Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
All right, I've finally got a good result up to entering the Kirisa Cave. I knew I was going to gain time here but I was surprised how well I was able to do. I'm now over a minute ahead of the test run. I'll post another WIP when I get a little further. It's going to be much easier going for a while so I should make progress faster.
Joined: 7/29/2004
Posts: 136
Location: Temple City, CA
tmont and axe both have my great respect for the love they've given to Crystalis. This thread, more than any other, is what keeps me coming back to the site.
"How can you prove you exist? Maybe we don't exist..."
-Vivi Ornitier (Final Fantasy IX)
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
I just figured out one optimization tmont did that I never noticed before. He unequips the shell flute in the sea to cut lag. I had noticed there was extra lag on sword swings on the dolphin but I didn't connect it to the flute. Good thing I checked back on tmont's run through this section. Even though I'll need to go to the menu just to do that it will still save a few seconds seconds overall.
Tmont's run is pretty old now and the site standards were quite a bit more lax back then. The FCEU he did it on probably didn't even have a lag counter much less all the lua functions and debuggers that I am using heavily. While it's easy to notice that it's improvable it is much, much more difficult to make a run that can't be improved. It is very time-consuming too. So you can't blame him for doing it the way he did.
Anyway, I have finished planning things up through the sea. I expect to pick up about 10 seconds more over my test run here. It's coming together quite nicely.
Joined: 6/23/2009
Posts: 2226
Location: Georgia, USA
Oh, of course! With the shell flute equipped, the game's probably checking, with each sword swing, whether you're trying to call the dolphin. Nice catch.
Used to be a frequent submissions commenter. My new computer has had some issues running emulators, so I've been here more sporadically.
Still haven't gotten around to actually TASing yet... I was going to improve Kid Dracula for GB. It seems I was beaten to it, though, with a recent awesome run by Hetfield90 and StarvinStruthers. (http://tasvideos.org/2928M.html.)
Thanks to goofydylan8 for running Gargoyle's Quest 2 because I mentioned the game! (http://tasvideos.org/2001M.html)
Thanks to feos and MESHUGGAH for taking up runs of Duck Tales 2 because of my old signature!
Thanks also to Samsara for finishing a Treasure Master run. From the submission comments:
Shoutouts and thanks to mklip2001 for arguably being the nicest and most supportive person on the forums.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
I missed it before because it seems that all the other items do a full "Nothing happens" message instead of a more subtle frame of lag. Anyway, it was really tmont who caught it and I am just picking it up from him. So thanks, tmont.
I'm aware that tmont didn't have the TAS tools that are available now back when he did the run. I feel confident that now, with current tools, I could do a better job than tmont did; however, I'm not a TASer so I wouldn't value my opinion much.
It was wrong of me to suggest that theAxeman is putting in more time in his run than tmont did, especially considering the tools that are/were available to them. For all I know, it's not true.
Anyway, that's good news.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
It was wrong of me to suggest that theAxeman is putting in more time in his run than tmont did, especially considering the tools that are/were available to them. For all I know, it's not true.
No, I am definitely putting in more time on this. The game mechanics are very complex and it is really heavy on planning and analysis so only a fraction of the time is actually spent playing the run. I have written a few hundred lines of lua, lots of notes and a few spreadsheets and optimization scripts.
The point I was trying to make is that it would have been pointless to put this kind of time in before. This sort of analysis wouldn't have been possible. Without proper insight into the mechanics you could spend an eternity messing around and never come close to what I have now. So it's not a matter of time.
It used to be that if a submission beat the current run it would be accepted without much question. Now if a run has as many known improvements as this one it would be rejected. That's how it should be. When you make a TAS you shouldn't just be aiming to beat the current record, you should be aiming for perfection.
It used to be that if a submission beat the current run it would be accepted without much question. Now if a run has as many known improvements as this one it would be rejected. That's how it should be. When you make a TAS you shouldn't just be aiming to beat the current record, you should be aiming for perfection.
In the name of publishing widespread improvements rather than possible publishing of improvements to a very small audience, I disagree. Improvements SHOULD be submitted and published, even if known improvements exist, especially if these improvements will take time to implement. It depends on the game, of course - in Crystalis one might as well include all improvements currently known as it isn't easy to go back and make changes. For some games, including these changes isn't feasable unless the run takes years.
