or less identical to ones from the original game. Of course I don't blame HappyLee for that... I find the original SMB TASes interesting for historical / nostalgia reasons, but I've never really found SMB to be a very exciting game for speedrunning. Therefore, although I'm sure the run is optimised, I decided to vote no. As for the whole debate about passwords/the rules etc. (which admittedly I haven't fully read), with respect to HappyLee, who is clearly a great TASer, I don't really find the run interesting enough for the discussion to even be very meaningful. I don't really know how the site works, but I would hope that the rules are as they are and continue to evolve so as to accommodate the most entertaining content, and to weed out runs that are obviously at odds with the site's mission. If a new run came along that was met with high praise, but was in conflict with some detail of one of the rules, I'm sure there would be majority support for amending the rules to accommodate the new run. I don't think this run is anything like that, and even if you could win the argument that this is the `true any%', having it obsolete the full game TAS would seem like a hollow victory to me.

Posted: 11/18/2022 8:05 PM
(Edited: 11/18/2022 8:06 PM)

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I have a very different method for solving the problem. It’s by no means as simple or elegant as FractalFusion’s approach, but I think it’s pretty cool and seems to get the right answer. I don’t know anything about martingales or Markov chains, but I do know a bit about scattering theory in physics, and I was able to translate the problem into a scattering problem. Apologies in advance as this will be a bit long and probably not very straight-forward to follow. Imagine a network of nodes and links that looks something like this – 1 – 2 – 3 – 4 – 5 – 6 – 7 – 8 – The numbers (nodes) represent how many coins you have, and the dashes, which connect different nodes are going to allow you to move left or right within the network. The physics problem I’m imagining might be something like a series of optical waveguides (lines) with connections (numbers) at which light can scatter (either forwards or backwards). The idea is that we’re going to inject light into the network from the left side and find the probability (or intensity) of the light that makes it to the end. Three things can happen at each turn in the game and each one corresponds to a different type of scattering in the network: lose a coin (p = 5/9) = scatter left in the network gain a coin (p = 1/3) = scatter right one node gain two coins (p = 1/9) = scatter right two nodes (more on how this works shortly) To model the light in the network I’m going to suppose there are 4 different channels in each connection, a, b, c and d. Channel a is for all light that propagates to the left. Channels b, c and d will all be for right-propagating light but have different functions with respect to the game. Channel b will account for the case of when you win 1 coin (move right one node), channel c for when you win 2 coins (move right two nodes), and d will let me ‘skip ahead’ so that the light I initially inject into to the system can straight to go node 5 (since the game starts with 5 coins). The light in each link can thus be described by a vector of four components and scattering at the nodes can be described by setting up 4x4 scattering matrices ( https://en.wikipedia.org/wiki/S-matrix ) at the nodes. I need three different scattering matrices: one for before the entry point (node 5), one for at the entry point, and one for afterwards. I'm going to use plain variables for channels on the left side of a node, and a prime for channels on the right side of the node. For nodes 1, 2, 3, 4 we have the matrix equation (forgive the formatting) a 5/9 0 0 5/9 b b' = 1/3 1 0 1/3 c c' 1/9 0 0 1/9 d d' 0 0 1 0 a' The way this works is that light injected into the skip channel (d) stays in there and won't be scattered into any other channel. This channel acts independently of all the others. Light that comes into the node from channel c on the left all goes into channel b on the right. This ensures that all the light that scatters into channel c from the left effectively skips ahead one node (so it corresponds to winning 2 coins in the game). At the entry node (node 5) we have a 5/9 0 5/9 5/9 b b' = 1/3 1 1/3 1/3 c c' 1/9 0 1/3 1/9 d d' 0 0 0 0 a' which is more or less the same matrix, but now light coming from the skip channel gets evenly distributed among all the other options, which means light is now entering into the network at this point. For nodes to the right of the entry point the 'skip' channel becomes useless and we just have a 5/9 0 0 5/9 b b' = 1/3 1 0 1/3 c c' 1/9 0 0 1/9 d d' 0 0 0 0 a' There may well be a simpler way to set everything up, but this approach seemed fairly intuitive to me. Now we need to get the scattering matrix for the entire network so we can relate the channels at the far left to the far right. Then we can just inject light with intensity 1 into the skip channel at the left end and see how much comes out at the right end. One way to do this is to convert each 4x4 scattering matrix into a 4x4 transfer matrix (see for example http://assets.press.princeton.edu/chapters/s8695.pdf ). You can work these out by essentially finding the 'partial inverses' of the scattering matrices ( https://en.wikipedia.org/wiki/Partial_inverse_of_a_matrix ), although you need to be a bit careful with the matrix block structure. Once you have the transfer matrices you can just multiply them together to get the transfer matrix for the whole system, which can then be converted back into a scattering matrix. If I use M to denote transfer matrices, the transfer matrix for the whole system is just M_right ^ 3 * M_entry * M_left ^ 4 Convert this back into a scattering matrix S and finally compute S * [0, 0, 1, 0]. My little Python program outputs [a, b', c', d'] = [0.45451636, 0.45483641, 0.09064723, 0] Finally, since the network stopped at node 8, we need to remember that I win if I'm in either channel b or c. b is light that entered 9 directly from node 8 (you won 1 coin when you had 8) and c is light heading towards an imaginary node 10 from node 8 (you won 2 coins from node 8). So we just need to add 0.45483641 + 0.09064723 = 0.5454836405947694 which seems to be at least close to the right answer.

