What I do know is that based on what people are saying in this thread, we are to ignore anything from the point of the payload onwards. At which point we're left with 97 out of 99 seconds which are identical. Those last two seconds are utterly boring and do nothing to add to the entertainment factor.
I think your interpretation of "we are to ignore anything from the point of the payload onwards" is way different from mine.
I think a comparison could be made with movies which end input early but still complete the game -- in particular, monopoly 4 cpus which takes 1 minute to set up a situation which will cause all 4 cpu players to lose within the following 1.5 minutes. Obsoletion of this movie could be done based on the time-to-last-input, but you'd be mad to make an encode that stopped at the last input because the resulting movie wouldn't show the game being beaten! In the same way, you'd be mad to make an encode of this movie which stops at 99 seconds, because it would entirely cut off the section of play which showed what the payload is and what it does.
For a total control TAS, I would say the goal is to deliver a payload that executes arbitrary code as quickly as possible, and to deliver as entertaining a payload as possible. (Yes, these goals may conflict.) Obsoletion in terms of delivering a faster payload should be done based on the input up to the point where the payload is delivered; but in order to understand the implications of the payload, you need to present the viewer with the resultant behaviour. For interactive payload, that means that some extra input will be necessary to demonstrate the possible interactions.
This run is identical to the existing SMW any% for the first 97 seconds. However, it is perverse to ignore the differences between them after 99 seconds in terms of entertainment value, because the post-99 second section of the video is where the goal is achieved: beat the game, or demonstrate the payload.
The implications of this line of thinking:
* a TAS which exhibited total control of SMW in 92 seconds would be a strong candidate to obsolete this one
* a TAS which simply played snake or pong better would not
* a TAS which took the same time to deliver a different payload, would have to be judged on its entertainment and goal merits relative to this one
There is also a significant unresolved issue which has been hinted at but which I'd like to try to state precisely: once a game has a total control glitch possible, total control may become the fastest way to beat the game. Is total control to execute an arbitrary payload a separate category from total control in order to beat the game? We currently have pokemon yellow any% and pokemon pi day which perform very similar activities for the first minute each, but end up delivering very different payloads and result in very different movies, despite the initial similarity.
I think there is value in a separate "arbitrary code" branch from an "any%" branch, because when a total control glitch is used to beat the any% category, it's not clear to the viewer if the glitch used allows arbitrary code or only limited glitchy behaviour (such as Mega Man's DelayStageClear glitch). Having the separate category demonstrates categorically that this is an arbitrary code glitch.
Big golem : I have to use blocking mail armor to be able to dodge the first hit and be invulnerable for few frames. Allow me to use Black panther soul almost the all time. Ascalon is the best weapon to increase strength of souls but I have to switch to claim’s solais at the end because the weapon does cancel the animation when soma get back on the ground.
You can cancel heavy weapons with backdash. Might be more entertaining than another pausescreen.
So my question is: How fast would a wheel have to spin in order for its outer edge to age, let's say, 10% slower than its surroundings? (In other words, for every 10 seconds that pass, the edge of the wheel only ages 9 seconds.)
I don't know the answer, but your question prompted me to find the relativistic rocket which just blew my mind.
My only contribution is that you probably need to consider the acceleration (from centripetal force) as well as the velocity. But I could be wrong.
PJ_Boy made me think of an interesting physics question.
If you derive acceleration, you get the rate at which your acceleration changes. (You could do this nearly infinite times, until you hit a constant, to which the derivative would be zero).
If you integrate acceleration (m/s^2), you get velocity (m/s), i.e. the summation of your acceleration over an interval.
If you integrate velocity (m/s), you get distance traveled (m).
If you were to integrate distance... you'd get (m * s). Perhaps I'm just too far out of physics... but that doesn't seem like it would have any meaning.
Integrating distance feels strange because you don't normally have situations where distance is a function of time. Velocity and acceleration are often a function of time, but distance?
