User Files from FatRatKnight

The only significant difference is the fact I rearranged the elements. The map/Boss HP is toward the top, so as a boss takes damage, the relation of its HP and that health bar should be more apparent by proximity.

It's still just a 40 pixel width display.

So, I did stuff at Stage 8 boss. Hopefully it's a better show. One key problem is that, somehow, an entire stage later, I get a desync at stage select from an extra lag frame in transition.

I had no idea this was possible. Anyone up for some deep analysis in the functions?

One could also try to take pieces of the older and newer movies. It's a good bet I left intact the exact positions of where I released each charge shot, so direct copies of antics between shots might work.

The Stage 4 boss should be approximately 1.4 times as entertaining as before. I'm sorry, but I wasn't able to make it 10 times like I did my Overlord submission. Of course, 10 times nothing is still nothing...

I have no idea how I come up with these numbers for entertainment.

I used an enemy in reach of the save platform to trigger meteors.

Apparently, spikeball sparkles in dirt form do not like being hit, and will summon meteors from the heavens when they are destroyed against an enemy. Strange destructor behavior for a projectile, but then again, we do have MurderBarrier...

How can one handle this lag, though?

ThunderAxe31 needed my help on something. Here's how I ended the boss fight.

Currently sitting on 1/8 delay due to dialog (burning the frame before loading the door), and we're not using the save platform!

Complicated business. Basically animation frame rule involving the door. By slowing down my ascent on the wall below the save platform, I trigger the next segment a few frames later. I also begin moving right a few frames later as well, but my movement delay is not as big as the segment delay. As a result, the door itself has existed one fewer frame by the time we reach it, which is enough to beat an animation rule, and by extension, use those savings to beat the dialog frame rule.

Complicated business. Advance with no regrets.

Inserted 3 frames delay just before save platform because I could. Well, mostly because it helps to have a standard spot relative to frame rule to trigger the boss fight from. Note the save platform route allowed me these 3 frames delay, saving 4 normal frames to get there, but 3 lag frames is the exchange.

Ideally, we find another frame somewhere, so we can take the sneaky route to avoid lag. But at least we have our frame rule here. Without the save platform route, we're looking at 7/8 dialog frame rule delay.

I dunno... I think it looks normal.
Hint: This really does not look normal.

Basically, I try to fill up the array as quickly as I can. By the time I start falling, it has already been maxed out. It has interesting effects on the appearance of pretty much anything.

This is what the game does when it gives up and reuses the first spot in the array for nearly everything at the same time.

Just some small stuff. Saved 2 lag on the miniboss by jumping one frame later. I also leaked less memory, if that helps any.

The next part looks terrifying. I think we have 3 frames to spare, based on keeping up with staying 1 dialog frame rule ahead of our prior run.

This is in case we can't manipulate our way into keeping those 8 frames. I did a lazy-sync in mid-stage and the movie goes clear out to where it ends.

Still, the original route with the better item drops has a lag frame in it, and this version, while it loses us 8 frames and compels us to match said item drops in the later half, removes a lag frame the other item drops were stuck with.

Oh, and I did some input clean-up in the rest of the movie. Nothing that affects sync, just marking out frame rules properly and all that.

P.S. The Stage 4 boss almost certainly will not let us avoid all damage. We are rather lacking in ammo, and need at least 5 hits of 8 damage from our spikeballs. The remaining 15 HP can be taken down in 6 hits of 4, 3, 2, 2, 2, and 2, in case you want to remove one Charge Storage, and we have exactly the ammo necessary for that.

A small little extra for Zook to do. Among my actions included sticking to the wall better so the camera is less bumpy.

Curious if "machine gun" dashing is as annoying as seasickness camera movements. I'll leave it to you to decide.

There is an array at 7C44 that gets filled when objects are allocated, and freed up when objects are removed.

Colors:
Black - Free
Cyan - Player bullet
Green - Item
Yellow - Enemy (w/ 0+ HP reported)
Purple - Enemy (w/ negative HP reported)
Red - No detected object -- Memory leak likely

The more filled the bar, the more likely the lag. At least, that's the idea I want to transmit.

... I'm just not feeling rather vocal right now.

Nonfunctional. Do not run.

