Posts for FatRatKnight

1 2
17 18 19
43 44
Posted: 9/5/2014 7:51 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
With all this, I am also angry. I come by hoping there's some sort of progress to Metroid Prime. ... At this point, the relevancy of loading times does not matter to me. I don't care if it is absolutely pivotal to the creation of a Metroid Prime run, and that all attempts at a TAS will be futile if it isn't fixed. What I do care about is that the general fury I'm feeling from the topic. I feel that, even if emulation is "perfected" right the snap now, what I'm reading implies to me the schedule for TASing Metroid Prime is pushed back a few years, as it appears likely a GameCube TASer or two will leave the site due to this one event, or at the least have bad feelings whenever the thought of TASing this game comes to mind, discouraging attempts even if they choose to stay. Well, that is as I see it. ... Can someone please split these recent posts off somewhere else? I don't want to come across them again when I browse through old posts sometime in the future.
Posted: 9/4/2014 10:38 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Shop RNG script v1 I feel it's worth making a new post for. That should speed up your search considerably. I hope my choice of two-letter identifiers make enough sense, as screen space doesn't really allow for 80 "One-Winged Angel" to fit in there while retaining any sort of useful vision of the game itself. Basically, make use of the script and hopefully any further tweaks to it isn't too difficult to make.
Posted: 9/4/2014 8:57 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Ghostwheel wrote:
I have reason to believe that GameFAQ's table is patently false (as is the guide it came from)
It is inaccurate, but the parts where it is correct enough supplied a great deal of information that sped up my attempts at figuring out what the game is doing. The numbers look correct -- Except some of them are in the wrong places.
Ghostwheel wrote:
The last thing I'm curious about is what pack type the free pack gives you. My hunch is that it's the highest available type, but I can't say.
The freebie I got from the Atlantica shop I generated from using the TAS as a base definitely looked like a brown attack pack, yet the shop offered one black bound pack of each type.
Ghostwheel wrote:
If you could actually give me a complete list of what the pack probabilities are for each band color for attack and magic cards that would be huge for SS running and be very helpful in determining what kind of cards are possible to obtain for the TAS. To do that you would need to go up to the last floor where basically all the cards are available.
Peeking at the precise-yet-inaccurate data at the FAQ I linked earlier, and from my own testing, here's my current list: 
Note, I haven't had a chance to look at Moogle packs yet.
The numbers I used for Moogle, as well as unknown cards, are guesses based on that FAQ.
Spots left blank are theorized to be unavailable to that slot.

Leaf  Brown Black Moogle
00-19 00-13 00-03 00-03 Three Wishes
20-39 14-27 04-07 04-07 Pumpkinhead
40-59 28-41 08-11 08-11 Wishing Star
60-79 42-55 12-15 12-15 Lady Luck
80-99 56-69 16-19 16-19 Olympia
      70-74 20-29 20-25
      75-79 30-39 26-31 Crabclaw
      80-84 40-49 32-37
      85-89 50-59 38-43
      90-94 60-69 44-49
      95-99 70-79 50-55 Divine Rose
            80-84 56-65
            85-89 66-75
            90-94 76-85
            95-99 86-95
                  96-99

00-14 00-09 00-04 00-04 Fire
15-29 10-19 05-09 05-09 Blizzard
30-44 20-29 10-14 10-14 Thunder
45-54 30-34             Cure
      35-39 15-24 15-19
      40-44 35-44 20-24 Aero     (Don't know about Moogle, but)
      45-49 25-34 25-29          (Aero shifted between brown and black)
55-74 50-54 45-49 30-34 Simba
75-84 55-74 50-57 35-44 Genie
            58-65 45-54
85-94 75-94 66-73 55-64 Dumbo
            74-81 65-74
            82-89 75-84
95-99 95-99 90-99 85-99 Cloud

