Here's a TAS I did of head shots a while ago.
Entertaining? You decide..
http://youtube.com/watch?v=f5ruG1-YF3Y
In my experience.. Multiplayer desyncs a lot.
I have many fond memories of playing this game back in middle school.
Unfortunately, the video doesn't look great. It looks like you spend time aiming, which is a huge no no for a TAS. Why not run single player?
Good stuff. I almost thought Palom was going to get killed right before the boss dies for more XP then finish it off with Cecil.... if what Bob A asked is true, it would make sense.
Tellah doesn't get Recall in this version, and I doubt any of the characters will level high enough for Stone.
Whoops. Guess it goes without saying I've only played the GBA version.
I thought that it was Tellah that cast the Meteor during the Milon fight, rather than the twins. Rewatching it, I see that Tellah does nothing (?), except for something maybe about an ether that goes by really fast... and eat all the counters. In any event, its really impressive. Manipulated Flares for the win.
I have one question though... do levels that you gain before people join your party matter at all (in this version or the gba)? Or does everyone join at a static level. If its the latter... that's really annoying to me for a normal play through, although awesome for a speedrun.
Just watched the whole thing, very very enjoyable. I don't know why people are complaining about the restarts and the pause so much, they detract very little from the enjoyment imho.
I never actually finished this one (although I'm half tempted to dig the GBA cartridge out now), so I'll be watching this thread carefully.
I'm curious as to later game strategies. Twin magic seems very useful and powerful, as well as Teliah's Recall. I'm just hoping there isn't anything as game breaking as L5 Death or Roulette in this game... as much as I enjoyed those runs and as good as they are, I prefer to see bosses die without the use of instakills ^_^
You're using the wrong ROM. You have to find the 1.0 version, which is preceded by 0095.
As an update on progress, I'm finding the next mission against Eagle extraordinarily frustrating. His AI does different things depending on where I move different units (to a much greater extent than others), so when I fix one problem, I break everything else. As such, its very difficult to control where the AI goes, and I lose important units.
It's just a matter of finding the right combination, so I'll keep working.
Why did you destroy the rocket in mission 6? It's 6 spaces away from the enemy HQ, so it won't be able to stop your infantry from capturing.
Thanks for catching that (and before I had started recording the next mission, which made it painless to fix). Saves me 147 frames. I guess I've been fighting Grif too much ^_^
samurai goroh wrote:
Probably because of the battle animation, sure it can't attack the Infantry, but it can attack a tank...
It kills the APC, and one explosion takes as much time as the other, but ignoring it saves me the time of moving to the tanks and giving them orders, which is the 147 frames.
Seems you have a FoW map next, the address code to change the fog is located at 0300431C, keep in mind that the enemy ignores the fow, but you can't...
True. I've seen an allied computer player fire on a unit in fog of war that I couldn't see with shared vision.
I think Grit-Andy-Sami might be the best combination to use on the final battle if you can get the following strategy to work:
Sami has the unique ability to get a relatively large number of decent units (mechs) over the mountains in a relatively short time. Grit presents the computer with a large number of tempting targets. If you can use Grit plus two of anyone's t-copters to lure Sturm's forces away from the capital, you can wipe most of them out with his special on the next turn. At the same time, emerge from the mountains on the east with Sami's infantry/mechs and t-copters. See if you can send a few of Andy's units to suicide up the center to draw fire from Sami's units. Then, move Grit's remaining artillery into Sturm's space to draw fire, suicide a couple more of Andy's tanks, but one way or another place a mech or t-copter on each of Sturm's factories and start capturing. If he can only produce air units and you manage to somehow come up with an anti-air squad in the same space, you win.
To further this, you should probably give Sami control over the center airport.
That looks like a fairly solid plan. The only problem is, it seems like its MUCH faster to play through Andy's route than Max's meaning I get Max for the final battle. Still, its something to think about.
