Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Hey Randil, great job on all the improvements so far.
Whenever I've fast-forwarded the run to reach certain points for testing, other stuff has been catching my eye, and so here is a list of more possible level improvements that I've noticed these past couple days.
Level 5-2. As shown in the above screenshots, a different strategy to save 15 seconds on this bridge/egg-float level: http://dehacked.2y.net/microstorage.php/info/276/Lolo2-5-2_southbridge.fcm. You can likely save at least another second compared to this demo run by making use of that extra shot and better positioning the floating eggs (see similar tip for level 6-2 below).
Level 6-2. The sooner Snakey's egg sinks the sooner it reappears, and waiting for the egg to appear is the biggest holdup in this level. Thus you should push the egg as far to the left as possible before floating it on the water.
I've tested and found that it can be put another full space to the left and still leave time to grab the heart, get back onboard, and shoot Gol.
Level 7-2. When playing this level on the console Leeper tends to be just an annoyance, but with some experimentation I think you could find a way to use him in order to save time on pushing emerald blocks.
Level 7-5. Two steps are wasted as you grab the last heart then walk over the egg towards the jewel box. Either grab the last heart and get back onboard the egg slightly before the egg reaches the end of the river, or switch the order in which you grab the last two hearts.
Level 8-3. In the current run, after shooting Gol you push him down three spaces then make a bridge at that row. It'll be less wasted walking for Lolo if you push the egg down only one space then put the bridge at the top of the lava.
Level 8-4. The current route involves twice waiting for Snakey to reappear and has 2 shots remaining in the end. You can change the route to shoot Snakey away earlier and make use of the second Snakey to get rid of all waiting time.
Castle-1. An alternative strategy as depicted below ought to save ~20 seconds.
Castle-5. Improvements on the current strategy: After shooting Leeper the first time, don't push him all the way down to Medusa. If you leave a full space gap between Medusa and Leeper, Leeper will have less distance to hop right before you shoot him the second time. Also, when you shoot the second time shoot as early as can be allowed, which means the egg should end up a half-space from Medusa's line of sight. Next, when pushing the egg down, you can leave a half-space between the egg and the wall before the egg hatches and Lolo puts Leeper to sleep. However, these frame optimizations are irrelevant because the optimal strategy for this level most likely involves a live Leeper. It's possible to beat this stage by relying on Leeper to hop down to the bottom when Lolo grabs the jewel box, and even if this doesn't time perfectly it could be significantly faster than having to wait for egg-hatching twice.
Randil wrote:
EDIT 3: I've redone Castle-2 with hanzou's strategy.... And I must be getting it wrong or something. I can only get it 15 frames faster. Take a look and see if I'm doing something wrong: http://files.upl.silentwhisper.net/upload4/C-2new.zip
Well, you do waste two steps to grab the bottom middle heart even though Lolo walks into that spot later on anyway, but that error was in your previous run as well. The main thing wrong in the new run is near the start of the level. The first thing you've done after waiting for the first Snakey to appear is to have pushed the emerald block down (like in the first run), but in the new run you should instead immediately grab the top left heart and then walk over to push the emerald block.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Randil wrote:
And about 3-4, could you post an fcm? Even if it's not perfect, I could get the idea.
Looks like I've lost the movie I made this past weekend, so I made another one just now: link.
By the way, Castle-3 looks like the most improvable level of all, likely on the order of 25+ seconds. The strategy used in Randil's current run requires walking the long way around twice. It's actually possibly to complete this level without ever walking around the top-right corner. I'll leave it up to someone else to try speedrunning this level (password QKDH), and these screenshots ought to convey the idea:
edit: Randil wrote:
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Randil wrote:
I looked through all levels closely, but didn't find even one single timesaver. I actually think that the levels on floor 1 are more or less frameperfected (except for 1-4 maybe). Try to prove me wrong! :) I'll get on to floor 2 soon.
Here is a run of 1-5 that saves 25 frames, and I wouldn't be surprised if there is an even faster way. For any level that involves Alma, Leeper or Rocky it's very hard to prove a particular run to be optimal beyond a reasonable doubt. The exciting aspect of this is that we'll always be able to find improvements on such levels if we look hard enough.
Another level that can be improved is 3-4. I was able to save at least 7 seconds by taking advantage of Rocky instead of relying entirely on emerald blocks.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Man, I thought I was done fooling with the pro levels, but here's one more demo. Hopefully this will be my last pro level vid for the time being.