00-49 00-39             Potion
50-99 40-79 00-19 00-14 Ether
      80-89 20-39 15-34 Hi-Potion
      90-99 40-59 35-54
            60-79 55-74
            80-99 75-89
                  90-99
Ghostwheel wrote:
Alternatively, I could do this testing myself if I knew anything about this kind of thing. If you could walk me through the steps you took to find the probability table you produced earlier, I could get to the end quickly and do the pack testing I need.
I can't think of any good way to do this on the actual console, but I can suggest a way that might help through VBA. 1) Get VBA 2) Get Kingdom Hearts: Chain of Memories to work on VBA. 3) Get my work-in-progress script ... Okay, silly stuff out of the way. Once you run the script (Tools -> Lua Scripting), it should spit out a whole bunch of numbers. The left side is just spitting out predicted dice rolls, but the numbers on the top-right are my first attempt at figuring out what it all means. The two-digit numbers in that line are the important ones to pay attention to, as it tells us what card it is. Find a shop, jot down those numbers, then grab a pack, either attack or magic. The order I have the numbers should match the positions of the cards on screen. If you plan to pick an item pack, the top numbers become useless as they assume three dice rolls per card, not two. That leaves the raw dice rolls in that mass on the left. From the top-left number, going down, every second number starting with the first one listed (1st, 3rd, 5th, 7th, 9th) is the identifier used for each card in an item pack. The script is no good at predicting mix packs ahead of time, but those just use the same tables as the pure packs, with a range of 0 to 299 used. If you need a new set of random numbers, just leave the shop. Jump a few times if you'd like to adjust them slightly, or enter and leave the shop again for yet another new set of random numbers. Editing the RNG addresses directly is what I've been doing, really. This would require knowing where the RNG addresses are (16 bytes at 02034030) and a little about hexadecimal. If you think you know what you're doing, the low 14 bits of 0203403C and the uppermost bit of 02034038 (read as 32-bit values) are the important ones to tweak to edit the outcome of the first card.
Posted: 9/3/2014 6:56 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
I can stop your TAS at whatever point I feel like. Like when you hit a door. I'll just pick a different card. Shouldn't be hard to predict which card I'll use at this point. What I did was create a Moogle shop in Atlantica, interrupting your TAS somewhere there. The freebie pack you get there can give these cards: 00-13 Three Wishes (14%) 14-27 Pumpkinhead (14%) 28-41 Wishing Star (14%) 42-55 Lady Luck (14%) 56-69 Olympia (14%) 70-74 Kingdom Key (5%) 75-79 Crabclaw (5%) 80-94 Kingdom Key (15%) 95-99 Divine Rose (5%) The numbers relate to the RNG state. In particular, the 100-sided dice rolls I talked about. Seems the game uses modulo 100 for all three aspects of generating a card. In any case, all the script needs to do is to spit out a lot of future numbers. While you can manually check the cards, having some onscreen numbers will greatly speed up your search, once you can recognize what they mean. I can probably get the script to correctly predict the number and premium state, if those chances really are constants, assuming you pick attack or magic packs only. This leaves just the card identifier which may take further work to decode, but at that point, giving a number from 0 to 99 will help your search anyway. Once you know a particular range of numbers gets you the card you want, you'll know to skip the RNG states that won't give you that card. I already have a decent portion of what I need coded up to make this analysis. Converting it to a useful display at this point shouldn't take a lot of effort from me. I will tweak with some numbers as I try to get exacting ranges of what value to expect from the cards, then create a display that I think will be useful to speed up your search. EDIT: Well, uh... What are those fancy looking bindings...? I didn't even know the quality of packs changed later on. Shows the sorts of things I tend to know. It should be clear I never really got far in this game. ... Just more to analyze... I have this script. I'll share what I've got so far. EDIT2: Alright, with the help of this FAQ from GameFAQs, I now understand a lot more about the internals of how the game selects a card. After picking a number from 0 to 99, the game basically looks up a table and picks a card off of it. If you are not allowed to acquire the card, you instead get Kingdom Key (attack), Cure (magic), or Potion (item). Other than that, I don't see any strange tricks behind it. The mix pack takes modulo 300 from the RNG, rather than 100. Naturally, numbers 0 to 99 gets you attack cards, 100 to 199 gets you magic cards, and 200 to 299 gets item cards, and from there, it uses the same look-up table using the low two digits. ... I think I can generate a useful script that will identify exactly what you get. Well, other than cards I don't have access to, as there are limits to what cheating with the RNG can give you, which gives me Kingdom Keys, Cures, and Potions for cards you don't have access to, yet. Something at the endgame should help, but I can still probably pick out a few ranges and indicate you're in that range, though not what's in it.
Posted: 9/3/2014 4:00 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Tounet01 wrote:
- Getting these cards will roll the RNG just once (I think).
15 RNG calls. I suspect the game needs to identify what the card is, how high that number is, and whether it is a premium card -- Three random values per card. Item cards do not make the premium RNG roll, leaving only two RNG calls per item card. Apparently, the mix packs still only make one identifier call per card, so for those, 10 to 15 RNG calls are made, depending on what number of them are green cards. The shiny premium cards use the RNG to shine so beautifully, with the occasional sparkle calling the RNG to be positioned. Tounet01 already knows this, so it's not like my saying this would help, but mentioning this for completeness. I might crack into the shop relating to the RNG for a bit. Whether the RNG numbers from this will be helpful for the TAS is another matter. In any case, when entering a room from any sort of screen transition (from another room, leaving the menu, finishing a battle, exiting the shop), the RNG is reset based on some rather volatile sources (perhaps a timer; I did not look into this). That will be largely impossible to deal with in a real-time speedrun. EDIT: My analysis of the RNG for the shop continues. I believe it is best to document my findings while I'm here. After generating a new RNG value, the game uses the lower 15 bits of this new value, then takes modulo 100 of it. Which is to say, in a sort of formula: (NewRNG AND 0x00007FFF) MOD 100 When generating the five cards from the pack, the first RNG value identifies the card, then the second RNG call gives it a value, and finally an optional third call may pick either normal or premium, with this third RNG value not getting used if it's an item card. Just a cursory glance indicates the following chances of each value: 0=5% | 1=15% | 2=15% | 3=15% | 4=15% | 5=10% | 6=10% | 7=5% | 8=5% | 9=5% 10% chance of a premium card I'm using the Agrabah shop used in the TAS to get these numbers, and right now I'm making a fair number of guesses on these percent chances, so don't rely heavily on these numbers. I have no information on whether these chances change depending on where you are in the game, as I have only one real data point in location. I can probably tweak my script to spit out the next 15 RNG values, and revealing the 100-sided dice rolls, as I do have the exact formula to produce the next step of the RNG call. With the possibility of premium cards, and the choice of item cards that take fewer RNG calls, means it will be difficult to make a useful script that will predict what you'll get for your second pack.
