Topic by Comicalflop

Posted: 8/12/2007 1:46 PM
(Edited: 4/22/2008 11:25 AM)

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

I think there is also one more part that is unclear (you've given helpful info when I've asked about techniques and such, so that's good) and that is that a while ago you used an emulator with slowdown to try and experiment with the game: "wow... i actually think it is harder to accomplish while slowed down :S... anyways... some findings... THESE ARE NOT VALID TIMES 1-1 6.47 2-2 14.80 2-3 8.87 2-4 35.15 2-5 8.92 2-7 17.55 2-10 15.30 3-1 13.95 3-4 14.37 THESE ARE NOT VALID TIMES all of these times used emulator slow down and the boost grab was taken full advantage of.. " I know that you used slowdown and got times faster than your console runs, but somehow on 2 of those levels you managed to beat Nitsuja's times, who uses frame advance, and thousands of rerecords. TAS 3-1 time: 19.19 Your 3-1 time: 13.95 TAS 3-4 time: 17.47 Your 3-4 time: 14.37 So maybe you can help shed some light on where you gained time in those levels, and how you play tool assisted differently that makes it faster.

☣ Homepage ☣ Retired

Posted: 8/13/2007 3:41 AM

Quote

Spacecow

He/Him

Joined: 6/21/2004
Posts: 247
Location: New Hampshire

comicalflop wrote:

I know that you used slowdown and got times faster than your console runs, but somehow on 2 of those levels you managed to beat Nitsuja's times, who uses frame advance, and thousands of rerecords. TAS 3-1 time: 19.19 Your 3-1 time: 13.95 TAS 3-4 time: 17.47 Your 3-4 time: 14.37 So maybe you can help shed some light on where you gained time in those levels, and how you play tool assisted differently that makes it faster.

I think you have your times slightly confused. It looks like Veysey got the 17.47 time unassisted http://www.youtube.com/watch?v=exoUMcAcLTk ... And looking at nitsuja's last WIP, he got 14.18 on that level, not 17.47 (or 14.37). nitsuja also got 16.42 on 3-1, not 19.19. Veysey's unassisted run is 27.27, though, so I don't know where the other numbers are coming from this time.

Posted: 8/13/2007 11:21 AM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

Edit: Veysey's times I posted that need verifying were using an emulator with slowdown, not console. *shrug*

☣ Homepage ☣ Retired

Posted: 8/15/2007 4:30 PM

Quote

Veysey

Joined: 8/16/2006
Posts: 24
Location: New Brunswick

I think I answered this somewhere before, but especially in 3-1 - I'm sure the emulator was at fault - when asked about this before, I went back to see and the screen didn't move up nearly as far as it does on console where the bombable blocks are - so I believe the difference in the time was credited to not having to go as high... I'm pretty sure it was done with project64? I can't specifically remember - but I do know I didn't need to take time to stop and further up at those blocks where Nitsuja did in his WIP (If I recall at least) Hope that helps

Posted: 8/17/2007 8:57 PM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

Yeah that seems logical. I also remember you mentioning that all of those times were achieved with almost perfect boost grabbing (not TAS frame perfect but slowdown achieved goodness), not going at all faster than normal slide jump speed, so that the theoretical limit was almost reached. Sub-pixel perfection really makes a difference in this game, making an in-game timer goal the best choice... somewhere in between 1-01 and 1-02 my 19 frame gain became 17 due to game randomness, but I still had gained 0.01" on the timer at the halfway star (I still need to redo that segment again anyway because of a slight sub-pixel error I made in slide boost jumping vs. boost grabbing. 307200 does not equal 327200 >_< So maybe I can squeeze in another 0.01 hopefully.) In the end this means that most of the levels will be improved by a minuscule amount, which hopefully is what viewers want to see; but with luck major time can be saved in some stages more than the already known ones. Nitsuja: Is the current TAS Input Plugin's combo feature working correctly? Those boost grabs, especially in testing over large distances for the perfect angle and approach, can be quite tedious.

☣ Homepage ☣ Retired

Posted: 8/18/2007 5:43 PM

Quote

Veysey

Joined: 8/16/2006
Posts: 24
Location: New Brunswick

While I'm not much into tas' that much... I am excited to see what times can be squeezed out of this game... especially worlds 4-5 (where Nitsuja left off) Shedding some light on some of the theories that we developed solely based off non-tool assisted speedrunning will be nice as well as I'm sure it will only help the MM community as a whole (as few of us as there are) Can you possibly further elaborate on the speed of slide jumping and boost grabbing? Like how long does it take you to achieve max BG speed? and those numbers you use - over what course of time will consistently and accurately BGing overtake a series of slide jumps timewise? When speed running on a console getting 3-4 perfect/near perfect BGs is about all you can realistically expect...