I'm not saying the aim shouldn't be perfection, but rather that perfection takes time, and sometimes we don't all have time. I have a belief that something is better than nothing but the leadership in this community doesn't see it that way. Doing better than what exists should be the absolute target, and perfection should be the goal.
I think things are simply questioned too much now. It's all about cool tricks or games we remember as a kid; if something is foreign or "boring," even if played well, it's rejected. If something has a very minor improvement possible and it's the first run of a game, it's rejected. If huge improvements are submitted for an existing game but one minor thing was missed, it's rejected. The submission-phobia, category-phobia, imperfection-phobia, publish-phobia really sucks. And this isn't even taking into account the subjective judging constraints ("entertainment" in a speed-based run) now put on players making runs.
(by the way, I am looking forward to your run =])
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
I have everything planned out pretty closely up to Swan. I tried lots of different things through the caves and finally settled on something that gets the xp I need pretty fast, at least a few seconds faster than test run. After discovering the thing about unequipping the flute I had to try out a few more things. Also got some other interesting ideas and found a new trick I'll keep for a surprise. So it'll just take some time to tweak the lag as I play out the final cut through here. I'll post a WIP when I get to Swan.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
All right, here is the WIP up to Swan. Now 79 seconds ahead of the test run and over 6 minutes ahead of published run.
http://dehacked.2y.net/microstorage.php/info/806877994/Crystalis-Final-toSwan.fm2
Saved a lot of grinding in the waterfall cave with many tricks to get more xp faster. Earlier I talked about how random and laggy those enemies can be. Turn on the lag counter to see how I was able to do in the end and compare to other runs. This took a lot of planning and work. Compared to the earlier part of the run there is now much heavier luck manipulation.
In the caves I messed around with a lot of different flows that would have different enemies spawning. In the end I settled on this flow that gets about the same xp as in the test run but a few seconds faster. I planned out when I needed to reach certain points or wait at certain points to make enemies spawn and used extra time to cut lag.
On the Angry Sea I saved a lot of time by unequipping the flute. Using the level 2 and 3 fire attacks really capitalizes on this. The level 3 firestorm saves about half a second on each octopus and I skip one because it's faster to just take out an extra turtle. I have just enough MP in the budget to use this three times. Near the end you'll see a new trick I call "dolphin warp". You can see the dolphin disappear and reappear in tmont's run. With a few tweaks to your movement you can make him reappear just about anywhere. This saves time because the dolphin moves so much faster. It also stops the ghetto flight mp drain, allowing me to do one more firestorm attack.
Wow, six minutes! And over a minute faster than the test run too. I'm looking forward to this. :)
I'm also surprised that the level-3 attacks are worthwhile for generic leveling, given how MP-hungry they are. Or does your MP get refilled on levelup? It's been so long since I played this game that I've forgotten.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
Derakon wrote:
I'm also surprised that the level-3 attacks are worthwhile for generic leveling,
If you aren't really good or making a TAS then you probably save most of your magic for healing. When you don't need that there is plenty of MP left over. I really just need it for teleport. Still, I wouldn't say that I am using the level 3 attacks for generic leveling. Just a few special spots where they can save time.
You don't get all your mp back on levelup but you get the difference between the old and new max mp. I really just need to worry about making it to the free healing in the Goa fortress. Another big consideration is that I don't pick up the other two bracelets (water and thunder) and I won't be using the wind and fire swords very much at all for the rest of the game.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
Time for an update. I am in the fortress and reached level 14 after grinding on the spiders. I found a few nice tricks to get more experience all over. So compared to my test run I am grinding less and gaining more and more time. At this point I am now 1 minute 48 sec ahead of the test run and over 9 minutes ahead of the published run. I expect the final time to be about 41 minutes and I've got 32 minutes of that done.
One interesting thing I found is how to make the bugs fly in on the big fortress 2F area with all the water bridges. Enemies and their weapons each have their own entries in the object table. Most enemy attacks are only in the object table temporarily until they are done stabbing or the shot goes offscreen. But some enemies are always twirling around a flail. In that case, the weapon is created right after the enemy is, using up that slot. This is true even when it happens offscreen. So the problem here is that those weapons are taking up the slots where the flying insects like to spawn. By preventing some of them from spawning (by hitting select) before the insects I can get those insects flying around for more experience. Flying enemies are great for xp in big areas like this because they come to you.