Posted: 8/23/2014 4:25 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
SoulCal's post was indicating the Wii version. Your posts (kusoman) are indicating the GameCube version. If you can make that initial jump in the Wii version, then please, go ahead. Otherwise, try to keep in mind what version is being talked about at the time, as we already know (or I hope we know...) that it's possible on the GameCube version. In the future, if you're going to add something else within a few minutes, and not in response to another post, it is much preferred that you edit your latest post instead of making a new one.
Posted: 7/7/2014 5:59 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Aside from implementing my own suggestions for Soul Blader (J), and never bothering to upload that, I have no progress worth noting. I'm usually pretty vocal about any progress I make if there's any, and quiet if I've stopped. It helps to have an occasional reminder, though.
Posted: 6/3/2014 9:47 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
maoushoujo wrote:
Is there a particular reason why you didn't skip the Alexis encounter in the Ruins? IIRC you can just teleport around the trigger point like in the Tower, and that would remove a fairly lengthy cutscene.
The Alexis encounter at Ruins Area 4? I am not in reach of an open square past the trigger spot, as the map doesn't have any explored spots past that trigger, and the only tiles in reach from the doorway are walls and the scene trigger. Warping to the scene trigger square does not let me skip it.
maoushoujo wrote:
At the Alexis encounter in the Fortress, is it faster to save her or kill her? (You have the option of attacking her yourself; she'll drop the key on death and won't show up in the ending.) I don't remember how the two versions of the scene compare in length, or whether you still have to deal with the Rakshasa afterwards. If Alexis is dead then the ending will be significantly shorter, but player input has already ended by that point so that part is moot.
Not sure which is faster (I never tested), but I do know that, if I ever go through the new route getting experience for Double Magic, killing Alexis is not worth as much experience as the thing trying to kill her, and we're very tight on experience. As for the ending, it does matter for in-game time, as the timer is still running up until the final time is finally displayed. (Actually it keeps running after that, but whatever routine displays the time doesn't care after it does its thing) Standard TAS timing isn't about the in-game time, but how long that input file is.
Posted: 6/1/2014 5:53 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Area 4 -- Well, half-beaten. Restarted the run on lsnes and dealt with motivation issues. One trick I've been abusing readily is the fact charged shots can "double hit" at times. Which is to say, the boss takes damage from it twice. This requires that the charged shot is "close enough" to do damage two frames in a row, which only works at certain times. It should come as no surprise I abuse this fact for bosses. The Area 3 boss has all four of its orbs hit twice per charged shot. About the charged shot: Against most normal enemies, the game skips the damage check and destroys it right away. Against bosses, it naturally doesn't keep the instant destruction, and instead does a certain amount of damage, depending on what is being hit. Some hits do only two damage (such as against those colored orbs around boss 3), while others do five damage (against the center part of boss 3). The double hit comes from the fact that it's within a certain depth difference to deal a hit, and still within that depth range when it continues on behind the enemy at the next frame. I really should measure this out at some point, but it effectively dictates what frame this can be abused on against bosses, as their relative depth is completely independent of your speed. Each of those colored orbs have 10 HP each. Charged shot hit does 2 damage to those. With a pair of double hits from my charged shots, and the fact it can hit all four orbs at once, meant I only needed two normal shots per orb. My limiting factor is when I can get my third charged shot, so I feel that battle is fully optimal. Now, Area 4. I've been getting through this place, salting the "off road" time carefully to minimize time loss while getting at every enemy. Those "fluffy snowballs" were rather difficult, as they spawn relative to your position and not based on an absolute position, so don't mind me hugging the left side in an attempt to manipulate their positions. It still took some careful positioning to even get them all, and once they start spawning, it's time-based, not distance-based, so slowing down wouldn't help at all. The part I stopped at is the real pain. Look at those green things. They grow. Each one of those segments is an entire object. Each fully grown "tower" is four objects. The limit is twelve. They are also durable and the bottom segment can only be reached by charged shots, which often can't destroy that particular segment in one shot, as for some reason it does 2 damage (they have 3 HP) instead of instant-boom, yet the other segments do go instant-boom. Optimizing my score looks like I'll need the brakes. This will give me more time to charge some shots. Making sure I don't destroy the "towers" behind the ones I want gone will be a tricky puzzle to solve. Making sure I stay within the 12 object limit will certainly make things real hectic. There will be "fluffy snowballs" that I want to spawn for maximum score, which just pushes the limit ever harder. It looks like I can't catch every enemy and spawn, at a glance, so I might have to concede a few kills. We'll see if I attempt to continue.
Posted: 5/24/2014 8:06 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
That movie is mislabeled. It does the DSS glitch, but does not glitch straight to Dracula like in this run. Perhaps there was confusion somewhere along the line when labeling that movie. ... I have no other comments to give at this moment.
Posted: 5/5/2014 12:56 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Hey, you're still working on this! It didn't take much of my time to produce that script, but it shows that a diverse set of skills can really help a TAS. I may have saved some work off your TAS, but by now, you're probably a lot more familiar with the RNG, or at least have an idea of some patterns in there. The run does look a lot smoother now. You did manage to save a few minutes, too. It definitely shows in your battles. Here's hoping Ghostwheel notices this update, as when I played this game, I was confused with the whole card thing and didn't get far, so I'm probably not a good source of strategy.
Posted: 4/28/2014 5:24 AM
Post subject: Hey, IRC. Fixed a tiny error in my 8-frame brakes text.
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Let's work out a few patterns while boosting with Fire Stingray. When I say "Bonus distance traveled", the units are in 1/256 of a pixel. This is how much additional distance I cover above and beyond what the usual maximum speed would allow during a single cycle of the boost pattern. I provide both hexadecimal and decimal values. I prefer hexadecimal, since I can very neatly line up speed values in my script, and I have no clue how to do so in decimal. 
No brakes
65 frame cycle
Bonus distance traveled: 0x2080 - 8320
Average speed: 0x0980.00 - 2432.00

