I have to agree that a 100% run would be... at least 100% more entertaining. That's not a reason to stop this from being published though, so I have to vote yes.
After reading 2 minutes on that page, I must say I don't really comprehend how it works either.
You spent 2 minutes on an 8-page article before giving up? Even the fastest readers in the world couldn't have read it in that timeframe, much less understood it. You certainly didn't click any of the links to acquire the necessary background information, either.
But that'd explain how you come up with your theories: refuse to actually spend time educating yourself, instead substituting crazy guesses. Ignorance through lazyness. Good luck with that.
@Derakon: MLC flash actually stores multiple bits in a single charge, usually two. This increases storage density (thus lowers the price), but reduces speed and reliability.
Congratulations on using the bible as a measure of storage density o_O
It's simply flash memory, usually the cheap NAND variant. The same stuff found in USB sticks and SSDs.
Every time you post, I'm impressed by the thoroughness of your research, and your willingness to redo large chunks of the game to save a few frames. This game is receiving a treatment and attention to detail that many games only get after several iterations.
Looking forward to the results.
On a semi-related note, your upcoming TAS and my discovery of the excellent Let's Play have motivated me to do some papercraft:
The arms and hands are still missing, the hat isn't glued on because I need to find a way to cram some LEDs behind his semi-transparent eyes first, and the feet won't be attached until everything else is done, but at least it's already looking a bit like a black mage :)
I supersized the original model, so he's a good 40cm/17inch tall. Almost life-sized. ;)
I think scientists might lie about these things because they don't want to accept the idea that they could have wasted huge junks of their lives on believing a lie, instead of following the Biblical Truth. They've made this huge investment in having faith in science for years, now they are in denial, refusing to accept any of the sheer endless amount of contrary evidence to their positions. ;)
I think religious people might lie about these things because they don't want to accept the idea that they could have wasted huge junks of their lives on believing a lie, instead of following the truths found in nature. They've made this huge investment into having faith in god for years, now they are in denial, refusing to accept any of the sheer endless amount of contrary evidence to their positions.
Not sure why you even considered using this as a one-sided argument, or why you think it's bringing this discussion forward.
If universes arise from immensly complex processes, it would be very much justified to call them God from our perspective.
You can also call it Berta, but neither is a very useful label. Most religions assign more properties to their gods than just being the creator of our universe. Properties one wouldn't assign to a meta-universe by default.
Is your meta-universe sentient? Does it provide moral guidance? Does it harbor love for humanity, create miracles and react to prayers? Does it give rise to an immortal soul in each of us, judging us after death to deliver eternal salvation or damnation based on our deeds, adherence to a book and sexual preferences? Does it have a human son? Does it explain the trinity and the existence of the holy spirit? Is it omnipotent, capable of influencing our universe on both large and small scales? Is it omniscient? Omnibenevolent? Does the meta-universe need to be worshipped?
If you answered "no" to any of the above, then it's not the christian god. (Fill in your own questions for other religions.)
If you answered "I don't know" to all of the above, then it's premature to label it "God".
"This sentence is false."
Try reading a book on Fuzzy Logic someday. As soon as you leave the confines of theoretical logic, things start to get somewhat grey.
Newton's theory of gravity is fairly accurate under most circumstances, so I'd call it a good approximation. But it's not entirely correct, thus - if we restrict ourselves to true or false - it must be false. Same goes for every macroscopic scientific law, since quantum mechanics always leave a probability > 0 of randomly tunneling into a state where the law is violated. Is that what you're trying to say?
Of course, the same goes for every theological text, too. But that's what you're trying to do here, right? Pointing out a small flaw in the bible, thus labeling all of religion as false. Right?
Not trying to distract you from the programming you do, but gimp/audacity and games only compete for two resources: hard drive space and your time. You don't run gimp and a game at the same time.
Hard drive space is cheap enough that buying anything less than 2TB is a waste of money. Filling 2TB is difficult even if you wish to retain HD encodes of every published TAS ever, including multiple redundant backups.
Actually, gaming would require a windows partition, so you'll have a dedicated gaming partition set aside which won't be taken over by programming stuff.
Obsoletion is merely an issue of game choice. We both know that you don't need the latest innovations in 3D rendering to enjoy a game, and if you're willing to tone down the resolution on the more intensive games, a lot of games will run even on an outdated machine.
There's also plenty of fun to be had with older games; I'm currently playing a game that's almost 7 years old and another that has virtually no system requirements. The only thing actually taxing my computer is skyrim.
