Posts for FractalFusion

Posted: 8/3/2012 1:06 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

OK, after watching a bit, I agree with a lot of people here. This run is nowhere near optimized and is very uninspiring to watch. For example, at around 2:50, there are two very obvious parts where the author could and should have easily rerecorded over, but he couldn't be bothered to do so. Not that those are the only parts; the whole run is full of parts like that. I will not go into detail unless WinEpic actively requests me to do so. Concluding comments about the submission: 1616 rerecords is simply not enough for this type of game and goal, and I do not want to see "I could probably improve my run by a few minutes, but since it was my first time TASing with DeSMuME, I wasn't used to all the tools" in a submission ever again. As for the encode, sorry to put you guys through SendSpace, but I didn't realize how obnoxious the download page was. I won't upload there anymore.

henke37 wrote:

The SHA1 checksum passes, the emulator version is correct and the indicated setting is enabled.

It may be possible that differing firmware causes it to desync. To make sure, set your firmware according to the movie file as follows:

firmNickname: yopyop
firmMessage: DeSmuME makes you happy!
firmFavColour: Turquoise
firmBirthMonth: July
firmBirthDay: 15
firmLanguage: English

(This is default firmware setting. Older versions of DeSMuME do not set firmware to movie file specification when playing back so it must be done manually. I don't know if 0.9.8 changed this.)

Posted: 8/1/2012 9:47 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Encode downloadable from Sendspace.com (~235MB). Click on link in the blue box below the filesize. Don't click on any other buttons; those are ads. http://www.sendspace.com/file/oke2ns The actual time of the TAS should be around 51:42.

Posted: 7/31/2012 7:29 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

I was hoping that more shots could be used, since it seems like shooting stuff seems faster than trying to avoid them. At a few points, it looks like you missed shots (extra bullets flying across the screen). Was there a reason for that? In the first level, it looks like you could just walk up the black slopes instead of climbing the ropes. Is there a reason you couldn't just walk up the slopes? Also, if it is possible, maybe kill off the first character to get to the second one (with full health) instead of switching him after the first level? Anyway, I like this version a lot better than the previous one. I tend to be in an optimization mood though, so I try to think of things that could make it faster. I wish that your character wasn't forced to stop and go so often.

Posted: 7/29/2012 7:19 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

To explain what AngerFist is saying, the quality of the voices in the youtube video is weird, as if nitrogenesis voiced it over himself. The voices should sound like the following: http://www.youtube.com/watch?v=JZRVsrlxHcc It doesn't help that the audio and video are desynced right at the beginning as well. I don't know yet if this is a property of FBA-RR v0.0.7, or if this is only how the encode came out.

Posted: 7/28/2012 6:49 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

NitroGenesis wrote:

Oops... the real TAS is here. http://tasvideos-nitrogenesis.googlecode.com/svn/tas-bayroute.zip

Thanks, nitrogenesis. Replaced the file. Just to let you know, changing the extension of your FBM file to .zip doesn't magically make it a ZIP file. Also, I'll give you back your table. ┳━┳ ︵╰（°□°╰)

Posted: 7/25/2012 8:05 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

gochaさんの翻訳でありがとうございました。このサイトの登録フォームのCAPTCHAを変えました（日本語と英語とフィンランド語）。日本語の登録フォームの例え：

管理者連絡リンクはadmin@tasvideos.orgに変えました。翻訳されない英語や「赤い → 赤」のなどは、今は全てを変えられないけど、　CAPTCHAの「コード先頭の5文字を入力してください」の翻訳されなかった英語が非常に大事で、　直さなければなりません。この事をここまで直さないのは、　申し訳ありません。

Posted: 7/25/2012 1:25 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

This TAS shattered the bounds of reality. Nothing makes sense anymore. Also, Peking Duck? Jamon Iberico? Yes vote.

Posted: 7/25/2012 1:14 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

[18:05] <Nahoc> what am I supposed to do with kyman's run? [18:05] <scrimpy> Nahoc: CRUSH [18:05] <scrimpy> KILL [18:05] <scrimpy> DESTROY [18:05] <scrimpy> SWAG [18:05] <scrimpy> CRUSH [18:05] <scrimpy> KILL [18:05] <scrimpy> DESTROY [18:05] <scrimpy> SWAG [18:07] |<-- scrimpy has left freenode (Remote host closed the connection)

Posted: 7/22/2012 10:18 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

p4wn3r wrote:

1) Prove that the smallest non-abelian group has order 6.