Posted: 2/4/2022 1:26 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I’m not sure if any of this really answers the question, but here are some random thoughts. I’ve always found questions like ‘what is the value of 8 / 2 * 4?’ a bit artificial. I think the problem is that, after a certain point, these kinds of expressions don’t really occur in the mathematics we usually do in school or beyond. When I (and I assume this is true for most people) actually sit down and do numerical or algebraic calculations by hand, I never use symbols like / or ÷. Instead, I tend to write everything using fractions and, if necessary, parentheses. Since a fraction is written vertically, there’s no ‘horizontal ambiguity’ when reading an expression from left to right; it’s clear that the top and bottom of the fraction are on equal terms and I never need to think about the order of operations (which, to be honest, I don’t remember anyway). When I see a question like ‘what is 8 / 2 * 4?’, my first reaction is something like ‘tell me what language you’re speaking first and then I’ll do the calculation for you’. I may even ask for clarification: 'do you mean (8/2) * 4 or 8/(2*4)?'. Like with Masterjun’s English sentence example, I don’t think it’s the responsibility of the listener to have to resolve ambiguities in the message they receive. It should be the responsibility of the speaker to produce a clear, unambiguous message in the first place. I suppose that someone might respond by saying 'the expression is already unambiguous if you follow the standard order of operations'. The only scenario I can think of where an expression like 8 / 2 * 4 would occur in my life is in computer code (or writing a message on a forum I suppose). In this case, given a particular programming language, we can appeal to the rules of the language to parse the expression unambiguously. I don’t have a lot of experience with different languages, but I assume these rules are consistent across most mainstream languages and also align with the order of operations we learn in school. As for why the rules should be as they are, I think for me the main reason is that the alternative seems confusing. If a division sign were to incorporate more than the numbers or expressions immediately before or after it, then things like 1 / 3 + 4 / 2 would quickly become very confusing for me. In practice, I get a bit paranoid about fractions as they are and I’m not afraid of redundant parentheses. If I need to write a fraction with something other than just numbers, I tend to write things like ()/() before filling in the contents of the parentheses, safe knowing that anything else I write in my line of code isn’t going to be affected by this particular use of the / sign. To summarise, I guess I don't really care about the question in the context of mathematics, as it's one I haven't had to worry about in over 20 years. I'd be happy to go with either result depending on what the person asking the question prefers, but I may heavily caution them against interpreting 8 / 2 * 4 as 8 /(2 * 4), as I think they are likely to run into problems doing more complicated calculations. In the context of computer programming, well just stick it into Python and see what comes out.

Posted: 11/18/2021 8:58 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Dacicus wrote:

NxCy wrote:
Hence a(n,p) = (p+1)*F(n-p-1) for p < n-1.
I believe that there are errors in the derivation of this formula. If you take n = 5 and p = 1, for example, it yields a(5, 1) = 2 * F(3) = 4. But that Wikipedia link shows that a(5, 1) should be 5.

Yeah, you're right. I forgot to include cases where the 1s are mixed between the numbers in the composition without the 1s (in your example my argument misses 2+1+2). I'll have to think about it some more.

