Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
The ending of Sky Room is quite sloppy to get hit there.
Your shot at 32200 missed in the lake. I'm not really a fan of skipping free points in the lake to get an exact score. Also, you could have probably cleared 10,000 in the plane level.
You seem to take damage a bit too willingly. Taking damage in Dark Wood was fine, but the rest was questionable in that regard.
You do some impressive things and then there's just things that hit me in the face. Your routes are pretty solid, just stop with the really mediocre mistakes/choices (to be blunt.)
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
The run could restart from entering Pandam most likely, though the Red Pison fight probably has a fair amount of time to save with scripting if it were to be redone.
We either have to recruit Lejes after losing magic, or get Lejes his magic back in Guanta. Doros, M Pison, Monmo, and Gariso are the fights of concern. Personally, I think Doros would be easier with Lejes, but then Lejes is possibly down a level. It would also be possible to kill off Lux instead of Lejes at the end of fights.
-
As Nitrodon has emphasized, Monmo takes much longer every level we lose on Lejes. He estimates that 1 point of magic is about 1.25 damage against Monmo per Fire 2.
Previous average Fire 2 damage was 33 at (average 87 magic.) Monmo took 3 minutes like this.
Level 23 = 87 magic
Level 22 = 82 magic
Level 21 = 77 magic
Level 20 = 72 magic
Level 19 = 68 magic
(87-68)*1.25 = 23.75 damage lost. Monmo takes about 3 times longer or about 6 minutes longer than the original 3 minutes at level 19 Lejes.
(87-72)*1.25 = 18.75 damage lost. Monmo takes about 2.3 times longer or about 4 minutes longer at level 20 Lejes.
(87-77)*1.25 = 12.5 damage lost. Monmo takes about 1.6 times longer or about 2 minutes longer at level 21 Lejes.
(87-82)*1.25 = 6.25 damage lost. Monmo takes about 1.2 times longer or about 40 seconds longer at level 21 Lejes.
1 S Brain takes about 20-25 seconds to kill. 1 trip to purchase 9 vacuums is about 40 seconds.
According to my calculations, 18 S Brains (as opposed to 27) and no regular brain fight, puts us 62 experience short of level 20 after M Pison. (Level 20 Lux = (Assuming here) Level 21 Lejes) This saves about 4.4 minutes at S Brains over getting to level 22, and loses about 2 minutes at Monmo. In theory this also loses about 25 seconds for an extra fight, making it actually only save about 2 minutes. We could grab a couple recoveries at brains though, saving maybe 5 seconds.
Alternatively, 14 S Brains and no Brains puts us just over level 19 after M Pison. (Level 19 Lux = Level 20 Lejes.) This saves about 6.2 minutes at S Brains and loses about 4 minutes at Monmo.
It takes 11 S Brains to get us over level 18 after M Pison. It definitely isn't worth it.
23 S Brains puts us over level 21 after M Pison (for Level 22 Lejes). This saves about 1.8 minutes at S Brains and loses about 40 seconds at Monmo.
-
Level 22 Lejes is associated with 1.1 minutes of gain. Level 21 Lejes is associated with 2.0 minutes of gain. Does that extra level save about 55 seconds at the other bosses? Nah.
Level 21 Lejes is associated with 2.0 minutes of gain. Level 20 Lejes is associated with 2.2 minutes of gain. Does that extra level save about 12 seconds at the other bosses? I think so.
I believe that 18 S Brains and 2 Brains for a level 21 Lejes is optimal for a run utilizing the glitch. Level 19 Lux is reached after Serpent. This makes a recruited apprentice two levels weaker than the current run. (For Olvan this means worse gear.)
So scratch 2 levels, scratch many of the recovery items, grab an extra exigate for the penultimate dungeon, and we save maybe 3 minutes in the present. Less grinding is a plus though. The glitch easily saves 15-20 minutes in the past.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Here is a new glitch I'm sure everyone will enjoy. =)
http://www.youtube.com/watch?v=1oKoFCiHE-E
831 frames per iteration, 6 frame window of precision.
Nitrodon and I intend to complete the run under the previous route, while we could finish this run in literally a week with this new trick, the run looks quite unplanned if completed that way.
After we complete the originally intended run, we may in the future run with this glitch, probably with the intention of obsoleting the run we create. That run would cap out levels at maybe level 18 at Gariso. Many less recovery items would be required, and the time saved would be about half an hour. About half would be saved in the present and about half in the past.
We intend to finish our "non-glitched run", as without this trick it is nearly as optimal as possible. Having the run published glitchless on the site is a good thing in my opinion, as it makes the statement that we really can slaughter a game so restricted by an engine. Obsoleting it is fine by me, I just think it would be good for our current run to exist under the circumstances. We are mostly done, it would be wasteful to dispose of our progress. I hope this run will be generation one, and I hope a future run applying the glitch will be generation 2. (Planned from start to finish.)