And most of the games currently released will have to run on XBox360/PS3, as long as your computer is a bit faster, those games will run fine until next-gen consoles arrive. Or if you buy a console, it'll also run fine until next-gen consoles arrive.
And in my opinion, something like 300 euros is definitely not a little money especially if invested on a regular basis.
Once every 3 to 5 years is "regular"? That amounts to 20¢ a day, which is probably less than the power bill for your current computer. (In fact, upgrading may save money in the long run, if you switch to a more efficient machine)
Is the PSX version sufficiently different from the GBA version? If so they could as well coexist, though I agree that the goal should indeed be different.
The gameplay, tricks, levels, physics, skaters and menus seem to be pretty similar, but there are two striking differences:
* The GBA version seems to be missing three levels. Instead, it got the boring warehouse level.
* The GBA lacks the 3D chip for a proper 3rd person view, instead opting for an isometric view where you just can't see where you're going. Oh, and they replaced all the nice textures with something bland that'll fit into a GBAs memory. Ugh.
While I'm actually impressed how faithful the port is to the original, the only thing it can show that the PSX version cannot are crappy graphics.
I'd vote yes on a PSX 100% run any day. But this? Meh.
Is it ok if a verification movie is over 15 hours long? (eg, verification movie is required when starting from savestate)
Maybe I missed something, but why do you need to start from a savestate?
Assuming you meant "sram" instead of "savestate", I don't think the length of the verification movie matters, but IMHO start from sram is only acceptable when using a save file that a regular player might create. If you're using weird glitches to create a broken save file, creating the save file should be part of your TAS. IIRC both the chrono trigger and symphony of the night TASes use save corruption, but they do it within their TASes, not beforehand.
I suggest writing a bot. Not only is the required luck enormous, the bot will also eliminate the requirement to remain conscious after hitting the wall to press rewind.
And you honestly have no problems with any of that? People owning other people as property, and having the right to beat them with a rod?
People do it with their children throughout the world. There are of course movements to stop physical reprimands as well as child labor laws. The two really aren't all that different, except with a slave it goes on throughout their lives.
There are so many things wrong with that statement that I'm beginning to hope you don't reproduce.
Slaves are owned for the benefit of the owner. Slaves are punished to work harder for the benefit of the owner. The slave does not benefit from being owned; if he received fair payment for his labour he'd be able to make a better life on his own.
Children are placed under the care and protection of their parents for the benefit of the child, because they can't survive alone. They're not dominated, they're taught. If they're punished, then it's meant for the benefit of the child, to learn from it.
Most important of all, children are released from the authority of their parents once the education for the benefit of the child is completed. They're released to have a life on their own. Slaves aren't.
This requires that no input from level x is affecting level y, though. Maybe someone could confirm that?
I can confirm that your assumption is wrong. As explained in the submission text, RATE does not reset between levels.
Derakon wrote:
In other words, distributed computing can help bots run faster but it won't help you make bots.
the other point was that transforming a bot into a distributed version has to be done for each bot individually. Reuse is only possible for bots that use similar algorithms, and different games are likely to require largely different bots. There's not much of a reusable "framework" you could build that goes beyond the capabilities of the existing frameworks.
Derakon wrote:
Is there any interest in improving Lunar Pool? My only complaint about it was that it occasionally made "dummy" shots just to reduce the score tally time -- a smart optimization for the overall TAS but it reduced the per-shot entertainment value.
The "smart" improvement for dummy shots would be to find a pool game with more tas-friendly rules. Surely there's more than one pool game in existence, but I haven't checked the others.
There really is no need for the creation of a distributed botting framework since there's literally hundreds of distributed computing frameworks in existence. Some are over two decades old, can be readily plugged into any C/C++ program with the flip of a compiler switch, they're high-speed, feature complete, cross-platform, work on heterogeneous networks, and they're easy enough to use to be forced onto students.
Writing another is really just re-inventing the wheel for a car you don't have.
While the last few posts have finally focused on the actual problem (writing a bot that works), there's still no suggestions on how to make writing such bots any easier. Because, frankly, there isn't. It's a hard problem that'll have to be evaluated and solved for each game or situation separately, and in most cases the evaluation phase will still end with "It can't be done in a useful way".
If I had to wager a guess, I would say that doing all of Norfair in a single trip would be much slower. The main problem would be the gold torizo and ridley fights.