Think of a particle on the x-axis whose postion at time t is x(t). Then the integral of x(t) is the function whose derivative is x(t); then the definite integral from a to b of x(t) dt is (b-a) times the average position of the particle during that time.
Derakon wrote:
Let's say you're in a hybrid automobile at the top of a hill. Your vehicle can convert velocity now to velocity later at a given efficiency rate of A (that is, e.g., if you slow down 5m/s now, then later you can speed up by 5Am/s), with A of course being less than 1. You also have to deal with air resistance, which as usual goes up with the square of velocity. Let's say your drag constant (area * drag coefficient) is B (see Drag Equation).
The question is, when driving down the hill, do you try to conserve as much kinetic energy as possible (i.e. don't use the brakes; just coast), despite increased losses due to air resistance? Or do you brake, losing some energy in the conversion process but also losing less due to air resistance?
Just to be specific, are we optimizing for total distance or for some other metric (time taken to reach a target, say)?
Okay, another suggestion, to add something to the discussion:
Freakonomics - Levitt & Dubner. Actually mostly about statistics and not that much about economics. How can we tell from match results that sumo wrestlers cheat? How does laws on abortion affect crime rate? The book is a bit uneven, if found the two chapters about the questions I mentioned the most interesting.
disagree that it's not about economics. Economics encompasses the study of people's behaviour under various sets of incentives - which is why nash got the nobel prize in economics for game theory.
This one seemsa bit...easy?
By symmetry, AD = BD = DE. Also arc AE = BE. So AED is an isosceles right-angled triangle, and angle AED is 45 degrees.
Alternatively, the angle between arc AE and line DE is 90 degrees. Not sure what exactly you were asking for so you get both.
I've heard very good things about marcus chown - I haven't read him, but I've seen him talk (at an event where simon singh was also talking) and out he's half as good a writer as a speaker then his books will be brilliant.
Bad science by ben goldacre is a perennial favourite. A bit more focus on medicine, quacks, how people misrepresent science and how science can actually determine something to be true. In a similar but more maths focused vein, there's innumeracy by john allen paulos.
A good but heavy tome is godel, escher, bach by douglas hofstadter - it's very difficult to describe because it visits so many topics, such as genetics, fractals, canons and fugues, mathematical logic, and many others; it unites these into an overall theme of symmetry and self-similar patterns.
autopickup is part of the game. it's not cheating at all. if you want optimal play, you have to use it, because it lets you pick up items without spending turns to do it -- it might be considered exploitative, but it's definitely not cheating.
Furthermore, what the NetHack community considers to be "cheating" differs wildly from what the TASing community considers as "cheating". Savestate abuse, for example.
As I understand it, only by changing which RNG values will generate souls. Here's my understanding of the "impossible soul drop" problem: a soul can be either 1 or 2 RNG outputs away. When Soma swings his weapon, the RNG advances by 1 if he doesn't shout, or 2 if he does (presumably to decide which shout he uses). If the soul is 1 RNG away and he doesn't shout, or 2 and he does, then all is well. However, if it's 1 away and he does shout, then you end up skipping the desired RNG value.
This is exactly right; when you're 2 away, you won't shout, and when you're 1 away, you will, so you can't get the soul either way.
It's perhaps overstating the case slightly to call these "impossible" drops; more just "awkward"; this is because you're not forced to have the swing happen so close in time to the monster dying; you could instead swipe much earlier at thin air, then move your extended blade into the monster after external factors have scrambled the RNG appropriately. Nevertheless, this strategy is often not acceptable because it will cause delays. (Perhaps cooldown could be managed to ensure a certain amount of "slack" time on the killing blow?)
Learn the assembly language of the system you want to TAS on. For example, GBA uses ARM Thumb instructions.
Thanks, rhebus. I've never programmed in ASM before, so would ARM Thumb be a good start, or should I start with PC architecture and then shift over after I know it?