Posted: 8/26/2007 12:47 AM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

Ok, so TAS wise you really should be BGing all the time, and slide jumping should be used as infrequently as possible (except when needed for obvious reasons.) This I deducted from memory watching. Perfectly timed BG's preserve the fastest possible SJ speed as long as you can go, and SJ's have exactly one frame where you are slightly slower than your fastest BG maintaining speed. So, in order to maximize sub-pixels (which I will be doing), and to hopefully attain 0.01" on the in-game timer after a distance, I probably will be BGing all the time. Of course, the one frame slightly-slower speed will probably not make any difference at all across short distances, but one can never be too sure (or I can test and leave in what doesn't lose any time. BGing can be boring.) The difference is not noticeable enough for normal speedrunning.... so you can Slide Jump to your heart's content. Achieving max BG speed can happen as soon as you want. Once you do a slide jump, you'll be flying in the air at a speed of X, and assuming you do 100% perfectly timed BG's, X will not change. I imagine that the number of secrets left in this game is really starting to decline... But there is one biggie left, and that is the other method of squeezing through walls that you told me about Veysey, by grabbing projectiles... do you know if you have a video example of this, or know of a "sweet spot" like in 4-01 where it can be done consistently, so I can do some research on it and figure it out?

☣ Homepage ☣ Retired

Posted: 8/26/2007 5:26 PM

Quote

Veysey

Joined: 8/16/2006
Posts: 24
Location: New Brunswick

I don't have a video of it... It happened only once on Clanball lift ... I was doing an extended slide jump to get over the first tall... pillar?... (where the two clancers in the floaty mechs are) and one of them shot at me... I'm PRETTY sure I ended up turned around grabbing the oncoming shot while I was close to the corner shaped like an upside-down L... I ended up going into the wall and popped out the other side... all this happened at the 12 second mark of this video http://www.youtube.com/watch?v=exoUMcAcLTk You might also note that there is a triangular block at the top which at this point in time seems to be a likely culprit of our collision issues >_< Edit: Thanks for the explination

Posted: 9/1/2007 1:05 AM
(Edited: 9/1/2007 4:58 AM)

Post subject: WIP, .m64 updated

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

It's gotta be those triangles. I get into those all the time, plus: they have upwards ejection routines, as well as ejection routines that place you inside other blocks. Hopefully I can learn how to make MHS create tile maps of all the stages so that I can scour each level for triangle blocks. A few people have requested a WIP, Here's one (youtube) to satiate appetites, completing stage 1-05. Edit: 1-05 is now 18.67". I'm going to retry 1-06 and see if I don't have that same problem with having to wait 30 frames. Link above is updated, download that for the latest 0.20" improvement.

☣ Homepage ☣ Retired

Posted: 9/1/2007 4:48 AM

Quote

FreshFeeling

Joined: 8/13/2005
Posts: 356
Location: Canada

Wow. That was very fun to watch... despite what it looked like on youtube. I'm going to watch it again now in mupen. This game really is amazing TAS material.

Posted: 9/1/2007 4:56 AM

Quote

SaxxonPike

Any

Joined: 6/8/2005
Posts: 236
Location: Madison, Wisconsin

The game seems a lot more interesting to watch a TAS of with a WIP attached to it. Thanks :) I'll definitely be looking forward to seeing more of the same essence of awesome.

Music - https://www.sanxion7.com Code - https://github.com/SaxxonPike

Posted: 9/1/2007 6:59 AM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

comicalflop: Nice run, it looks good, are you ahead of Nitsuja's old WIP. I had hoped he would finish that run, but I'm glad that someone is still working on this game. It is an extremely TASworthy game too.

Posted: 9/1/2007 7:10 AM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

I am ahead of Nitsuja, both in frame count* and in in-game time. I've so far beaten 5 of his times, even in ones I thought were incredibly unimproveable, like 1-01, and 1-04. *I'm behind by roughly 1 second so far in frames. The frame counts for this game is weird. Not only can lag do weird things (I once saved 3 frames from lag, and the in-game timer went up by 0.09") but the input counter is higher than the real-time counter. What is even stranger is that in recording mode I might do something at, say, frame #12354, and then when I watch the run in read-only that action might happen at frame #12370, and when I go back to recording mode, I will get to that action at frame #12370. Frames are being inserted in some places for strange reasons, and yet the in-game times I'm achieving are not changing at all. And everything still synchs up fine, but I still don't know what the deal is. In the end, because the run is aiming for overall fastest in-game time completion, I'm not going to worry about it, since it's not screwing up the times. It would seem that what Nitsuja wanted most- 1) nothing new to be discovered anymore and 2) memory watching- has been achieved, which is why I'm working on it. I know it's not a good karma thing for me in particular to say, but I really want this one fully completed, I hope that I can beat if not match every TAS time that Nitsuja had made, blow the 4th and 5th worlds' real time records out of the water, that I can do this game justice, and to (fingers crossed) submit the entire movie as a very worthy N64 addition to the site.