WIP updated, now has Zero through Five. Six is done but I want to redo it, so its not included.
samurai goroh wrote:
Nice, I'll watch this WIP later & comment if I think there's an error somewhere...
I'll definitely check out that website. I've been doing all of my strategies from scratch as I get to the map, so even though I made every effort to make it perfect, I'm sure at least some (if not all) can be optimized.
As for the glitch, I'd never heard of it before this. I think I'll refrain, but it's something to consider for a different run.
And no, I have no idea what the differences are. My ROM is 0095_-_Advance_Wars_(U)(Mode_7), if that tells you what version I'm playing on. I don't even know. People seem to be able to play my WIP without issues.
Oh, so the best way (S/A rank) isn't the fastest way? That does sound like it could make for a good watch.
I'm just going for whatever beats the level fastest, regardless of rank. Usually I'll end up getting a good rank anyway, but sometimes its just faster to let units die rather than maneuver in a way to save them, so my technique is usually sub-S levels. The best way to get max technique is once you have the units you need to clear a map, keep building infantry back home to pad your technique score, something I don't bother with since I don't care about rank.
I'll also need to do at least one of the Kanbei missions slower than the requirement, to avoid the 3 Sonja levels. I can't remember if that prevents S rank.
If someone wanted to make an all S rank playthrough (including every bonus mission), that would be awesome.
Edit: Andy's version of Mission 4 is almost 2000 frames faster.
Oh well. Still, will be fun to watch the game get beaten, maybe. Won't it feel a bit too walk-through-ish, since it's like a puzzle of where to move what each turn?
Since I'm going for the lowest time, you'll usually see endings very different than a normal playthrough. Generally I'll capture the enemy's HQ under heavy fire, or be down to my final unit, or something else exciting.
Nitrodon wrote:
While watching with input display on, it seems that you're using turbo a lot. This is generally considered a sign of laziness, but it especially hurt in those instances where you used turbo with the wrong button and didn't cancel the enemy's targeting cursor.
Also, I thought you said you were using Max in the final battle, but you seem to have used Max in mission 4 here.
I see what you mean now. It happens once each during Missions One through Three. I wasn't thinking about canceling the cursor I guess, since I was canceling the dialog. It's a dumb mistake, but little enough (90 frames total) that I don't want to redo everything to fix it. More likely, someone will come up with a faster strategy for one of the maps (I'd like to think they're optimal, but hey), and then the fix can be added in.
As far as Turbo being laziness... it helps me keep my sanity, and I don't really care if it /looks/ bad on the input. Generally the window between when input begins being accepted, and when it must be in for the command to be processed, is at least 1-2 frames, thus making using turbo fine. There are one instance, however, where I've been using it and it causes me to lose frames. There's a 50% chance to lose a single frame if I use turbo on the screen at the end of a unit's turn, because for some reason the window is shorter here (starting accepting input only the frame when the window appears). I'll try and be better about this, but again, its not going to kill the video.
And for Max and the final battle... yeah. Stupid mistake. I'm redoing 4 now. I don't know why I got it confused. Maybe it was my hatred of Andy overwhelming my rememberence. Although honestly, now I'm not sure. Andy's levels are going to be longer than Max's, but will Max being around at the final level make up for it? If I have to play through the last level with Grit, Andy and Sami, I'm going to cry. It's going to be agony even with save states.
Damage is variable, but only by a very very little. And it usually makes no difference (unless you're playing as Sonja or Nell, obviously). And given the same board, the AI will make the same decisions each time you play through a map, regardless of frames or button mashing. That's about as much as I can tell you about the RNG, since I know nothing about memory and whatnot. My guess would be that if someone found a way to better optimize a single level, it would be possible to replace that section of input, move everything else up, and not have any desynchs (hex editing, I believe its called?)
Nitrodon wrote:
Some parts of the WIP looked unoptimized, but that could just be indicative of the lack of manipulation available in this game.
What parts specifically? I'll go back and check any spots that look bad, but its optimized as far as my brain can tell.