Posted: 11/17/2021 2:42 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

This turned out to be harder than I thought. I wouldn't be surprised if there's an easier way to do it, but here's what I found. If we call the duration of an eighth note 1, then a bar has duration 8. Because of ties, it's possible to construct notes of any duration from 1 to 8. Suppose first there are no rests, then a rhythm will be an ordered sequence of notes whose durations sum to 8. For example, four quarter notes would be 2 + 2 + 2 + 2 = 8. These are the compositions of an integer https://en.wikipedia.org/wiki/Composition_(combinatorics). The number of compositions of n is given by 2^(n-1), which is the number of rhythms that don't include any rests. To deal with rests, note that rests can't be tied (as there is no sound). If we look through all the compositions of n, we can replace every instance of a 1 (which corresponded to an eighth note) with an rest of duration 1 to get a new rhythm. We only need to do this for 1s though: replacing a 2 with a rest of duration 2 would give the same rhythm as the same composition with the 2 replaced with 1+1 and both 1s replaced with rests, which we have already considered. If there are p 1s in a certain composition, all choices of notes/rests for the 1s gives a total of 2^p rhythms. The total number of rhythms of duration n is hence given by sum from p=0 to p=n of 2^p * a(n,p) where a(n,p) is the number of compositions of n containing p 1s. Now we need to find a(n,p). It's easy to see that a(n,n) = 1 and a(n,n-1) = 0. For p < n-1, a composition of n containing p 1s contains precisely p 1s and a composition of n-p that does NOT contain any 1s. Since order matters, there are (p+1) choices for the locations of the 1s. The number of compositions of n-p not containing any 1s turns out to be equal to F(n-p-1) where F(n) is the n'th Fibonacci number (F(0) = 0, F(1) = 1, F(n) = F(n-1) + F(n-2)). Admittedly, I got this part using a recurrence relation I found by google searching. Hence a(n,p) = (p+1)*F(n-p-1) for p < n-1. Finally then, the number of rhythms of duration n is given by 2^n + sum from p=0 to p=n-2 of 2^p * (p+1) * F(n-p-1) When n =8 I get 1257. This does include the rhythm containing only rests (i.e. silence). If you don't count that I get 1256.

Posted: 11/16/2021 11:34 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Dacicus wrote:

FractalFusion wrote:
Just for clarification: Is a tie between two eighth notes equivalent to a quarter note, or are they distinct? (similarly for other ties of two or more notes)
They are equivalent.

I also have a question. Are two rhythms distinct if they're related by a shift/delay in time? For example, if I have quarter notes on beats 1 and 3 (rests on 2 and 4), is that distinct from having quarter notes on beats 2 and 4 (rests on 1 and 3)?

Posted: 10/28/2021 12:32 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

You beat me to it :P I did it more or less the same way. Let p be the golden ration. Starting from 5.12.3 here https://dlmf.nist.gov/5.12, I made the substitution t = z^p. Then setting a = 1/p, b = p - 1/p = 1, you get 1/p * B(1/p, 1) = I, where I is the desired integral. Since B(1/p, 1) = p, it follows that I = 1.

Posted: 9/30/2021 3:24 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

This was quite fun indeed, although finding a geometric solution probably took me longer than it would have taken to just do the differentiation :P The function f(x) can be thought of as the sum of the magnitudes of the vectors (x+2, 3) and (x-4, 5). If you draw a picture and think about it, you'll see that minimizing f(x) is equivalent to the following problem:

what point C on the central horizontal axis in the figure is such that AC + CB is minimal? This is solved by reflecting the triangle through the axis and constructing the straight line from A to B' (purple line in the diagram). In this case, the angles to A and B from C are equal, which means (4-x)/5 = (2+x)/3. Solving this gives x = 1/4, which is the same as what Dacicus found. edit: just realized my solid/dash styling is inconsistent, but I can't be bothered changing it now. Sorry!

Posted: 7/12/2021 6:01 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I found by experimenting that the final two digits of powers of 7 cycle through four different values: 01, 07, 49 and 43. Therefore we need to reduce the exponent of the bottom 7 modulo 4. This can be done using an application of Euler's theorem. Basically in reducing numbers mod 4 you can reduce powers mod phi(4) = 2. The exponent of the bottom 7 is itself a tower of 7s. Writing it as 7^(7^...^7), clearly the thing in brackets is odd, so 7^...^7 = 1 mod 2. Therefore, modulo 4 we have 7^(7^...^7) = 7^1 = 3. This was the power of the bottom 7 of the original number and 3 corresponds to an ending of 43.

Posted: 5/31/2021 2:29 PM
(Edited: 5/31/2021 2:33 PM)

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

I also don't think this category makes a lot of sense. A natural comparison can be made with the recently published minimum A presses run, as both runs restrict the use of the same button. For the most part this run looks like a normal any% run, but Mario has to keep doing wall jumps to climb up pipes and walls. For me this doesn't add anything interesting. The game ins't traversed in a new, innovative way; it's just that an extra step has been added to otherwise regular movement. I felt like the minimum A presses run required more of a rethink of how to get around different obstacles and pushed the game more to its limits with the strategies involved. I don't think this category is better than minimum A presses, and I don't see that it is different enough from other runs to warrant publication as a new category. Therefore I voted no.