☣ Homepage ☣ Retired

Posted: 9/1/2007 6:52 PM

Quote

Bag_of_Magic_Food

He/Him

Active player (278)

Joined: 5/29/2004
Posts: 5712

Hmm, weird, do you think one of the counters is falling behind for some reason?

put yourself in my rocketpack if that poochie is one outrageous dude

Posted: 9/1/2007 7:22 PM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

I dunno. But I just found that the input frame counters is the more accurate counter which I should be watching, which increments by two's because the game reads it like this: frame input 1 A 2 A 3 4 5 C> 6 C> 7 S 8 S 9 10 In real time frames (the left counter) I am 54 frames behind Nitsuja, but by the input frame counter (the right one incrementing by 2) I am 57 ahead.

nitsuja wrote:

I think I've gotten as far as I'm able/willing to in modifying the glN64 plugin. The result is that - the indoor textures in Mischief Makers wrap correctly - the boulder in level 4-1 of Mischief Makers is visible - the plugin reclaims texture memory at a reasonable rate (based on time of last use) - very old video cards should get more accurate results than before (on all games) I'll make the plugin available now in case someone wants to try it out: plugin dll (mirror) source code (mirror) EDIT: But don't use this for games besides Mischief Makers. I haven't tried it on a lot of them so chances are good that it makes the graphics worse in some other game(s).

Is this going to be a problem? Should I have used this plugin to start with? Do I have to redo to implement this (which I hope to all god that I don't have to do, I just had hex help to complete stage 1-06 the way I wanted it to), or will recording in Jabo's 1.6 still make it watchable with this new one? If I don't have to redo and can keep using 1.6 then that's great, but if it's worth it to use this one, lemme know now.

☣ Homepage ☣ Retired

Posted: 9/1/2007 8:20 PM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

comicalflop wrote:

Do I have to redo to implement this (which I hope to all god that I don't have to do, I just had hex help to complete stage 1-06 the way I wanted it to), or will recording in Jabo's 1.6 still make it watchable with this new one? If I don't have to redo and can keep using 1.6 then that's great, but if it's worth it to use this one, lemme know now.

There's a chance that it can still sync even if you use another graphics plugin. Some games are more picky than others, what you can do is try to replay it using the new plugin, and if it still syncs the same, continue using it. You can probably hex-edit the m64 to say the correct plugin too.

Posted: 9/1/2007 8:39 PM
(Edited: 9/5/2007 2:16 AM)

Post subject: demo WIP of 1-08 present

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

I just tested it and it seems to work just fine. Recording, replaying, whatever, it doesn't affect synch at all. I may switch to using recording the gln64 4.1.1 in the chance that I need it to see certain textures, like the math questions, etc. But for the meantime I'll just use Jabo's (also since when encoding movies it goes very, very, very slow and freezes things up when i try to use the 4.1.1.) In the sub text I can probably say that the recommended plugin to watch is with the gln64, but that it was recorded with Jabo's and works fine either way. A published avi should probably be made with the enhanced gln64, but for me it was a very aggravating experience with encoding. But the publishers will deal with that one I guess.

☣ Homepage ☣ Retired

Posted: 9/5/2007 2:15 AM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

One last, final WIP for the masses. Phew! World 1 done. Me making tangible progress? Who'd have think it! Every stage beats the previous TAS time, except for the last. Phooey. Enjoy. Now I'm going to go eat me some lava. And then snow after that. Then rocks. And then finally some metal.

☣ Homepage ☣ Retired

Posted: 9/5/2007 2:51 AM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

comicalflop wrote:

Every stage beats the previous TAS time, except for the last. Phooey.

Is this because of a difference of real time vs game time? Is each level faster on the framecount, or were you just not able to match the previous time? Now is the time to try again, it will be more difficult to fix later. Although with the separate levels, hex-editing seems possible, but probably painful anyhow.