One spot that might look bad but isn't... in mission two, I move an APC forward for seemingly no reason right before the end. If I don't, the artillery fires on my tanks and sets me back a turn on the final kill, but if I move the APC into the artillery's range, it fires on /no one/. I don't know why, and if I could isolate whatever causes the glitch in the AI, it sure would be nice.
Edit: Added Missions 3-4, sleep now.
Any chance you will have the final mission with that blue army long range man and um.. Hawk, or Falcon, the air supremacy green army guy? I've been stuck with them for years... Or maybe it was that high cost yellow army person... Bah, it's been quite a long time since I last tried.
So close, but the answer is c) Eagle ^_^
Anyway, my planned final setup for the last mission is Max, Andy and Sami on the right. Drake might be interesting to use on the right (I've never actually tried it), but using Andy for 4 of the hardest missions in the game sounds like agony.
As much as I love Kanbei, in order to get him, you need to play through three extra missions against Sonja, which will surely not be worth it ^_^. And Eagle gives you an extra mission at the end, so he's out too.
WIP (Missions 0 - 6):
http://dehacked.2y.net/microstorage.php/info/1842023824/0095_-_Advance_Wars_%28U%29%28Mode_7%29.vbm
Video of WIP (Missions 0-2):
http://video.google.com/videoplay?docid=-3013477591911033044&hl=en
My goal is to complete the normal campaign from a fresh start. That requires going through the final training mission, which I'll call Mission Zero for simplicity. There's no luck manipulation (or at least very very little), so the strategies you see are usable in their normal handheld, if you can make out what's going on ^_^
Mission Zero: Starts on frame ~1300. Unfortunately, I can't turn off the battle animations yet. That makes this about twice as long as it needs to be. Don't worry, the pace of the video doubles after this. Not much to say about the strategy... I have to minimize the number of fights because of the long animations, but I must destroy both pieces of artillery to capture the HQ. Destroying all units takes longer because of the silly animations, otherwise it would be worth thinking about.
Mission One: Starts on frame ~13500. A quick dash for the HQ. I actually do not have enough firepower left at the end to destroy that last tank while its sitting on a city, so I just surround it and win. My MdTank kills the rocket, then baits the enemy's MdTank away from their HQ. The reason why the tank resting right next to my capturing infantry doesn't kill me is because the AI refuses to use a unit which has been severely injured until it goes back up to full hp (most of the time).
Mission Two: Starts on frame ~23000. In order to capture the enemy's HQ, you have to take out both artilery and both rockets, plus the tanks nearby, so its faster just to kill everyone. Grif's CO Special is absolutely devastating, except that I force him to use it when nothing is in range the first time, then clean up as quickly as possible.
Mission Three: Starts on frame ~33500. This mission is all about guarding my transport copter from any damage (it needs to be able to survive a hit from a light tank to guard the capturing infantry at the end). I found the best way was just to surround it. I also had to make sure to do little enough damage to avoid triggering his CO Power until he flew his Fighters into my AA guns. When his CO Power does go off, it is actually quite helpful, killing a useless tank in exchange for moving the bomber into a perfect spot for my to kill it.
Mission Four: Starts on frame ~40000. Nothing interesting here. I delay my landing a turn to prevent the MdTank from coming back and kicking my ass. I use my artillery once to prevent them from killing any of my units only to save myself the explosion animations.
Mission Five: Starts on frame ~46500. I tried a bunch of strategies, but the most boring one worked best, unfortunately. I suicide forward with 2 LtTanks and 1 MdTank, and prevent my HQ from being captured. Any further input to my units only slows down completion, even with how much I die.
Mission Six: Starts on frame ~53000. I can't win this by elimination, (it makes me play a bonus level if I do), so I must capture his HQ. I issue at least once command to every unit except one (the lander), a record so far. The battleships butcher everything they can reach, so I get all the infantry into tree cover, except one that I load into the lander to save a death animation. I quickly eliminate everything Olaf has on land while sprinting towards the HQ with my APC in the minimum number of turns, all while dancing my units out of battleship gun range.