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8-9C4-
9C0-9BC-9B8-9B4-9B0-9AC-9A8-9A4-9A0-99C-998-994-990-98C-988-984-
980-97C-978-974-970-96C-968-964-960-95C-958-954-950-94C-948-944-
940-93C-938-934-930-92C-928-924-920-91C-918-914-910-90C-908-904-
900 - Repeat

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504-2500-
2496-2492-2488-2484-2480-2476-2472-2468-2464-2460-2456-2452-2448-2444-2440-2436-
2432-2428-2424-2420-2416-2412-2408-2404-2400-2396-2392-2388-2384-2380-2376-2372-
2368-2364-2360-2356-2352-2348-2344-2340-2336-2332-2328-2324-2320-2316-2312-2308-
2304 - Repeat
This one's purpose is only really any good when at the tail end of your boost, when there isn't time to accelerate again under boost power. Also, there's enough time to go through four of these, then your boost cuts off shortly before you get your fifth. 
8 frame brakes - Begin braking from 0x0970 - 2416
45 frame cycle
Bonus distance traveled: 0x1C20 - 7200
Average speed: 0x09A0.00 - 2464.00

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8-9C4- Coast
9C0-9BC-9B8-9B4-9B0-9AC-9A8-9A4-9A0-99C-998-994-990-98C-988-984- Coast
980-97C-978-974- Coast
970-962-954-946-938-92A-91C-90E- Brake
900- Accelerate (Repeat)

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504-2500- Coast
2496-2492-2488-2484-2480-2476-2472-2468-2464-2460-2456-2452-2448-2444-2440-2436- Coast
2432-2428-2424-2420- Coast
2416-2402-2388-2374-2360-2346-2332-2318- Brake
2304- Accelerate (Repeat)
A lot better than simply coasting your way down, as you get rid of most of the slow half of the boost cycle in favor of a faster cycle with more of the high speed intact. Brakes reduce speed at 0xE or 14 per frame when you're traveling somewhere between 0x900 (2304) and 0x97F (2431), and reduce it at 0x0F or 15 at higher speeds (for most practical sorts of speed to brake at, anyway). If I were to brake for two more frames, they'd be reducing speed at the 0xF or 15 rate for one frame, leaving my speed at 0x8FF or 2303 before I can accelerate to 0x9FF or 2559, which leaves me traveling slower for the next cycle. 
11 frame brakes - Brake at 0x099C - 2460
37 frame cycle
Bonus distance traveled: 0x17EF - 6127
Average speed: 0x09A5.98 - 2469.59