Also lack of space jump, both for general movement and for reaching the gold torizo.
Cpadolf wrote:
unless there's some convenient E-tank laying around that I'm forgetting.
The closest one is the early e-tank that recently spawned in blue brinstar. It's just a room away, but still takes more time than refilling 100e at the pipe bugs. Unless you find another use for the tank or the nearby missile pack.
Do the wraparound shots require charge, or does it also work with uncharged wave?
If charge+wave is acquired early, has anyone tested wave-PB combos against phantoon to sneak in some damage before the missiles start hitting? Will he just disappear before you can stunlock him? Or could you possibly time the wave-PB combos so that they hit at the same time as the final super missile?
Can't be more than a second saved, but it may be worth trying to tip the scales in favor of early wave.
Of course, you'll have to postpone spazer until before maridia then.
For example SMB1 (not very random choice) we have pretty clear checkpoints set out for us:
Still, those checkpoints are too far away from each other. Sections longer than a second are impossible to brute force.
crollo wrote:
I guess we more or less could disable left button on the D-Pad along with Start/Select
SMB needs the left button for several tricks, like fast acceleration or glitching through walls.
Not that removing two or three buttons would help much: Warp has already demonstrated that even 2 seconds on a 1-button-gamepad isn't possible.
crollo wrote:
I was hoping to get you guys thinking in this direction in order to see if it has any potential at all.
We already did. Several times. Laws of physics won't comply with the ideas.
Writing a bot for a game is very difficult, it's a lot of work, and it takes a good understanding about programming, complexity, algorithms and data structures. And even then, it's often physically impossible.
Pitching an old idea as new and throwing around buzzwords like distributed computing or GPU offloading does not solve any of these problems.
thommy3 wrote:
[..]Kwirk[..]
You've linked the thread where a bot has been attempted. Apparently it went nowhere. Despite lots and lots of manual optimizations, the amount of CPU/ram required would have been too much on the longer levels.
Why do you think it'd work this time?
Also check out Lunar Pool. This is as simple as it gets for a bot: pick angle and force, fire. Repeat one or two times, end level. Thus you have very short levels, and only a few frames of input that matter during each level. And yet it had to use lots of heuristics to finish in a reasonable timeframe..
Stop dreaming, people. Bots are tools that may be useful in some limited circumstances. They're not the solution to TASing, nor can they replace TASers, nor will a generic bot ever be able to finish a game on its own.
One point that some people missed: RBO *has* been done on console, for example here:
http://www.archive.org/details/Hotarubi_SuperMetroid_RBO_54
He's picking up a lot more items (46% instead of 33%) and taking twice the time (0:54 instead of 0:27) though, but he is doing the bosses in reverse order and he skips all the suits.
I completely agree with hero's list and assessment.
Of all the categories, RBO is the one that offers the most radical change, and is thus the best one in terms of unusual routes, new tricks and the unique challenges of being suitless.
It's a simple goal definition that has been popular even for unassisted runners (IIRC both red scarlet and hotarubi did an RBO) simply because it's SM on Hard Mode.
Most of all, it makes for an exciting, unique run. Neither GT code nor PAL roms nor unglitched low% nor timing changes could offer much that hadn't been seen in the runs already published at the time. This one can.
Brute forcing a complete game is out of question measured in today's processing power
Even measured in the theoretical processing power of the whole universe, it's way too much. The math has been done a couple of times, brute-force-TASing will never ever work.
This means that for every single game, we need to check whether the game is suitable for botting, write a game-specific bot and run it. The specific algorithm that works for the game may or may not be suitable for parallel or distributed computing.
If you end up creating such a bot, and need more CPU power, then I'm sure there will be enough volunteers to help. But a generic framework is likely useless since we'll never have a generic bot. All we can do is handle a few special cases.
lrn2troll. You contrast killing-instructions with Exodus 20:12 and Deuteronomy 5:16. If you're trying to determine peacefulness, large-scale war and mass-murder is a lot more fun than a little capital punishment, see Deuteronomy 20. I particularly like verses 13 and 16. Also contrast with Lev 19:18, that's always fun.
The arms and hands are still missing, the hat isn't glued on because I need to find a way to cram some LEDs behind his semi-transparent eyes first, and the feet won't be attached until everything else is done, but at least it's already looking a bit like a black mage :)
I supersized the original model, so he's a good 40cm/17inch tall. Almost life-sized. ;)