After the last PROA improvement I thought about how it would be best for Gol to reappear on any of the lower squares. After a bit of testing, I now conclude we have been making this level way more complicated than it needs to be. Check this out: http://dehacked.2y.net/microstorage.php/info/267/Lolo2-PROA_twoshotsleft.fcm. I waste a couple squares by walking around Gol in the beginning and not shooting him when I could have, then I uselessly doodle at the end before the jewel box, and yet this still comes out 11 seconds faster than in Baxter's submission.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Michael Fried wrote:
I haven't watched it yet so I don't know how good it is, though you might want to take a look at some of these vids to see if there are any tricks you didn't know about:
http://www2.loops.jp/~neka/index.html
OmnipotentEntity wrote:
Michael Fried wrote:
Edit: Btw I see people saying that Rainbow Road was good, I'm too lazy to watch that far, was it better than in the vid at the link I posted?
Actually it's about 15 seconds slower... Now I want to retract my vote.
Michael Fried wrote:
If you want to see good Super Circuit vids, watch the vids at the link I posted. They're much better and as far as I know they were done without save states, so if you're gonna TAS that game then at least make it better than those vids.
Michael Fried wrote:
What happened to the "must beat all existing records" rule? All the tricks that Genisto did are in the non-save state vids, and the non-save state vids have more tricks in addition to that, and were also more skillfully played.
Michael Fried wrote:
Anon wrote:
And where are these vids that you speak of?
I linked to it earlier in the topic.
Michael Fried wrote:
I guess I have good general Mario Kart knowledge. Anyway, it doesn't matter because you can tell it wasn't played well just by comparing it to those Japanese vids.
Michael Fried wrote:
It feels kind of weird repeating myself so many times, so I'll just ask this question: Is there anyone who watched the Japanese vids and still thinks that Genisto's vid should be published?
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Bag of Magic Food wrote:
Hey Randil, do you think we should start a new topic for a complete Adventures of Lolo 2 run where we try to get the best times by competition on a few levels at a time?
I'll go ahead and start this now as I'm about to post some level strategies. For reference (edit 2005-11-12):
Randil's latest main levels submission: http://tasvideos.org/forum/viewtopic.php?t=2931
Randil's save states for all 51 levels: http://files.upl.silentwhisper.net/upload7/lolo2savestates.zip (applies only to variants of above submission movie)
Baxter's latest pro levels submission: http://tasvideos.org/forum/viewtopic.php?t=3054 (updated)
Randil, can you post a list of the levels that you are planning and already know how to improve? That'll help everybody concentrate their efforts in the right places. I've watched your movie and here is a list of small and big things that stood out to me. These alone ought to add up to saving at least 30 seconds.
Level 3-2. Push the egg on the water, then do some final task like grabbing a last heart or pushing around a necessary emerald block (might require dodging Gol fire), then catch the egg before it reaches the jewel box.
Level 3-5. Grab the final heart then shoot the skull, instead of vice-versa.
Level 7-1. You don't need to fully trap the last Don Medusa. For example if you push the final emerald block just one space to the right (instead of two) it'll set up the timing so you can directly dart for the jewel box. Of course, the timing will be different if you redo other parts of this level.
Level 8-2. Instead of playing it safe, grab all the hearts first then run into the brawl and blast everyone on your way to the jewel box.
Level 9-1. You don't need to spend time pushing that last emerald block around since you can just fake out Gol to get past him. For the same reason you also need not rely on Leeper to get to the jewel box.
Castle-2. Push the egg on the water, walk around and grab one of the bottom hearts last, then dodge fire from the three Gols on your way to catch the egg. This way you need not ride the slow floating egg or stall for the top-right Snakey to appear.
Final battle: You can definitely blast him much faster. =)
edit 2005-11-12: List of progress so far as shown at http://tasvideos.org/Randil.html.
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Sup. Couple more concept demos here, and now that I've tried all four stages hopefully I won't be so addicted to these pro levels anymore. :) I agree the regular levels are a greater challenge altogether due to their quantity.
New strategy on PROA: http://dehacked.2y.net/microstorage.php/info/263/Lolo2-PROA_golbottom.fcm. By making Gol reappear at the bottom the second time, you can shoot him again sooner. Although PROA in is completed here only 7 frames faster than in Baxter's recent submission, much of the first part in this demo is completely unoptimized (including a totally useless wobble at the start as you'll see). If you can make an optimized version that better times Gol's reappearance it will be at least a few whole seconds faster.
In addition, if the timing works right you can save another second by using Don Medusa to cover one of Gol's spots, and you could maybe save a couple more seconds if later parts sync better with Don Medusa's movements. The main drawback of this new strategy is that it removes the stylistic moment when Lolo fires away an egg while running straight at Don Medusa.