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8-9C4- Coast
9C0-9BC-9B8-9B4-9B0-9AC-9A8-9A4-9A0- Coast
99C-98D-97E-970-962-954-946-938-92A-91C-90E- Brake
900- Accelerate (Repeat)

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504-2500- Coast
2496-2492-2488-2484-2480-2476-2472-2468-2464- Coast
2460-2445-2430-2416-2402-2388-2374-2360-2346-2332-2318- Brake
2304- Accelerate (Repeat)
The best way to remove the slow half of your speed without handy rails. In order to make up for the fact you can only decelerate at 0xE for one more frame, we decelerate 0xF for two frames in order to line up our speed to accelerate optimally. 
12 frame brakes - Brake at 0x09B0 (2480), with one frame coast gap
34 frame cycle
Bonus distance traveled: - 0x15DA - 5594
Average speed: 0x09A4.87 - 2468.53

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8-9C4- Coast
9C0-9BC-9B8-9B4- Coast
9B0-9A1-992-983- Brake
974- Coast
970-962-954-946-938-92A-91C-90E- Brake
900- Accelerate (Repeat)

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504-2500- Coast
2496-2492-2488-2484- Coast
2480-2465-2450-2435- Brake
2420- Coast
2416-2402-2388-2374-2360-2346-2332-2318- Brake
2304- Accelerate (Repeat)
Minimally slower than the 11-frame brake's average, but intentionally inserts one coasting frame in the middle of braking in order to avoid nasty stuff with speed no longer lining up cleanly at 0x0900 (2304) exact. However, this does allow a way to tweak with the exact timing in case the course goes through certain twists and turns where the difference in timing is critical. 
R2RB4 - Rail 2 frames, Rail Brake 4 frames
27 frame cycle
Bonus distance traveled: 0x13D0 - 5072
Average speed: 0x09BB.DA - 2491.85

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8-9C4- Coast
9C0-9BC-9B8-9B4- Coast
9B0-99C- Coast on rail
988-96A-94C-92E- Brake on rail
910- Accelerate on rail (Repeat)

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504-2500- Coast
2496-2492-2488-2484- Coast
2480-2460- Coast on rail
2440-2410-2380-2350- Brake on rail
2320- Accelerate on rail (Repeat)
This is a very fast pattern. It works because the rail slows you down by 0x10 (16) before applying your acceleration or brakes. So braking at 0x988 slows me down by 0xE rather than 0xF, which helps to align my speed nicely. Then I can accelerate at 0x910 rather than going all the way to 0x900. I love these rails! 
R3RB4 - Rail 3 frames, Rail Brake 4 frames
23 frame cycle
Bonus distance traveled: 0x10E8 - 4328
Average speed: 0x09BC.2D - 2492.17

A00-9FC-9F8-9F4-9F0-9EC-9E8-9E4-9E0-9DC-9D8-9D4-9D0-9CC-9C8- Coast
9C4-9B0-99C- Coast on rail
988-96A-94C-92E- Brake on rail
910- Accelerate on rail (Repeat)