I hope everyone agrees with this decision.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
It's quite watchable. TASing didn't add too much to a human played run really. You skipped one cutscene twice. Everything else seemed to be just sort of point A to point B. It's short enough.
Weak yes I suppose.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
It costs around 10 seconds or so to acquire and gives Lux 4 more guard. Guard is absolute garbage of a stat. The gear stats are not doubled by B Protect. Defend cuts damage in half on top of that.
1 inherent defense stat = 2 defense stats with B Protect. Defending then cuts damage to be dealt in half. Grabbing the Brown Coat would be about the same as forcing an extra 1-2 guard through leveling.
After all the calculations are done, the Brown Coat lessens damage by literally 1-2 hit points, which is an extremely insignificant amount after Red Pison. At Red Pison it may have opened up one or two other options in the battle, but the time saved is not nearly enough to make up the ~10 seconds required to get and equip it.
Guard is overrated. Power and Speed are the necessary stats to be concerned with.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
I have mixed opinions. Up until about 0:50 it was pretty solid. It got a bit worse musically in my opinion after you grabbed the Clavi.
I'm not sure what note range you have (8 notes it seems?), but would it be possible to play ocarina themes or old NES game themes?
The TAS itself is quite bland. The stages have little action and the boss you fought, though seemingly optimal, wasn't exciting.
Musically it has potential, but I'm not sure how much potential. I would only watch this run for the interesting musical aspect, so that is probably what you should pay the most attention to. (while being optimal at the same time) From my viewing thus far, I don't think I would consider it publishable, but if you think you can do more with the music, I may reconsider. See what you can do I suppose.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
LOL at waiting for the platforms before getting the flight spell.
The damaging boosting towards the end of the run was really cute. I wonder how many more opportunities exist earlier in the run.
I would vote yes on this right now, but I'm sure you know ways to better optimize it and maybe slightly better route it after this test run.
Great stuff, looking forward to your final run. ^_^
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Daltone wrote:
I'm not using auto-fire I am doing it all frame by frame, I don't know an effective way of finding the first frame i can input.
I usually pick a frame in the middle, see whether or not that works, cut the search space in half, pick half of the rest, pick half again, etc. It takes a bit, but once you get used to TASing you can mash through the buttons pretty fast.
Daltone wrote:
The password starts the game on level 1 of the second quest. (After you beat the game once it sends you back to the start similar to ghouls n' ghosts)
Ah okay, so is the second quest harder than the first presumably?
Daltone wrote:
I looked at evilchen's old WIP I improved on some if not most of it.
If you expect to get published, you should improve on all comparable sections. (If you lose time in one spot solely due to the fact that you gained time in another spot, this is okay.) You should be able to pick reference points and compare sections. For example just compare times exiting given rooms quantitatively.
Seems like you have the right idea.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Bumping old threads is always okay if you have something new top contribute to discussion.
Using Autofire is bad habit in a finished product. You have a 50% chance of losing one frame every time you do something with it. Unless you can prove that there is a two frame window that the button results in the same completion, you should always find the exact frame.
What does the passcode do? Passcodes are generally frowned upon unless they are well justified. Does it raise game difficulty? That's one of the few good justifications.
Health Planning seemed okay, though it seems like you might have died later than you could have. Death seems to be productive to gain health and save time. The damage boosts seemed pretty good from inspection.
I actually have never seen this game before, so I can't really say anything about how optimal it is. I don't even really know the goal of the game. You should check out the other runs in this thread and see how you compare on a frame by frame basis.
The game is alright, probably publishable if it is fairly short, and depending how variable further gameplay proves to be.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
while memory.readbyte(0x7e0001) == 0 do
snes9x.frameadvance()
end
This will hold the script in a loop before actually executing the bulk of the code when byte 0001 becomes nonzero. Similarly this could be tied into the joypad.get function in any emulator for reading game input or the input.get function in Gens for reading keyboard input.
As most of these questions remain unanswered by the currently incomplete LuaScripting page, it is obvious to me that I need to work on a bit of an overhaul soon.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
I assumed you were on an emulator with 'word' length values. You want readbyte. All values in the NES are bytes. SNES can have 'word' type values, and GBA/others can even have doublewords.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
While we are on the topic.
I created this page a while ago: http://tasvideos.org/LuaScripting.html
It should very briefly cover the basics of what wicked is looking for.