Minimizing lag on PROB: http://dehacked.2y.net/microstorage.php/info/264/Lolo2-PROB_minimizelag.fcm. This takes the recent update a step further by killing the lag as early as possible. This movie is 48 frames faster than Baxter's recent submission, although since the difference is small and this time I had to play carefully with frame advance throughout, it's not clear whether this improvement comes mostly from minimizing lag or getting completely different (and more convenient) Alma movements.
Both movies are rerecorded from Baxter's submission that this thread pertains to.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Well done, again.
I think it would make sense to publish this once the main Lolo 2 TAS is available as well. Hopefully this will motivate Randil to finish up the next revision of his regular levels run. =)
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Great job! Will vote yes when I have enough posts.
However, I'll finish this day showing that we're never quite done. =) Here's one more technique that I trust you'll try and make frame-perfect when you find the time. Early lag prevention on PROB saves about 40 frames: http://dehacked.2y.net/microstorage.php/info/260/Lolo2-PROB_lesslag.fcm.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Here's a new strategy for PROD that demonstrates a shorter wait period, kills one more Alma, traps another Alma, and gets one of the bottom hearts without using an emerald block: http://dehacked.2y.net/microstorage.php/info/254/Lolo2-PROD_newstrategy.fcm. Again, this is built off Randil's run so if you have a saved state from that run you can load it to the start of the last level. With some combination of these demonstrated techniques, I'm sure somebody will soon find a way to beat PROD without any stalling.
PROA will inevitably be the slowest since it's the only one that requires waiting for an enemy (Gol) to reappear. It's good to hear you managed to speed it up a lot, Baxter!
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Here's a significantly faster strategy to finish off PROD: http://dehacked.2y.net/microstorage.php/info/253/Lolo2-PROD.fcm. I rerecorded this from Randil's run, so be sure to fast-forward past PROA-C to see what I've changed in the final level.
Unfortunately, this does not address the waiting period where Lolo stalls so that the two Almas trap themselves. Until a strategy is found that remedies this, I feel this run is not complete. In comparison to the regular Lolo 2 run, the pro levels--involving side-stepping Don Medusa (PROA), rapidly blasting Almas (PROB), and dodging Gol fire (PROC)--capture the essence of a speedrun rather than an optimized-puzzle run. Although the undesired waiting period has been made somewhat entertaining, the proper climax for this run would be to avoid this by quickly manipulating and dodging the Almas (PROD), and I am quite certain that this is possible.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Slotermeyer wrote:
I know you probably didn't try to totally optimize the pro run - but just to show off a bit, Pro B can be done considerably faster (a little over 500 frames).
http://dehacked.2y.net/microstorage.php/info/220/Lolo2-PROB.fcm
My version isn't particularly optimized either, but you can see the idea.
I know you probably didn't try to totally optimize the pro run - but just to show off a bit, Pro C can be done considerably faster (a little over ? frames).
http://dehacked.2y.net/microstorage.php/info/250/Lolo2-PROC.fcm
My version isn't particularly optimized either, but you can see the idea.
-hanzou
P.S. This is how I do it on the console.
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Tilus wrote:
As mentioned several times before,
It's not mentioned in this thread or the submission text, but I'm sorry I didn't read through the other thread until now.
Tilus wrote:
With my current path, neither Earthquake nor the Missingno glitch in any fashion will be faster than what I'm doing now. I currently don't go more than a couple seconds out of my way for anything I don't need to complete the game, and the Missingno glitch will take too much time to execute to obtain any timesavers out of it.
You're right. I tried the coast trick again just now and it's surprising how long it takes. I had thought removing the dealings with Onix and X speed would have completely made up for this, but admittedly those tasks are short by comparison. The real holdback seems to be the old man himself, who is slow in pressing his buttons and can't be forced to do it faster.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
flagitious wrote:
It is listed as "Does not abuse programming errors in the game." Are there any errors that would save time, and if so what would they be?
Possibly the best known glitch in this game is the old man+Cinnabar coast trick. After reaching Cinnabar in the game, the player could capture a level 100+ Mewtwo (or some other pokemon of choice) and from then on use it as the primary battler instead of Blastoise. The time taken up in flying to Viridian and back would be made up for alone by taking out the tasks of buying X speeds, getting the TM for fissure, capturing Onix, teaching Onix fissure, switching to Onix, using X speeds, and switching back. Much additional time would be saved due to the player knocking out most trainers' pokemon in a single hit rather than multiple turns, as well as being able to get rid of virtually all 'critical hit!', 'it's super effective!', and level-up messages.