2560-2556-2552-2548-2544-2540-2536-2532-2528-2524-2520-2516-2512-2508-2504- Coast
2500-2480-2460- Coast on rail
2440-2410-2380-2350- Brake on rail
2320- Accelerate on rail (Repeat)
The fastest pattern, though the R2RB4 is only barely slower. There's room for 11 of these on a single boost, with only two spare frames after 11 cycles. You can bet this cycle is used very often whenever I'm boosting on straightaways. Specifically any of those that go straight vertical or horizontal, as the diagonal ones have a jagged edge for the rail, interfering with this pattern. During a boost, the accelerator is useless for maintaining speed, so you only need to press it for one frame in order to accelerate. If you're not turning, and the accelerator is off, your momentum stays constant, whatever direction you happen to face. This little detail gives a lot of control when it comes to dealing with the rail in order to move into it and back off of it at exactly the right times. Without that, this cycle would be very difficult to pull off consistently even in TAS conditions. Now, here's the fastest sets put together: 6x R2RB4 4x R3RB4 1x No Brakes Total bonus distance: 56064 Base speed: 2304 Frames needed for base speed to cover this distance: 24.33 Number of seconds by in-game clock: 0.365 seconds More time than that should be saved due to taking corners more sharply, where going over the dirt or rail would otherwise ruin your speed and impact your time worse than what you'd have saved.
Posted: 4/27/2014 9:29 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
I prefer to comment my own stuff, though I do appreciate the offer in case I really am burnt out. I've been looking into Double Magic for a bit (well, mostly Ilari, so thanks). Without it, the game uses Arm Strength + sword power to determine your damage. With it active, the game instead uses 255, completely ignoring whatever you've got equipped or your strength. The hydra boss? Dead in less than 5 hits. The final boss uses a different formula entirely, taking 6~7 damage instead of 4~5, but this will still help out. The much speedier hydra kill and somewhat faster final boss victory makes it worth considering going back to see how much EXP I can get. I need to be level 27.00 to use Double Magic. When I enter the hydra floor, I'm level 13.97, so that leaves 13.03 levels to go. If I assume I'm level 20 the whole time, the convenient monsters in the labyrinth (aside from lobsters, whose defense defines them as inconvenient) are worth 8.83 levels. Searching for less convenient sorts of monsters, I'm still left with a gap of a little over a level. Before my Arm Strength training, I don't have enough Knowledge to spam lots of fire, but there aren't enough monsters after that (skipping the lobsters due to their defenses) to get the levels I need. So now I'm left with looking for the fastest M.Potions in reach. There's a 20-dose potion in the store I get my Invisibility Potions, but at 10,000 I can't afford it after selling my Ramela. No other M.Potion in easy reach -- Before the Dark Zone, I already took all those just for the ninja, and the Fortress has a few in rather painful locations. There is, however, a very easily obtained Sword of Flames in a chest in Cave B1. Uncle Mario took it for a faster ninja fight. I won't slay a ninja with it, as I need low strength for the training to be effective, but it alone should buy me all the potions I will need. My Knowledge is less than 40 by the time I enter Fortress 1F, so the damage output and regeneration of my magic is terrible, but at least I can do a few dozen damage per shot, and the potions will take care of the regeneration, so the shots will at least come pretty fast. Now it's just a matter of how quickly I can make the needed kills. On a side note, I'd have Ramela equipped to the end. The defense is hardly meaningful as I lack the HP to survive anything, but it would have me hang on to it as putting the Leather Armor back on takes a few frames and the spare Sword of Flames should make selling the Ramela pointless.
Posted: 4/23/2014 2:29 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
User movie #14190878786303873 Cool, I've got a start somewhere. I'm getting a bit of stuff together for a decent "no pause glitch" run, so getting a fast lap should help with the practice. (And to hide the fact that I still don't know how to even get the pause glitch working. Wait! Stop reading my small text!) At the start of the lap, I trigger a boost one frame before actually entering the lap. It's probably common knowledge at this point that you can have three stocked boosts plus one ongoing boost, so this really shouldn't be a surprise. I also shift around a little, intentionally wasting two frames during lap 4 in order to optimize my position and the timer for lap 5. The lap doesn't trigger until my Y position reaches 0x0880, so one frame I'm at 0x087F and the next I'm at 0x0889. As for the timer, it seems to switch between ticking up "01 or "02 every other frame (and apparently take around 66 or 67 frames to tick a full second). For sake of in-game time, I mess around with that detail. Other TASes I'm aware of has done this before, so it isn't really anything new and exciting. Normally, you can't accelerate with your boost until Fire Stingray's speed goes down to 0x900 (2304). If you're on a rail, you can accelerate when the speed reads 0x910 (2320). This is probably due to the fact the rail slows you down by a flat 16 before handling your acceleration related stuff. The rail brakes are interesting. There's a few different patterns of braking I could do, but I like this one: Coast on normal roadway for 15 frames, drive on the rail for 3 frames, brake while on the rail for 4 frames, then accelerate while on the rail for 1 frame. A 23 frame pattern that can repeat 11 times before the boost runs out, letting me accelerate to the high speed one last time. Diagonal rails are not so friendly to this bit of trickery, as it's more like jagged edges rather than a smooth line. Which is to say, as I drive toward the rail, I will experience a few frames of normal roadway and electric rail mixed together. This messes up any plans I have of rail-braking. Normal braking is a little slower, but at least I won't have to worry about losing some distance to tweaking with my direction constantly. Jump plates do not agree with boosts. Limit 0x910 speed, sorry. The "dirt" doesn't actually slow me down any faster while I'm above 0x900 speed, but it does prevent me from accelerating while I'm driving over it, even boosted. However, it does affect my turning rate, so I opt for the rail during the final turn. The electric rail applies a mean friction, but leaves your turning alone.