Posted: 5/6/2021 1:08 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

chillex_drsado wrote:

I appreciate the compliments and the thanks! As for that level, there are many gaps and enemies that Diddy can't really scale himself in a way that's faster. While most of them should be minor, the biggest timeloss as Diddy, as well as the dealbreaker for me was the enemy at 13:25. To progress past that as Diddy, I had to both swap at the beginning, which itself takes up 1.25 seconds, and wait for the enemy to be in a position where I could jump over it. Dixie's hover ability also allows me to take full advantage of her roll speed and skip over krockhead platforms, whereas I couldn't do it with Diddy. By the time I had progressed far enough into the level, it became less worth to do the rest as Diddy because I could goal swap at the end. I do plan to write more tidbits in my level-by-level breakdown tomorrow. I have some errands I need to run, so it won't be today.

Thanks for the response. Just out of curiosity I tried playing around with the level I mentioned a bit with Diddy and I think you're indeed correct. I also didn't realize when watching that those levels have ice physics and so some of your movement that I thought looked odd actually makes sense now :P

Posted: 5/6/2021 12:55 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

This was a great TAS. Although I really enjoyed the first stage, I was wondering if there would be much more to the run or if each subsequent stage would just be more of the same. Thankfully there was a lot of variety overall and I think the length was just about perfect: long enough so that you got to show off a lot of cool ideas with each stage feeling different, and short enough so that it didn't have a chance to feel repetitive. My favourites were probably stages 7 and 10. Easy yes vote. Thanks for the run!

Posted: 5/5/2021 8:51 AM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Acumenium wrote:

I'm not sure why this is so hard to understand. Right now, it's not a challenge, because it's a TAS, there's no such thing as difficulty. At the same time, the goal is bleakly arbitrary. This isn't relatable. The last sentence is also hypocritical. If I have no right to an opinion or right to express it, neither do you. You're accusing me of a bigger podium than I have by doing the same thing you are accusing me of.

It's hard to understand because you say things like "But this isn't even a TAS" and then "because it's a TAS". In any case, I don't really see hypocrisy in what I said. Whether or not something is a TAS, insofar as it can be considered publishable on TASvideos, is decided by the community (or at least by the judges). I wouldn't claim that a movie shouldn't be published because it isn't a TAS unless I were confident that my saying so is consistent with the community's standard of what a TAS is. I think I've already explained why I'm not impressed by the word 'arbitrary'. As for relatability, I personally think most of the ACE movies are extremely unrelatable and uninteresting, but I think they deserve to be published.

Acumenium wrote:

tl;dr: In a vacuum, this run doesn't have an issue. It isn't very interesting to me but it doesn't have to be. It being accepted represents serious flaws in the publishing system as a whole. It should've been immediately rejected given what many other runs and branches have to put up with, but it wasn't. It was rather accepted for every reason those types of runs are rejected.

Perhaps this what really mattered to you all along. I only initially responded because I felt comments about how this movie isn't a TAS or is arbitrary/unrelatable were unfair when discussing this run in isolation. I won't comment on broader points about the site's publication system because I don't really have an opinion on it (I'm more of a casual viewer these days), but you should certainly bring it up in the appropriate place if you have an problem with it.

Posted: 5/3/2021 12:06 PM

NxCy

Experienced Forum User, Published Author, Active player (340)

Joined: 11/19/2007
Posts: 139

Warp wrote:

NxCy wrote:
Yes, and I think we can use the same example. f(x) = exp(-1/x^2) g(x) = x^2 Then lim(x->0) [ f(x)^g(x) ] = lim(x->0) [ exp(-1) ] = 1/e but [ lim(x->0) f(x) ] ^ [ lim(x->0) g(x) ] = 0^0 is undefined
I don't think that's a valid argument because you could use the exact same argument to argue that the rule lim(a/b) = lim(a) / lim(b) is invalid (even though it's one of the basic laws of limits).

It's been a while since I've studied analysis, but that rule you mentioned has some assumptions behind it. I believe the simplest version goes: if lim(a) = A and lim(b) = B where B is not equal to 0 then lim(a/b) = A/B. It can probably be extended to cases where lim(a) or lim(b) = infinity, but you have to be careful with indeterminate forms. As for one function to the power of another, we have lim(x->c) f(x)^g(x) = lim(x->c) exp[ g(x)*ln(f(x)) ] = exp[ lim(x->c) g(x)*ln(f(x)) ] where I've probably assumed f is positive and used the fact that exp is a continuous function. If the limits of g(x) and ln(f(x)) are nice then we can use the limit product rule to get the limit to be exp[ g(c)*ln(f(c)) ], which then unravels to get f(c)^g(c) as you wanted, but I think it's clear that there are several things that can go wrong, for example the limit of f at c may be 0, in which case you would end up with ln(0) and (at least the simplest version of) the limit product rule no longer applies. So I think the answer is 'sometimes yes, but not in general'. If anyone has a more clear-cut answer that'd be nice.

Posts for NxCy