These other two pages need integrated with each other, and then referenced within the Lua Scripting. These may be helpful in finding your position value.
http://tasvideos.org/MemorySearch.htmlhttp://tasvideos.org/EmulatorResources/RamSearch.html
If anyone has ideas on what to do with these pages, I'd be interested in discussing it. Obviously I need to finish the LuaScripting page sometime. It became a bit time consuming to write simple code for games everyone has played and then explain it simply. I'll try to work on this some more soon. I have some notes since I last updated it. Obviously I need to include qfox's command list as a reference. Does anyone know any commands that should thoroughly be covered? I think everything I listed is the most necessary, but other things that I don't typically use are probably still useful to list.
Correction:
while true do
gui.text(0, 0, "Boss: " .. memory.readbyte(0x1C5E))
emu.frameadvance()
end
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
dave_dfwm wrote:
Overall, I think there is still a lot to learn about the game, such as fight queues, and how using heal, running from fights, or doing various actions in the fights (and then winning or running) affect where you are at in the fight queue. Often, it is easy to find the quickest way through a fight with Crags or Babbles, but it is often harder to quickly get to the fight in the first place.[/i]
How many RNGs does getting into a fight actually increment, is it that much faster than just walking around and waiting? I also still think it is probably possible to kill a metal babble or two at level 1 if you wait for the right RNG. If you killed 3 in one fight at a higher level, this would save about 30-60 seconds as well.
The RNG is at 001C and 001D, right? What is the incrementation method?
The overworld non-walking increment is at C3D2 I think? Could someone interpret the calculation for me?
EDIT: The following supercedes most of my previous discussion, but not all of it.
Oh wow, this RNG isn't really even random on the overworld. The only thing you can do is wait, and even then you can only skip RNGs by waiting 64 frames.
From what I gathered there are 3 increment types (you probably knew these)
Walking (16 frames per step) +1 RNG
Overworld (64 frames) +2 RNGs
Town (128 frame cycle, a constant number of times per town in 16 frame intervals (5 times in Rimuldar)) +1 RNG
Note: All party members must be off of the town or dungeon square before the 64 frame increment occurs.
I now see that in caves you will hit every RNG unless you manually skip them.
The battles are essentially the only way to radically change the RNG enough to produce a rare outcome.
Walking on plains has the minimal encounter rate. On the RNGs you will get engaged on plains, will you always get engaged on any tile?
Also on a separate note, Kathias had spare MP to use Return to exit Rimuldar after using the inn. I am fairly convinced that walking around outside before going from Bomb Crags to Rimuldar and Returning out would have been much faster than advancing the RNG inside Rimuldar. Maybe this could be hexed in?
Byte 6a65 goes from 00 to 01 after Ortega dies, but overwriting this doesn't seem to remove him from the game, so I don't think this is the one we are looking for.
I have a better understanding of what is going on now. Based on these observations, I still feel you could plot the whole run.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
I'm going to be mildly nit-picky. Consider it constructive criticism.
What's with the autofire during dialog? Unless it can only close every two frames, you could be losing 1 frame per dialog box. Are you possibly losing time here, or can you verify that you are not?
Submission Text wrote:
The Ortega fight is faster in this video. Using just one additional HEAL before the battle caused him to chant HEALALL just once in battle and die much quicker.
Submission Text wrote:
General bad luck - I had to do some screwy things to manipulate a Bomb Crag fight and the second Metal Babble fight. This is more time consuming than in version 1.
I feel like some RNG delving could make this run a lot more precise. You should be able to see things far ahead of time if it really is this rigid.
I feel that battles 3,4,7, and 8 (at a glance) aren't necessary in an optimal run. I also feel that since you feel the possibility may exist to kill a metal babble at level 1, this should be further investigated.
This RNG is supposedly very rigid, what is stopping you from calculating everything beginning to end on a future run before even starting? If every action increments it a set amount, you should be able to plan everything before even starting, and then simply execute it. Correct me if I'm wrong.
I can't accurately judge how much effort was put into the RNG manipulation, as it wouldn't necessarily show through the rerecord count. The fact that you're saying 'bad luck' as opposed to 'unfeasible scenario' leads me to believe you don't know exactly the path of the RNG necessary to make this feasible. I think it could fairly accurately be investigated if a metal babble fight is possible at level 1 based on the approximate RNG at that point in the run. This would save over 30 seconds, and should be investigated on a future run. It seems like you're saying it is beyond your ability to check this currently, which is acceptable for now, but you should learn how in the future. Games like this are much better calculated than using situational trial and error methods.
The non-RNG related portions of the run seemed optimal. I feel that more should be done in planning the RNG path for a future version.
Overall thoughts: Good movie, thanks for making it. The glitching and overall route were good. I question how optimal your RNG manipulation execution was. Feel free to deny this accusation by citing facts, but from viewing, this is how I observed it. Really I'm directly questioning, "Were you just doing the best manipulation through mediocre trial and error methods, or were you actually observing and calculating the RNG effects and how you could best advance them to the next point for a more optimal outcome in battle?" It seemed like the former. Please correct me if it was the latter.