The first asm you learn should be the first asm you want to use. If you have no reason to use PC asm, don't learn it. Find a game you want to TAS, and learn the asm for that platform.
DarkKobold wrote:
I think this path will find you very discouraged, as a warning. Attempt to TAS the game with just ram watch, and at most lua.
I partly agree with this viewpoint, because reading disassembly is not for the faint of heart. However, if you have a specific goal, and you have either good skills or good support (by asking questions here) or both, then you can get quite far by trying to investigate specific questions. For example, I recently discovered the algorithms for item drops and soul drops from Castlevania: Aria of Sorrow; this was entirely done by disassembling the code, using vba-sdl. It took bloody ages, but it was worth it. Having a specific goal - the soul drop code - helped keep me focussed and keep my mind on the task at hand.
Most of my AoS discoveries, however, are by RAM search rather than by disassembly. RAM search is far, far easier, and can tell you much about the game.
Learn the assembly language of the system you want to TAS on. For example, GBA uses ARM Thumb instructions.
ARM Thumb has the advantage that it's a rather nicely designed assembly language -- orthogonal instructions, no implicit arguments, general-purpose registers. Having said that, the most important thing is that you're TASing a game you love, and that will drive your desire to learn the assembly language on that platform.
It will also help to get a copy of an emulator with an assembly debugger in it, so you can actually step through the assembly code as it executes. I have done this with GBA before; I don't have time right now to find the version of VBA which allows this, but remind me and I'll do it later.
Awesome stuff, Kriole! I've just watched your last 3 WIPs. I really enjoyed your arc demon fight, and the triple soul grab in the flame demon corridor. Also, nice improvement to the legion fight! More action-packed wthout those pause screens.
Looking forward to Graham!
There are a lot of fast algorithms which will create pseudo random input, but it's extremely hard to create a shuffling algorithm that will give an even probability for all of the distinct possibilities.
Oh come on, Fisher-Yates is not that hard; it's not like implementing cryptography or mersenne twister or something like that.
You did miss shots in both driving stages, any reasons for this?
The Terminator's actions may change depending on when and how you ram him, when you shoot bullets (missed bullets also affect him, making him sink farther away down the screen than usual)
Both observers can see the light coming from (the same) distant stars located in directions orthogonal to their relative motion, so their observable universes overlap. Their interaction universes will also overlap in a similar manner, but in the future instead of in the past; but the two observers' will be outside each other's observable and interaction universes. The observable and interaction universes may overlap in some regions and they may be disjoint in others.
I didn't quite understand the explanation. If a star is observable from both A and B, couldn't an observer at the star transmit information from A to B (which ought to be impossible)? The star can observe light coming from A and rely the information it gets to B.
A point in the observable universe is a point in space-time, not simply a point in space. Since as marzojr said, the observable universe of A is defined as being every point which could have affected A's current state, then a point in A's universe is a point in A's past. As a result, A cannot transmit something to the star in A's observable universe, because this would involve sending the signal backwards in time!
A' B'
\uAB /
\ /
uA \/ uB
/\
/ \
/ u \
A B
In the diagram above, A is A at some point in the past. B is B at some point in the past. A' is A some time later; B' is B some time later. uA is the universe that A can affect. uB is the universe that B can affect. uAB is the overlap. u is the universe that neither can affect.
A can affect B' but not B.
The same diagram can demonstrate observable universes: uA and u are the observable universe of A'; uB and u are the observable universe of B'.
Even you ought to acknowledge that there's a very substantial difference between the two-slit and the three-slit experiments, compared to dropping different objects whose masses differ in some micrograms.
Actually, I think this is the crux of our disagreement. I don't think there's a substantial difference, given everything else we know about electromagnetic radiation. In particular, none of the observations from our use of multiple radio transmitter broadcast has resulted in anomalous results.
I know next to nothing about quantum mechanics, but I find it curious and interesting why the particle behaves like it passed through two of the slits even if there are more available.