Posted: 3/31/2014 11:06 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Hmm... It appears the wall turns fake right as you walk up to it. Just one problem: MP After some quick checking, Double appears to need 50 MP. At the level I get it (12), I only have 23 MP. You get 1 MP per level until level 16, where after that it's 2 per level (until level 60). Level 27 is where I'll have 50 MP. After a quick estimation, this means I must beat up 30 endgame monsters. This might have been a great spell, but it appears I'm far too low a level to make use of it. Good suggestion, though.
Posted: 3/31/2014 6:14 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
maoushoujo wrote:
Did you get the hidden Double spell late in the Fortress? IIRC it's at the end of Fortress:8. Does the strength boost from Double cap your str at 99, or can you go above that?
The Double spell is hidden in a 2x1 chamber in Fortress 7F, somewhere in that 3-floor maze. I don't know how to get in there short of cheating my way inside. I first spotted it in my item listing through my scripts. The final boss doesn't appear to care about my Arm Strength, as it increases up two points at no change in damage. I don't know if Double would change that, but it doesn't seem likely. The hydra is affected by my strength, and there are still a few normal enemies to fight by the time I leave the three-floor maze.
maoushoujo wrote:
Also, how does that southeast pit on Ruins:1 actually work? Sometimes it takes me to the small normal area, sometimes to the big secret area, and I don't know what controls that.
Hold B as you fall in. Seems you often jump in and sometimes don't keep the jump button held long enough.
Posted: 3/28/2014 5:50 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
http://tasvideos.org/userfiles/info/13628304150493123 I now have a complete run. The hydra boss can still be improved on, but I wanted to see whether there would be any problems after that fight. Turns out there aren't really any horrible problems to worry about. The fact those six headless ones just wandered around in those spots was just plain luck. I did not expect them to line themselves up so nicely in my route through that room. And yes, all six must die, and it's always the last one that drops the key. The run could be submitted, I suppose, but there's some potential improvements at the hydra. Whether I can be lucky again with the headless is another question. I still want to see how many frames can be shaved off, though.
Posted: 3/27/2014 2:58 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
http://tasvideos.org/userfiles/info/13592281639251372 Huh. So apparently, when I claim further progress will be soon, I really mean 9 months. Sorry about that. Regardless, I'm back in the mood to clear though this game. The glitch with getting past those teleporter things requires a message, such as the stat increasing messages, to show up while you're moving, and to then press a button to show the entire message at once on the frame you would teleport in order to somehow skip the teleport routine. Since my Arm Strength is still in the process of increasing, I was able to glitch two of these teleporters in good time. I have the Fortress Top to get through. I know I need the Planet Blaster to damage Berebus, and the only way to get that is to fight through something else to get it. Are the legendary armor and shield even needed? Can't I just TAS-dodge everything? I'll work out the answers later.
Posted: 3/26/2014 11:44 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
After a short talk with adelikat, BizHawk doesn't implement "False" correctly in joypad.set (it sticks the button as unpressed for an extended time even after the script has ended). This should be fixed soon. Until then, use nil and avoid manually pressing the buttons you don't want when the script is doing its thing.
Posted: 3/24/2014 11:03 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
http://tasvideos.org/userfiles/info/13538527443458224 - Lua script for lsnes (fixed something small) I suppose it could be ported easily enough, as I'm not using any lsnes-specific functions. I was making an Expert mode TAS, and did first lap of Mute City I slower than the WR opener. I forgot to save the movie, so I'd have to recreate the end of the TAS with my latest save state. Not that it matters much, since I didn't get such wonderful boosts from Expert GP mode racers early on (my technique for these opening bumps probably wasn't done right anyway). It's probably best to produce verification movie to get SRAM for Master, and TAS that. I don't know how much more aggressive Master is compared to Expert, but presumably it would save another few frames at least, and it's the hardest mode as well. I haven't actually done any hard analysis of the pause glitch, but my first attempts at it failed. I have seen no proof that it will short-circuit GP mode and currently have a weak impression you keep your place even though you skip laps. You can't win races if you're in 9th. 3rd or higher is required in the final lap for a completed race. I will ask for assistance (and will analyze that movie file) in order to confirm or deny just how much it actually shatters GP mode. I have made some other analysis. I'm going to define some terms. Using words I like, anyway. Facing The direction your machine is pointed at. Turning affects this directly. Momentum The direction your machine is actually going. It tends to shift toward your facing. There are 49152 possible directions, as this is the finest precision the game uses. However, the game only uses the upper byte to determine how fast you're going on each axis, which leaves 192 practical values. You can still tweak with the lower byte to make your turns slightly more responsive, as needed. When going straight along either axis, there's