Yes vote, glad to see this movie on the workbench, but I'd like to see you be a bit more wary of the RNG in future executions if my speculations are correct.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Yeah, it desynced for me on snes9x-1.43-improvement13, and it apparently then also desyncs on v1.43 v17
Nitrodon says it was working on v1.43+ v16 (interim), and I've been using snes9x-lua-v0.06.
On 1.43-improvement13 it desynced right after grabbing the moonlight for me.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Here's the video up to just before Gariso due to time constraints. (The WIP kills Gariso.) Nitrodon is responsible for the well-executed Monmo fight.
http://www.youtube.com/watch?v=Y0Rcx-BdEJQ
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
Based mostly on the debate that has taken place in this thread, I remove my intentions of creating a formalized monetary bounty list. I'll try to outline what caused this decision.
Baxter wrote:
It is hard to estimate a time for an optimized TAS.
I agree that clear time is not a good measurement.
Mukki wrote:
Must be accepted [...] and worthy of at least an 8.5 technical rating from me.
Heisanevilgenius wrote:
I could drive myself crazy doing a Metal Gear TAS and submit it and then you could say "Ehhhh, Technical rating: 8.4" and now I've done it for nothing.
I also agree that perceived, subjective technical quality is not a good measurement. While I'm sure Mukki has a pretty solid system for awarding a rating to a movie, the user will not know exactly what this means. The only thing that would be a good gauge is if it gets published or not.
Baxter wrote:
If people are making a TAS to claim a bounty, it is even dangerous to fully optimize it, as others might claim the bounty before you with an unoptimized verison. I don't think bounties will lead to better quality TASes.
I agree that if someone placed $100 on a run that there is a fair chance it would not be done to the best of the user's ability.
Kuwaga wrote:
Who would pay for a run, like OoT, that's going to be done [...] anyway? (when they could rather put their bounty on another one)
This is a good point.
Kuwaga wrote:
Doesn't the concept of bounties discourage team TASing as a whole? If there was a bounty for the first one to go sub-something on a certain game, wouldn't it harm the community? People would then keep new tricks secret and totally hide their progress, I guess.
We don't want the whole site to turn into the Super Metroid thread. =p This would probably be a rare case though.
-
As stated, encouragement is probably the best (and most rewarding) way to motivate a good TASer. I think bounties (for some) would ultimately detract from the main intention of TASing as a hobby for others to enjoy, (though only mildly).
It just seems that this idea as a whole will only create unnecessary tension amongst site members at worst, and will only produce a few new, wanted runs at best.
adelikat wrote:
Having said all this, if people want to work out monetary bounties between each other, then that is a freedom people have, so by all means.
I think I will leave this discussion at adelikat's statement.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
This game and system are just bad. The music and graphics are both quite painful. Gameplay is very bland. The physics engine sucks. Honestly, the recent Somari TAS was far more entertaining than this (ignoring flaws).
I know adelikat wants to TAS all systems, but I think this is just an unreasonable goal. Next we're going to TAS a Tamagotchi because it's a newly emulated system. You should redefine your goal to be 'TAS all currently published systems on the site'
Honestly I would rather see Atari 2600 runs than this, at least they could provide nostalgia and beat records. I feel kinda bad for anyone who had this system as a child.
Conclusion: I don't think that 'publish a lynx game' is a good argument.
For the record, I have discussed this briefly with others in previous months, and got around to posting on it possibly from reading either that post or some other post. It's not like you were the first one with the idea.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
My 2 cents:
Chrono Trigger is too broken to bother with having any sort of story preservation. I was able to read most of the Lufia 2 TAS while watching it.
The best idea is to create lua scripts to pause just before closing the text box, or possibly set timer delays. Both of these would be pretty easy to implement, but would require a somewhat active viewer, also the music would stop a lot. This script is not hard to write at all if the relevant text box bytes are found.
At a glance putting delays in seems like a good idea, but I don't find it to be a good idea in implementation. Hacking a modified input file to drag at text boxes is a plausible idea, but I do not support having users TAS intentionally waiting at text boxes.
Experienced Forum User, Published Author, Player
(208)
Joined: 7/7/2006
Posts: 798
Location: US
I'm still looking for more feedback on this thread. Post if you have a positive or negative opinion of it.
Regarding due date, I may as well just put the date posted.
djchilxxn wrote:
2: Run type: 100% (all worlds and at least one of every item collected [and preferably used also])
Are you sure this is the best definition for 100%? I'm not going to be a fan of runs that are almost publishable but not quite due to user stipulations.