I don't think this is an accurate description of the situation, from what I can see. The particle acts like it went through all three slits. The resultant interference pattern is a linear combination of the single-slit and double-slit experiments for each slit and pair of slits in the triple-slit experiment. This isn't a surprising result.
EDIT: the abstract of the paper on arxiv says "Born's rule predicts that quantum interference, as shown by a double slit diffraction experiment, occurs from pairs of paths." - which I think is basically saying that interference patterns occur as linear combinations of individual waves. In effect, it's saying that A+B+C == (A+B)+C == A+(B+C) - ie a three-way interference is equivalent to a two-way interference plus an extra wave, and doesn't behave in any "special" way. It's *not* saying that the particle only goes through two slits.
EDIT2: wait, my physics teacher in secondary school did an analogous experiment - in the late 1990s, I might add. She shot a laser into a diffraction grating which is basically a multiple slit experiment. I expect physics teachers everywhere are showing their kids this supposedly amazing result.
One would think that an experiment that confirms a prediction made by a scientific theory is extremely interesting and useful.
I just dropped some toast on the floor. That confirms a prediction of gravity. When do I get my PhD?
I was, rather obviously, talking about confirming a prediction that hasn't been confirmed experimentally before.
That's a nebulous concept. What counts as a prediction which hasn't been confirmed experimentally before?
I just dropped a piece of toast weighing exactly 82.569321 g on the floor. It hit the ground. That confirms a prediction of gravity which has never been confirmed before.
Of course, nobody would expect that a slight variation in toast mass would change the results significantly. The point is that although the triple-slit experiment had not been confirmed before, enough other, related experiments had not shown any anomalies to suggest very strongly that we already knew how it was going to end up.
For the triple-slit experiment to produce anything other than the result it did, there would have to be a new kind of radio wave interference operating on a three-way basis independent from a superposition of two-way interferences. In other words, interference would have to result from something other than a linear superposition of component waves. But radio transmitters already operate using exactly this assumption, with transmitters in numbers of far greater than three, and we haven't run into any problems yet of three-way or higher-way nonlinear interference effects. I would therefore say that the results of this experiment have been indirectly confirmed already. But as I said, it's a nebulous concept.
The kind of experimental observation of a theoretical prediction which is interesting is one which hasn't previously been observed in any form - such as the finding of antimatter after Dirac predicted it, or the confirmation of special relativity by meson atmospheric penetration. These experiments demonstrated that a whole new section of a theory was viable. The triple-slit experiment has far more in common with my toast experiment than with these great experiments.
(It turns out there was no nearby seed that let us bring all five wraiths down nearby, so we get four of them, and one of them gets away. We might nab it later, or gain the 14th level via polyself rather than wraith; there are plenty of options.)
I thought that wraiths can still leave corpses on graveyard levels such as the Valley, it's just 1/3 as likely. You can kill and eat that last wraith in the valley for the 14th level.
Thanks for the link, things are much clearer with it.
marzojr wrote:
Sinha et al just had very good PR... and lots of hype.
The abstract confirms this. 2/3 of it is talking about unification of quantum mechanics and gravity; the final third admits shamelessly that they haven't got any closer to said unification.
moozooh, I don't know where you're getting your official advice. Advice here in the UK about bird flu and swine flu - that is, official government advice as well as advice from university scientists in an official or unofficial capacity - was basically "to prevent spread of bird flu: practise good hygiene, isolate known and suspected cases, and otherwise go about your life normally." Drugs were not recommended by the public authorities and university experts. The media did stir up a hysteria and a panic stockpiling of Tamiflu, but government and university experts alike requested people not to so that those actually suffering would be able to get treatment. Arbidol is not approved for use in Western countries, so we didn't get any of that pushed on us.
Perhaps things are completely different in Russia, but as far as what the university and government experts here recommended with respect to bird and swine flu, I haven't found their advice to be inconvenient or personally expensive.