The symmetric group on 3 elements (S₃) has order 6 and is non-abelian, since (cycle notation) (12)(13)=(132) and (13)(12)=(123). For a group G with a smaller number of elements: |G|=1: That is the trivial group, which is abelian. |G|=2: That is Z₂, which is abelian. |G|=3: That is Z₃, which is abelian. |G|=4: There are two groups. If there is an element of order 4, then it is Z₄, which is abelian. Otherwise, the group is K=Z₂xZ₂, which is abelian. |G|=5: That is Z₅, which is abelian.

p4wn3r wrote:

2) How many homomorphisms f : K -> Z₄ are there? (K stands for the Klein four-group)

K=<(1,0),(0,1)>, so the action of f on (1,0) and (0,1) uniquely determine the homomorphism. In general, if f:G->H maps g to h, then the order of h in H divides the order of g in G. So (1,0) can be mapped to 0 or 2 in Z₄ and (0,1) can be mapped to 0 or 2 in Z₄. This gives 4 homomorphisms.

p4wn3r wrote:

3) Let D be a group of order 2n, where n is odd, with a subgroup H of order n satisfying xhx^-1=h^-1 for all h in H and all x in D\H (the set of elements in D but not in H). Prove that H is abelian and that every element of D\H has order 2.

Let h₁,h₂ in H and x in D\H. Then h₁h₂=xh₂^-1h₁^-1x^-1=xh₂^-1x^-1xh₁^-1x^-1=h₂h₁. So H is abelian. Let x in D\H and h in H, h≠e, where e is the identity. Then x²hx^-2=xh^-1x^-1=h. Since h≠h^-1 (otherwise h has order 2, but |H| is odd, contradiction), then x² is not in D\H, so x² is in H. Then x²=xx²x^-1=x^-2. So x⁴=e. If x²≠e, then x² in H has order 2, again a contradiction (similar to first bracket text). So x²=e. Since x≠e, then x has order 2.

Posted: 7/21/2012 8:48 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Pointless Boy wrote:

I'm a well-read native speaker of American English. It's a reasonable assumption that my lifetime of experience with the language has better prepared me to make judgments of English aesthetics than yours.

Pointless Boy wrote:

For example, just in this thread you have committed numerous blunders:

Pointless Boy wrote:

Since you are (I assume, based on your location information) a native Russian, a logical explanation exists for why your English is worse than mine. If you believe I am acting in good faith, it would be reasonable for you to defer to my opinion on this issue.

The opening post could be considered marginally funny, but here you are going way too far. There are things you should not say, and these are a few of them.

Posted: 7/21/2012 5:31 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Patashu is on the right track, but the conclusion isn't complete. Using notation above, B=(-d²+2bd-a²-b²)/2bd³ and dB/dd=(-d²-4bd+3a²-3b²)/2bd⁴. We check when dB/dd>0, dB/dd=0, and dB/dd<0. By completing the square on the numerator of dB/dd, we have dB/dd=(-(d+2b)²+3a²+b²)/2bd⁴. If we solve dB/dd>0, we get 3a²+b²>(d+2b)², or d is in the range (-2b-sqrt(3a²+b²),-2b+sqrt(3a²+b²)). Since d is only valid when a-b≤d≤a+b, we can restrict d to that range. We only consider the numerator of dB/dd because the denominator is always positive. If we plug in a-b into 3a²+b²-(d+2b)², we get 2a²-2ab=2a(a-b)>0, and so dB/dd>0 always when d=a-b. If we plug in a+b into 3a²+b²-(d+2b)², we get 2a²-6ab-8b²=(1/2)(4a²-12ab-16b²)=(1/2)((2a-3b)²-25b²). If a<4b, then dB/dd<0. If a≥4b, then dB/dd≥0. * In the first case (a<4b), since dB/dd>0 at a-b and dB/dd<0 at a+b, then exactly one local maximum -2b+sqrt(3a²+b²) lies in the range (a-b,a+b) and so the maximum occurs at distance d=-2b+sqrt(3a²+b²). * In the second case (a≥4b), there is no local maximum on (a-b,a+b) and the derivative is positive on this range, so the maximum occurs at distance d=a+b. This problem was motivated by the greatest brilliancy of Venus, which occurs about 5 weeks on either side of inferior conjunction, of which the position roughly indicated in the above image is one of the two positions relative to Earth. On the other hand, greatest elongation of Venus (θ=90°) occurs about 8 weeks on either side of inferior conjunction. So Venus is brightest at its crescent phase (θ>90°). On the other hand, Mercury is (usually) brightest at its gibbous phase (θ<90°). Using the assumptions of the problem above, greatest brilliancy coincides with greatest elongation when b=a/sqrt(5)≈0.447a. For comparison, distances from the Sun are 1AU for Earth, ≈0.72AU for Venus, and the range from ≈0.31AU to ≈0.47AU for Mercury. Sorry for the wall of text.