As for other glitches like encountering Missingno. to duplicate items early in the game (with the gambler trick) or surfing through walls, these might also have a use.
-hanzou
Experienced Forum User, Published Author, Player
(20)
Joined: 10/14/2005
Posts: 317
Bag of Magic Food wrote:
Would you need to have Professor Oak evaluate it? I'm really not sure.
Although Oak does evaluate the pokedex when our hero reaches the hall of fame, I recall it's just the same text as if you evaluate the pokedex using Oak's PC in a pokemon center. Thus, such a movie ought to just stop shortly after catching all 151 unless beating the Elite Four and rival is explicitly stated as a second goal.
FractalFusion wrote:
It is possible to get all 151 Pokemon, but whether such a run is interesting is another question.
Technically, it is possible to get 151 Pokemon, a bunch of glitched Pokemon (called Missingno.), 127 of every item in your inventory, and a broken cartridge.
Here's how I would picture a 151-pokemon run:
Our hero would play up through the first third of the game taking a somewhat scenic route, capturing various small pokemon like Nidoran and Caterpie that can later be evolved. He would continue on to the part with the gambler which can be abused for the so-called 'Mew glitch'. Recall that this glitch enables the player to capture not just Mew, but any pokemon in the game. Next, he would use the glitch to battle Missingno and create 100+ poke balls. He would then use the gambler glitch repeatedly to capture dozens of other pokemon who are in a first- or only-stage of evolution. These pokemon include the most familiar ones like Charmander or Bulbasaur, but also legendaries such as Moltres or Mew. This is a long and repetitive process, although it's probably faster than playing through the storyline and finding these pokemon in their natural places.
Next, he would use the gambler glitch to view Missingno again and get 100+ rare candies. These can then be used to level up various small pokemon into their evolution forms (which would probably be faster than directly catching them through more of the gambler glitch). In the midst of this, he can make appropriate use of the 'ditto trick' when a pokemon has an appropriate special stat (see above link), and this allows the trainer to catch any remaining ones out of the 151 that haven't already been caught.
This would not require playing the second half of the game, going into the safari zone, delving into dungeons to catch the legendary birds, or battling the Elite Four.
AceOfShades wrote:
Level 5-2. As shown in the above screenshots, a different strategy to save 15 seconds on this bridge/egg-float level: http://dehacked.2y.net/microstorage.php/info/276/Lolo2-5-2_southbridge.fcm. You can likely save at least another second compared to this demo run by making use of that extra shot and better positioning the floating eggs (see similar tip for level 6-2 below).
Level 6-2. The sooner Snakey's egg sinks the sooner it reappears, and waiting for the egg to appear is the biggest holdup in this level. Thus you should push the egg as far to the left as possible before floating it on the water.
I've tested and found that it can be put another full space to the left and still leave time to grab the heart, get back onboard, and shoot Gol.
Level 7-2. When playing this level on the console Leeper tends to be just an annoyance, but with some experimentation I think you could find a way to use him in order to save time on pushing emerald blocks.
Level 7-5. Two steps are wasted as you grab the last heart then walk over the egg towards the jewel box. Either grab the last heart and get back onboard the egg slightly before the egg reaches the end of the river, or switch the order in which you grab the last two hearts.
Level 8-3. In the current run, after shooting Gol you push him down three spaces then make a bridge at that row. It'll be less wasted walking for Lolo if you push the egg down only one space then put the bridge at the top of the lava.
Level 8-4. The current route involves twice waiting for Snakey to reappear and has 2 shots remaining in the end. You can change the route to shoot Snakey away earlier and make use of the second Snakey to get rid of all waiting time.
Castle-1. An alternative strategy as depicted below ought to save ~20 seconds.
Castle-5. Improvements on the current strategy: After shooting Leeper the first time, don't push him all the way down to Medusa. If you leave a full space gap between Medusa and Leeper, Leeper will have less distance to hop right before you shoot him the second time. Also, when you shoot the second time shoot as early as can be allowed, which means the egg should end up a half-space from Medusa's line of sight. Next, when pushing the egg down, you can leave a half-space between the egg and the wall before the egg hatches and Lolo puts Leeper to sleep. However, these frame optimizations are irrelevant because the optimal strategy for this level most likely involves a live Leeper. It's possible to beat this stage by relying on Leeper to hop down to the bottom when Lolo grabs the jewel box, and even if this doesn't time perfectly it could be significantly faster than having to wait for egg-hatching twice.
edit: 