I did make a discovery that could save some time but won't be in this run because it's too far back. The glitch that lets you climb the slope without falling can be done with some spells as well as the select menu. In particular, telepathy and teleport both work, even if you cancel out before actually using the spell. Not only does each of those take slightly less time than the menu but it would allow for using the sword charge glitch. Unfortunately, it would also prevent using the rabbit boots to jump because the A button would trigger the spell.
Anyway, I think this would probably be useful in both mountain areas. In Mt Sabre S we would go in with glitched fire sword and telepathy. Grinding and getting the key would be slightly slower without jumping but then we could go right up the mountain.
In Mt Sabre N we can avoid the backtrack after the boss. Go in with glitched fire sword and teleport because it's a little faster than telepathy. Climb up and then go straight to the prison area for the key. After that I'm not sure if it would be faster to go back out or take the back way that connects into the area where I grind. Anyway, after taking down the walls we can cancel the glitch by switching back to wind and grind the same as before. After beating the boss we already have the key so we don't need to go back for it. One thing though is that the backtracking gave me some experience. So there will need to be more grinding to make up for that. But I am pretty sure it is an overall gain.
Sounds great! Nine minutes is a huge amount of timesaving, but it's clear your detailed breakdown of how the game works is paying off. :) Shame about discovering a potential improvement after it's really too late to take advantage of it, but them's the breaks, I guess. Keep it up!
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Joined: 7/29/2004
Posts: 136
Location: Temple City, CA
The amount saved is truly impressive. Every time I think that you've found every possible way to save time you come up with another interesting trick. I can't wait to see the finished product.
"How can you prove you exist? Maybe we don't exist..."
-Vivi Ornitier (Final Fantasy IX)
Joined: 6/23/2009
Posts: 2226
Location: Georgia, USA
Wow, you're really close to the end of the run! Keep it up! With your latest work on this game, you've taken what's normally a grinding-heavy repetitive game and made it genuinely surprising.
Used to be a frequent submissions commenter. My new computer has had some issues running emulators, so I've been here more sporadically.
Still haven't gotten around to actually TASing yet... I was going to improve Kid Dracula for GB. It seems I was beaten to it, though, with a recent awesome run by Hetfield90 and StarvinStruthers. (http://tasvideos.org/2928M.html.)
Thanks to goofydylan8 for running Gargoyle's Quest 2 because I mentioned the game! (http://tasvideos.org/2001M.html)
Thanks to feos and MESHUGGAH for taking up runs of Duck Tales 2 because of my old signature!
Thanks also to Samsara for finishing a Treasure Master run. From the submission comments:
Shoutouts and thanks to mklip2001 for arguably being the nicest and most supportive person on the forums.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
It's coming along nicely. I have the overall plan done for the rest of it and am working on tweaking some lag out of Mt Hydra right now. I also want to make a nice viewing lua script before I submit so I might not finish before the end of the year. This is the last really annoying section though, because the floating eye enemies are the last ones I need to deal with that have completely random behavior.
I found a few improvements over my test run near the end. There is going to be even less grinding and about 12 seconds saved. So the final version will be about 40:50, a little more than two minutes faster than the test run.
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
Here's how the big dragon fight works:
He moves back and forth, making a random roll every 32 frames. That roll determines if he does an action or not. The type of action is determined by the almighty $0008 counter. Now if you don't take a hit the action will never be lasers. But if you get hit a flag gets set that lets him do lasers. In that case, he does lasers if the counter is 0-7F and bouncies otherwise. So getting the right behavior requires manipulating the counter and random seed. Also, this means that there is a two-second window for lasers followed by two seconds that are no good.
I really wanted to reach another breakpoint on the counter but ended up quite a bit short. So I used that extra counter time to wring out lag from pyramid dungeons. The result is pretty good. Now I just have the big dragon fight and the tower left. I think the toughest part is going to be figuring out how to get DYNA to use its lasers. It really must be done because it makes a difference of about 1.5 seconds of lag.