Posted: 9/5/2007 3:05 AM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

The entire run is based on in-game time. The first 9 stages all improve the in-game time shown. A few times I've made real time frame sacrifices for in-game frame sacrifices (lag, or something else.) At one point, I was 54 real time frames behind, and 57 in-game time frames ahead. (I forget what count I'm at now, but I *have* to be ahead by real time frames now as well. I've gained so much time.) For 1-10, there is just not enough inside the level to make me gain any time. I wall clipped a few times, and boost grabbed most of the stage, but no gain. I was gaining a pixel or two and some sub pixels, but I think that's the absolute limit. DeHackEd was able to hex one of Nitsuja's stages onto mine. I think this game is incredibly hex friendly, because so much of it is in-game time based and not frame based, so that I can redo levels as needed. (I just need to learn hex editing so I don't bug DeHackEd too much.)

☣ Homepage ☣ Retired

Posted: 9/5/2007 3:30 PM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

It would seem the in-game timer is not updated on frames with lag. This make the decision between the two very icky as you could have a very laggy run with perfect time, or a faster, smoother run with more IGT. I haven't confirmed this, but I may search some RAM addresses to try to observe the effects of lag.

Posted: 9/5/2007 4:41 PM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

Yeah, lag doesn't update the in-game time unfairly, same as with Super Metroid and other games. I used this to my advantage in 1-03 I think, where I tested one route that saved 3 real time frames and lost 0.09" on the timer (the left frame counter isn't too accurate to begin with anyways, I use the input counter for measurements), as compared to what I did which increased lag. While you're at it, could you find the RAM address (or addresses, if each stage has its own separate timer address) for in-game time? Deign found one that sort of conked out halfway through 1-02 and ended up being more inaccurate than I thought. (I use MHS Tool for watching btw) In terms of lag, if I can avoid it without wasting in-game time I probably will do so, especially since lag slows things done and ruins entertainment a little bit. I lagged things up a bit in 1-08, but it was for Marina to do some impressive maneuvers so I'm keeping it. Some things are just naturally laggy like Western World and one of the world 2 levels, so that I can't help. Removing lag is not going to be a chief goal of mine in general, but I probably won't purposefully lag up the game like all hell just to lower the time slightly.

☣ Homepage ☣ Retired

Posted: 9/5/2007 6:09 PM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

I found what I believe to be the counter for the total in-game time. This should hopefully be more stable than the previous times. I guessed correctly that it tracks time in frames, and each frame is 1/60th of a sec. This is the time that is displayed on the pause screen as well. Game Time Unsigned Long Addr: 00A20638

Posted: 9/5/2007 6:35 PM

Quote

Comicalflop

Editor, Experienced player (730)

Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA

Very nice (and quickly found, kudos.) However, the only thing is that it seems to be merely a frame count that is not affected by lag (which is helpful, but not what I was looking for). When the screen is paused, and during cutscenes, and when selecting levels, the in-game time is not supposed to change at all, yet this address is increasing by 1 each frame whenever the game is running. A more accurate frame count if you will. (I will add this address though, since it should be useful.) What you found is similar to what Deign had found in that it increases by 1 each frame... except his didn't start increasing until 1-01 started, but then halfway through 1-02 it stopped completely. This game tracks in-game time by hundredths of a second, so the in-game timer should be increasing by 0.01666667" a frame, since the game runs at 60 FPS and 100/60 = 1.666667. What I need is an address that only increases when the in-game timer is increasing, and very helpful if that address resets once a stage is done. This in particular is what I am most looking for, as it gives me the absolute best way of comparing to Nitsuja's levels and to test routes/strats. if you can find one that does that, I'll be most appreciative.

☣ Homepage ☣ Retired

Posted: 9/5/2007 9:12 PM

Quote

zdude255

Joined: 6/26/2007
Posts: 147

IIRC The S-Rank guide says that the level clock does run during the pause screen, so accidentally pressing pause will ruin a potential run. This doesn't effect a TAS at all though, since we can pause the emulator. It looks like that value refers to "Play Time" on a file. It does start incrementing at the title screen, but is loaded with the file. It is the original data as well, because you can set it and reset the time spent in-game. Even though it displays time in milliseconds, that doesn't mean it tracks it in them. It's usually simpler to keep track of frames and convert it to time when it's displayed. So far I haven't found an address that directly corresponds to level time in frames, but I have done some experimenting with delaying the B-button to grab the end star, a frame and then two frames to test the granularity, and it does indeed look like 1/60th of a second. I'll probably try again tomorrow, I've found a few (possible) frame counters, but haven't had their values match the level time yet.