Been looking over my older stuff to find this script. It is an attempt at botting (or more accurately, simulating) the first battle, so that I can see this fight perfected, but it is decidedly incomplete, and it is not structured to treat the fodder with ambiguous stats, so it needs a redesign anyway.

Note that the script shouldn't be running emulator functions. It's just a scripted simulator, not a savestate bot.

Should involve less seasickness.

Mind, ideas are rather minimal for one-dimensional movement. If there are better ways to dance to the music between the jumps, go for it.

I thought I gave up on this stuff. Okay, fine. Apparently I just can't keep away. Also has the collected 7 frames delay before s8 door.

I hope intro and Stage 8 have better looking miniboss fights. Well, better than standing there staring out into space, anyway. Splice them into whatever other entertainment stuff that was done.

Moved the delay frames elsewhere. The spot I chose is before the door object itself loads, not the short spot of floor leading up to the door. The exact position is when we landed on a platform, where if we insert a single blank frame, our two frames of not moving right would overlap with one frame of landing lag, so that the one blank really does mean one frame delay, not two.

We are now delayed 7 frames due to frame rule. I adjusted the boss for the case where we have zero frame delay on the frame rule, so I eat the 7 frames before loading the door.

Animation frame rule after the boss is still 1 frame loss. It is nice to know the slightly different boss timing could have tolerated killing him one frame later, but we're fine either way and didn't need that tolerance.

Was this the thing needed for sync?

f36966 *** STAGE 5 ***
f37699 Segment 0: 733
f38175 Segment 1: 476 (pr: 479)
f38482 Segment 2: 307
f38605 Segment 3: 123
f38682 Segment 4: 77
f38797 Segment 5: 115
f39250 Segment 6: 453 (pr: 456)
f39962 Segment 7: 712 (pr: 734)
f40256 Segment 8: 294 (pr: 277)
f40373 Segment 9: 117
f40611 Segment10: 238 (pr: 242)
f42017 Segment11:1406 (pr:1407)
=== END OF STAGE 5 (Frames: 5051) === (pr:5067)

Seg1: ThunderAxe31's fault. Shot Counter glitch lets us switch out of bubble barrier without needing to pause.
Seg6: ThunderAxe31, same basic method.
Seg7: Different route, we don't climb wall right away, loading next segment sooner.
Seg8: It started sooner thanks to route change, so SegmentCounter claimed it took longer.
Seg10: Used to be on bad side of frame rule. We're good now.
Seg11: We're on the good side of the frame rule now. Oh, and I tried avoiding damage.

Sitting on 16 frames delay after the stage now.

ThunderAxe31 went and did Shot Counter glitch stuff. This saves us from having to pause when refreshing our next two bubble barriers.

Do make sure the boss is made fun of. I haven't put in the time to do that, someone else might want to.

I hope I understood the branches right. Should be one that clears the start of s13 by wall jump rather than ladder. Another frame rule over ThunderAxe31's successful try.

You know, reveal a potential, and I end up exploring it. And then I come back after I turn the potential into something. Boss has not been synced, that still needs to be taken care of.

Some black magic with wall jumps were involved. I suspect frame rules, but it seems the apex of a wall jump actually has different velocities depending on what frame you wall jumped. I merely delayed it two frames, and got the exact black magic pixel we needed. Oh, and a short dash so I wall jump from just one pixel further. I think it was needed.

Once things become difficult to find, I don't get to them. As others are still scanning over it, each thing that is found can then be analyzed by me, and results can still happen.

Exonym found a small improvement. I went and implemented it.

Somehow, 2 frames turned into 4. Currently at 4 frames delay in stage select.

I've also taken a moment to measure stage entry frame rules. I've added Select presses at each stage select to indicate how many frames are delayed by these frame rules. Basically, things should sync if you delay just the A press up to as late as the select presses exist.

Something interesting happened at the end of this lap. The Rival Radar indicated that the game has no problems with spawning a CPU car on top of another CPU car! One rocketed forward to my position, and I got a relatively soft bump from it.

Yes, one of my problems is trying to figure out how to get 600 km/h bumps. As such, I'm too slow. I'm one frame behind WR opener. What did they do and how can I match that? I'm uploading this primarily because of an interesting event at the end of the lap, not because of my poor driving skills.

I want to take a break. I really do. Yet I come back and here's what I got.