Mission Seven: Starts on frame ~61500. Working on it now.
Comments, improvement suggestions and encouragement welcome.
Sounds like a good run to me, if Golden eye got accepted and half the time the game is spent staring at the floor and the walls (one of the most boring runs I've watched on this site) running through the stages why couldn't this? I like no damage 100% runs.
A lot of the time might be looking at the floor, but the other half is spent shooting enemies in the head with breathtaking precision. I'm guessing you've never played Goldeneye before? Because that was a game that defined the N64 and a number of people's childhoods, and watching it get so thoroughly beaten was extremely enjoyable. In fact, I'd say its probably one of the best runs on the site.
And "takes no damage", assuming its slower than a regular run, is quite silly in a TAS, unless it differs significantly from the main run.
Your new video desynchs for me right at the beginning. Don't worry, its me, not you. I remember this was happening before, and I fixed it, but I can't for the life of me remember what I did.
I'll have more comments once I've actually watched it.
The reason why I believed that whipping while jumping pushes you back is because sometimes I'll be able to make a jump over a monster if I don't attack, but then if I whip, I'll get hit on the way down. It seemed reasonable, but it's not true.
I tried testing it in a vacuum. I started the game autofiring Start and A, then saved during the darkness. I ran flat along the ground until the beginning line of my head sprite touched the first candle I saw. It was on frame 1713.
Then I tried it short jumping four times. My head hit the candle on 1721, a 2 frame loss per jump.
Then I tried with 3 high jumps. I hit at 1719, or a 2 frame loss per jump. Note that it didn't help any if I timed my jumps to cancel the landing animation (this came as a surprise to me, I assumed the time loss was from landing).
Then I tried it high jumping while whipping, and it also came out as 1719, meaning my original assessment was wrong. However, I can still repeat my observations about getting hit at the end of a jump while whipping while making it while not whipping, so while the net frame loss from whipping while jumping does not change, it might make the loss occur earlier.
In any event, the less enemies that you have on the screen, the less the gameboy lags, so I think your method is right in the end. Avoid jumping unnecessarily, but since you're jumping anyway, you should kill the maggots.
Here's an example of the awesome collision detection I was talking about earlier.
I'm trying to jump up onto the rightmost platform while killing the bat. I can easily kill the bat, land, then jump up, but combining them is obviously more optimal. However... as you can see, even though the candle is 10 pixels behind my head, I still manage to break it, and collect the heart a few frames later.
I don't mind so much struggling with these issues, I'm just worried people are going to think my routes are poorly optimized because of the detours I take. But as you can see, they are sadly needed.
Indeed :D
Not being allowed to pick anything up certainly makes you think about the game differently. For example, in the first level, the mostly vertical room with the two spearmen at the top... I had to try a bunch of different routes to finish that room with a decent time. You get attacked by various things as you land on platforms, and each platform is covered by a candle, so actually managing to kill the enemies without making the platforms covered in hearts is challenging.
This is speculation, but I think the way the game calculates sprite intersection is based on an invisible square around the sprite's total potential positions. You'll notice that you can hit a ghost from a decent number of pixels away from the actual body sometimes, its because the way the ghost wiggles back and forth means the actual box it is contained in has room for both positions, and just hitting where he could be works. I've done some game coding in flash, so I'm familiar with both types of collision detection, and their style requires significantly less processing power, so it makes sense.
Unfortunately, that works against me quite a bit in regards to avoiding hearts from candles I might want to break for speed's sake. Since the hearts do a similar wiggle, when you include the spacious box around my sprite, its extremely difficult to avoid picking up a heart once its out, bordering on impossible in many situations. So far I've just been forced to avoid breaking any candles until I'm already past them. This allows for sneaky things like jumping over a ghost who is using a candle for cover, then turning around and killing him before he catches up to me.