Posted: 7/20/2012 11:07 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

A star has two planets, planet A and planet B, which are distance a and b, respectively, from the star, and planet B is closer to the star than planet A (b<a). Assume the planets orbit in the same plane and the orbits are circular. An observer on planet A is looking toward planet B. Assume the brightness of planet B is directly proportional to the fraction of the visible surface that is in light (which is directly proportional to 1+cos(θ), where θ is the angle formed by star-planetB-planetA) and inversely proportional to the square of the distance between the two planets, where the distance is d. So the brightness functions as a rate of (1+cos(θ))/d². Find the distance d which maximizes this quantity. By the way, the image above, for what it implies in real life, is obviously far from drawn to scale.

Posted: 7/20/2012 7:26 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Mediafire download: http://www.mediafire.com/download.php?fdlyrsx1w1dvzaw (~26MB) Nicovideo (account): http://www.nicovideo.jp/watch/sm18397216 Nicovideo (no account): http://www.nicozon.net/watch/sm18397216

Posted: 7/19/2012 10:23 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Lately, I've been experimenting with a new kind of run. The goal is to beat the game while having two Pokemon only (starter, which has to be Mudkip, and Castform), with evolution not allowed, and some other restrictions that I am not set on yet. For testing purposes, my current test run only allows the use of level-up moves and HM moves (necessary only), and absolutely nothing else (no TMs, no held items, no items used in battle, no bike, etc.). It's difficult but I made it to the E4 with Mudkip at Lv42 and Castform at Lv31 or so. However, the E4 is extremely difficult and I have not yet completed it or have a full idea of what to do. Based on my research, I have collected the following information about RNG (most of this is obvious): - Whether an added effect of a damaging attack occurs (cause freeze, cause paralyze, cause confusion, etc.) is determined after the damage has been done and after any critical hit or super/not very effective messages. - Whether a Pokemon hits itself in confusion or not is determined right before the message "X is confused!". - The number of turns a Pokemon spends in confusion state is determined right before the message "X became confused!". - The number of turns a Pokemon spends in sleep state is determined right before the message "X fell asleep!" - Whether a Pokemon defrosts from freeze condition is determined right before the message "X is frozen solid!" or "X was defrosted!". - While accuracy for damaging moves (except OHKO moves) is determined before "X used Y!", accuracy for non-damaging moves and OHKO moves is determined after. - The amount of PP reduction Spite does is determined after "X used Spite!". - The number for Magnitude is determined after "X used Magnitude!" I also set up some "rules" that the AI tend to follow. Note that the AI is pretty much the hardest thing to understand and finding all its quirks is like trying to play Mastermind. - If the current NPC Pokemon has an attack that will always KO the current player if it hits, the attack will always be used on the next turn (except when said Pokemon is to be healed due to low health). - If the current NPC Pokemon has two or more damaging moves and decides to use a damaging move, the AI will select the most damaging move if it is always more damaging than any of the other moves. Only exception is if the immediate KO rule is in effect for two or more moves, in which case the situation is not clear-cut. - If any Pokemon in the NPC lineup satisfies all of the following: * is able to switch in, * has an attack that will always KO the current player if it hits and is super effective, and * resists the last attack used by the current player Pokemon, then the AI has a positive probability to switch to that Pokemon, provided that the immediate KO rule is not in effect. - If the current NPC Pokemon has a weather-changing move and the current weather is not a result of that move and is not "no weather", then the move will always be used, provided that the immediate KO rule is not in effect. Also, I found something weird. At one point, Mudkip and Castform both had 60 Speed. Glacia's Sealeo (the one with Surf, Body Slam, and Ice Ball) had 65 Speed. Yet Mudkip and Castform had first attack. I do not know why this is so. The numbers are given according to my script. For comparison, the other Sealeo had 67 Speed and had first attack.

Posted: 7/19/2012 8:48 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

About a week ago, I heard from sparky that he wants to do a TAS of Mega Man X2 on the Japanese version, using Snes9x v1.51. So I quickly hacked the existing MMX Lua script into a MMX2 Lua script. I did it back then but I should post it now: http://tasvideos-moviefiles.googlecode.com/files/megamanx2.zip There are still some bugs in it, but the essential stuff works. Also, since X3 has very similar memory structure to X2, the script might work on X3 as well.