The last of the end stages. Applied MurderBarrier. Shot charged star to the right instead of to the left. End movie before the segment has transitioned to the boss room. Oh, and I changed a wasteful bubble shot against a boss to the dirt weapon, so that I had enough ammo for the barrier.

It's a nice amount of frames. 81 compared to before. That's one easy improvement out of the way. At last, MurderBarrier finds use!

f62756 *** STAGE 9 ***
f63314 Segment 0: 558
f64671 Segment 1:1357
=== END OF STAGE 9 (Frames: 1915) ===
f65031 *** STAGE 10 ***
f65604 Segment 0: 573
f65784 Segment 1: 180
=== END OF STAGE 10 (Frames: 753) ===
f66149 *** STAGE 11 ***
f66731 Segment 0: 582
f68353 Segment 1:1622
=== END OF STAGE 11 (Frames: 2204) ===
f68709 *** STAGE 12 ***
f68829 Segment 0: 120
f68884 Segment 1:_ 55
f69303 Segment 2: 419
f69485 Segment 3: 182
=== END OF STAGE 12 (Frames: 776) ===
f69831 *** STAGE 13 ***
f69977 Segment 0: 146
f70079 Segment 1: 102
f70490 Segment 2: 411
f71954 Segment 3:1464
=== END OF STAGE 13 (Frames: 2123) ===
f72306 *** STAGE 14 ***
f72934 Segment 0: 628
f73121 Segment 1: 187
=== END OF STAGE 14 (Frames: 815) ===
f73459 *** STAGE 15 ***
f74082 Segment 0: 623
f74256 Segment 1: 174
=== END OF STAGE 15 (Frames: 797) ===
f74613 *** STAGE 16 ***
f74728 Segment 0: 115
f74938 Segment 1: 210
f75232 Segment 2: 294
f75425 Segment 3: 193
=== END OF STAGE 16 (Frames: 812) ===

There's probably plenty of obvious time savers in there. I just wanted a complete movie so that all the skips are decidedly included in it.

Probably should have shot to the right for the finisher. A note for later.

The difference of this movie and the leading team (T4) is 8364 frames. I expect general optimizations to improve this by a few hundred more.

f56625 *** STAGE 2 ***
f56849 Segment 0: 224
f58448 Segment 1:1599
f58509 Segment 2:_ 61
f58610 Segment 3: 101
f58838 Segment 4: 228
f59032 Segment 5: 194
f59814 Segment 6: 782 (pr: 778)
f59905 Segment 7:_ 91 (pr:_ 96)
f60033 Segment 8: 128 (pr: 127)
f60124 Segment 9:_ 91 (pr:_ 92)
f60394 Segment10: 270 (pr: 272)
f61717 Segment11:1323 (pr:1328)
=== END OF STAGE 2 (Frames: 5092) === (pr:5100)

Implemented Exonym's idea, and ended up with this mess. Apparently slower in the immediate segment, oh but we make it all back up in the next. Yet we lose a frame and break even on the next transition, but don't worry, the better camera makes it all better before the door, and you saved a frame rule.

I mean, sure. I did end up with something faster overall. I'm just really confused how it works out.

More specifically, I did a turnaround landing before Flame Charge. I did not Flame Charge out of the air and onto the slope. Thing is, if you're not on the ground when doing the Flame Charge, the next time you come into contact with a floor tile, you will stick there and be forced to wait it out until the Flame Charge ends. The slope tiles don't block you, and the "magnet" tiles found around slopes stick you to the slope, so we're good there, except again, once the slope ends and we move on to floor tiles, we get stopped. Walk a few frames so that we don't end our dash before that Flame Charge out.

Oh, and this file has my hastily attempted parts of the end stages tacked on. Yes, I'm having trouble manipulating ammo drops properly. I'm a bit too slow on s12, the Stage 4 look-alike.

#5361: Dooty's Genesis QuackShot: Starring Donald Duck in 19:21.63
The script appears to be working. Just run that submission on the ROM the submission is for, and start this script on the first frame. I can't say with 100% certainty the script is working as intended, but it should be pretty good.

Spikestuff, please let me know if this script is good enough to automate data collection for that lua-assisted subtitles. I've isolated functions that I expect you to change at the very top, and other functions that might need to be changed near the top. Other stuff below includes me trying to hack in the code that will do the stuff, and some ctrl+H stuff to make things work.