actually five possible values of the high-byte to momentum that gives you the maximum speed (128/128) on that axis, so you can move slightly on the other axis without slowing down. I have the speeds of the first 90 degrees tabled in my script, and I use reflection to get at the other 270. Turn & rotational momentum Turn by simply holding left or right on the D-pad. The game keeps a rotational momentum to your machine, which gradually speeds up or slows down over several frames. If your facing and momentum are far enough apart, having non-zero rotational momentum will slow you down. Blast Turn Does not exist, in the F-Zero: Maximum Velocity sense. Not that I've observed so far. Releasing A and pressing it again does not shift your momentum any faster than usual, at least when turning normally anyway. However, releasing A still has a positive effect on turns: The Fire Stingray turns much slower at maximum speed, but at any lower speed, even just one km/h lower, and it gets its full turning speed. Releasing A slows down your machine for exactly that effect. Maybe it exists, but I certainly haven't seen its effect, anyway. Snap Turn As I haven't seen this mentioned anywhere, I'm coining a term for it. For turning right: Frame x+0: Rotational momentum must be zero. Do not press right on the D-pad. Hold R. Frame x+1: Press right on the D-pad. Draw your own conclusions for left turns. Hint: It involves L and left. The Snap Turn instantly sets rotational momentum to maximum from zero, making your turns much more responsive. In other words, hold R before you turn right, and you will notice your machine reacting much quicker than if you tried turning right without R. Drift Holding L or R shifts your momentum to the left or right, but won't shift it past a certain point away from your facing. Does not affect the machine's momentum if there's any non-zero rotational momentum, but it does affect the maximum rotational momentum if you are currently turning. There has been apparent talk that it slows you down. Technically no; All it's doing is shifting your momentum in one direction. If you're driving "diagonal", you're going just as fast, no problem. Still, as it oscillates between a few different values of momentum, it's less of a perfectly straight line and more of a series of really small broken jagged lines. If your momentum is mostly along one axis, you can still keep full speed as these jagged lines all use 128/128 if they stay within one of those five values.
Posted: 2/28/2014 12:27 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
I think what we're looking for is a way to modify the input stream without promptly discarding it on a load state or barring the user from any control in read-only. And the inconvenience of a hex editor means you need to reload the movie just to see if there's any sync, when modifying on the spot what you have in the emulator itself would much more quickly let you know what's happening, while serving as a guide where you're headed as you have the future input right there. Refer to TASEditor of FCEUX, the desired TAStudio of BizHawk, my MtEdit script, and maybe something else I'm forgetting. I'm not aware of anything like that for Dolphin (on the other hand, I'm not familiar with Dolphin).
Posted: 2/27/2014 2:46 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
From my knowledge playing these games: - Attack, Level, and power of moves add to your damage rather than multiply. - Defense subtracts from this damage - STAB and type match-ups multiply Attack stages and Defense stages multiply the respective stats. There are different... "forms" of stages. Screech reduces Defense differently than Tail Whip from what I recall, and will actually work if the "form" that uses Tail Whip is already maxed out. ... How reliable this memory of mine is in question, but being Lv16 means you're practically guaranteed to do one damage against anything near the end, and any multipliers you use from type advantages for that one damage is still going to do no good. Your only real hope is to use ways to reduce or negate the defenses of late enemies so that your abysmal offense does more than one damage. I can confirm the formula is far different from the main series. In fact, if anything, most of the similarities between this and the main series are largely cosmetic. Effectively a Pokémon game in name only, as I see it.
Posted: 2/26/2014 7:59 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
http://tasvideos.org/userfiles/info/12964102343777807 My fast lap: 19"48 - WR: 19"28 Whatever crazy precision they've done impresses me. Especially since Training doesn't have the obstacles I can boost off of. Any experts of this game can probably point out where I'm slow, but at least I have something to start with. Now might be a good time to work out a few scripts. This will take a lot of digging into memory, coming up with an intuitive display, and punching out what might be useful to have.
Posted: 2/26/2014 4:51 AM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
Quota left: 40704kB ... It'll be a while before I crash into problems. Unless I suddenly go into a lua scripting frenzy and produce several MB of scripts all of a sudden. So, yeah. I'm probably being paranoid about that line of thought. Still, I'm asking in case, you know, the movie gets published and the verification movie that I had available suddenly vanished due to something off going on in my end. Chance of that type of insanity happening is low, so the problem is likely negligible.
Posted: 2/25/2014 11:20 PM
Quote
FatRatKnight
Editor, Experienced Forum User, Published Author, Skilled player (1174)
Joined: 9/27/2008
Posts: 1085
My main concern is... What if going through my list of user movies, I end up removing the verification? Sometime in the future, my quota is getting maxed out, and I start picking out things that I have little to no reason to keep. And accidentally... A submission guarantees a spot where the movie can be found, or at least, prevents me from erasing it by some mistake, but my user movies list is only a near-guarantee. ... Ah, whatever. I'm probably thinking to hard on it. User movies, it will be stored in. Verification movie progress will be slow, as I'm getting myself busy in writing stuff more.
1 2
17 18 19
43 44