Posted: 7/19/2012 4:53 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

p4wn3r wrote:

I looked a lot of threads back and saw no problem in abstract algebra, so here goes one. It can also be solved by looking it up, if you know groups well: Let G be a finite nontrivial group such that, for every two elements a and b in G different from the identity, there's an element c in G satisfying b = c^-1ac. Prove that |G|=2.

This is easy. Two elements a,b satisfying b = c^-1ac for some c in G are in the same conjugacy class by definition. Since this holds for any a,b in G other than the identity, then there is a conjugacy class of size |G|-1. From the Orbit-Stabilizer Theorem (using conjugacy as a group action), the size of any conjugacy class divides the order of the group. So |G|-1 divides |G|. This implies |G|=2.

Posted: 7/18/2012 4:09 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

This game is weird. I actually like it. The platform ride in Stage 4 is rather long, unfortunately. Also, I didn't catch the warp to Stage 3 first time around so I found myself asking where "Stage 3" was. What's more interesting, there was actually an older 14:17 version of this run on Nicovideo by the same author iiro2 that has over 150K views. It came in #2 spot on June 2012 Nicovideo TAS Rankings. (All links above are actually two links, first one for account, second one for no account.) No wonder the final boss fight seemed to be familiar to something I watched in the past.

Posted: 7/17/2012 7:07 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

RocWind, Unfortunately, I must remove the link to your site in your advertisement. Please note that starting one's posting life on any forum with an advertisement is not a good idea.

Posted: 7/16/2012 7:25 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

This appears to be a problem with DirectDraw. To correct the problem, use Direct3D instead (Config -> Video -> Render Method -> Direct3D).

Posted: 7/15/2012 8:11 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Hi, ctu. The movie requires an older version of VBA (VBA v22). It does not work on VBA v23 or v24. I am not aware of any other settings that may affect playback.

Posted: 7/14/2012 6:22 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

arflech wrote:

This one may also be a matter of "looking it up": Find an example of a real-valued function of a real variable that is nowhere continuous and yet still satisfies the conclusion of the intermediate-value theorem (that is, the image of an interval is another interval), thereby showing that the converse of the intermediate-value theorem is false in any reasonable sense.

Correct. It is a matter of looking it up. http://en.wikipedia.org/wiki/Conway_base_13_function If you want it defined on the whole real line, then define the function to be the Conway base 13 function of the fractional part if not an integer, or 0 if it is an integer.

Posted: 7/13/2012 9:18 PM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Is it Liouville's Theorem? Edit: The page also links to the Little Picard Theorem, which is a stronger form.

Posted: 7/10/2012 6:07 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

I wish the original SMW music was kept. The hack was innovative, and the Special World with the Green Hill Zone and others got me by surprise. That being said, I think that Super Demo World is a better hack.

Posted: 7/6/2012 1:21 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

Thanks to cstrakm for giving me the VBM file: http://dehacked.2y.net/microstorage.php/info/338836704/pokemonFR_j_cstrakm.vbm In his TAS, Squirtle's IVs are 16/24/21/23/15/20 and has Atk+ Def- nature (determined using my Pokemon script, of course).

thedarkone wrote:

I saw someone that found a GREAT route for speed running through leaf green. Without doing a TAS he got it done in 2:31:39. I was mostly planning to do a test run for the route they gave: http://pastebin.com/TLBhVJLa ...

While speedrunning routes may be helpful, remember that a TAS can do so much more. A speedrunner has randomness as an enemy, a TASer has randomness as a friend. Or enemy, depending on the circumstances. But anyway, with sufficient control, you can do things in a TAS that a speedrunner would dare not to consider.

thedarkone wrote:

I was wondering, how long before making an attack would I have to make a save state so I can keep on reloading and ensure full critical hits?

Use the script I posted a few posts back. As long as you are able to change the alignment of the random stream before it determines a critical hit, that would be a good place to make a save state. Critical hits are usually determined at the first "RNG overclocking" (excess of normal RNG cycles per frame) after the move message "X used Y!", provided that the move did not miss. Edit: In Fire Red/Leaf Green, the first attacking move in the first battle cannot be a critical hit. In other games, tutorial battles cannot have critical hits.

Posted: 7/6/2012 12:59 AM

Quote

FractalFusion

Editor, Experienced Forum User, Published Author, Skilled player (1941)

Joined: 6/15/2005
Posts: 3247

NitroGenesis wrote:

I tried the most recent version on the SVN and I do not have this bug.

Hm, last time I tried, both r421 and r422 have the bug. The bug appears to occur when you loadstate and then frame advance, or loadstate multiple times. It seems to be possible to switch the parity as well but I do not know for sure what causes it. (I tried stopping the movie directly after loadstate, toggling read/write, and closing VBA, but that doesn't switch the parity.)