As a side note... I was looking through your website, and Philosoma run looks really interesting, but the Filefront link gets redirected to the front page. I tried searching but no dice. Any chance I could get ahold of it somehow?
You manage to use the extra speed from burning mode to grab the rope that I have to jump to. That's the difference.
I've watched through your video and I'm really impressed. I took a lot away from it which will be useful for my run. Some comments of things I noticed if you ever feel like improving it.
Whipping while jumping pushes your sprite back slightly, roughly a frame's worth of movement. Therefore, its slightly more advantageous to just jump over enemies instead of killing them if possible, assuming you aren't also getting hearts or something.
At ~8500, you stop and crouch to kill a bat which you can just jump over.
At ~9600 you get caught slightly on the platform above you while jumping... if you wait until your head is more than halfway past it before jumping, you wont lose any forward speed killing the spearman.
And also... I'm not sure if this is useful to you, but... if you look back for a single frame, you can freeze the Giant Skulls in place. If you time it properly, you can get them far enough back so that they despawn with two glances. Glancing back doesn't cost you much in terms of frames, but its still something. The big advantage is in hearts. I'm not sure if its better than what you did, but its something to consider. Keep in mind that the short hop you do to cast spells costs you two frames worth of forward momentum. Not sure how much looking back costs you...
Finally, I just wanted to say that I think I'll be doing a run where I don't pick /anything/ up if I can help it. No hearts, no whip powerups, no food, etc. I had the entire first level done this way but I deleted a large section by rewinding while viewing. This has taught me the value of back-ups :P. But doing it again will probably only improve the quality.
1 + 2: I'm not sure why I didn't skip those... I'll make sure to
3 - This I did know. I tried to avoid having the whip out too late, although I see a decent number of times when I screw it up, now looking at the video. On flat terrain, you have to whip a minimum of 12 frames after a short jump, 22 frames after a long jump, although landing at a different height changes this obviously. Rule of thumb is 23 frames before landing, you have to have already whipped or you'll be stuck in landing animation for at least an extra frame. I knew all this, but yet still messed it up quite a few times. I'll be more careful.
4 - I'm pretty sure this is not true, but I'll double check.
5 - You won't believe this, but I've never actually used it on anything but a boss (I've beaten the game on a cartridge back ages ago... borrowed it from a friend). I always assumed for some reason that it increased her damage. I feel pretty silly finding out this isn't true. Darn assumptions...
6 - More criticism is always welcome. This isn't going to be perfect just with me working on it alone, that's for sure. I've already overlooked some pretty obvious stuff, so make sure to kick me into line ^_^
First post here.
I've been a huge fan and avid watcher of TSAs for a long time now, but this is my first project. I was daunted at first by what I assumed would be technical issues (I'm not a computer programmer by any stretch) but I found VisualBoyAdvanced to be very intuitive.
I do have one issue, which is that when I pick up the file again, I have to fast forward all the way to the end, then hit 'begin recording here', to pick up where I left off. Am I missing anything? Seems a little clunky.
I'm kind of amazed that no one has tried Castlevania Legends. I did a search on the forums, but all I found was a request for it. But I'm not complaining, because running this game is an awful lot of fun. I've finished the first level, and I think its fairly well optimized. There's a few frames lost here and there but I'm satisfied for my first attempt. Actually now that I watch it, the boss looks really sloppy... I'll go back and fix that.
Let me know what you think.
VBM: http://rapidshare.com/files/64889354/Castlevania_-_Legends__G___S_.vbm.html
Video: http://www.youtube.com/watch?v=wOV6vuWtuUM
Edit: Thanks for the heads up Squ4ll-
I'm trying to jump up onto the rightmost platform while killing the bat. I can easily kill the bat, land, then jump up, but combining them is obviously more optimal. However... as you can see, even though the candle is 10 pixels behind my head, I still manage to break it, and collect the heart a few frames later.
I don't mind so much struggling with these issues, I'm just worried people are going to think my routes are poorly optimized because of the detours I take. But as you can see, they are sadly needed.