Submission #6193: Flip's Genesis James Pond 2: Codename: RoboCod in 23:39.58

(Link to video)
Sega Genesis
baseline
Bizhawk 2.2.1
85065
59.922751013550524
108386
Unknown
Submitted by Flip on 12/16/2018 12:14:39 PM
Submission Comments
Pond is just in time to save Christmas, and save a nice 2983 frames over the previous TAS while he's at it.

Game objectives

  • Emulator used: Bizhawk v2.2.1
  • Aims for fastest completion
  • Takes damage to save time

Improvements

SectionDifference
Intro+1
Castle Movement-8
Level 01-6
Level 02-78
Level 03-2
Level 04-120
Level 05-152
Level 06-37
Level 07-228
Level 08-332
Level 09-46
Level 10-386
Level 11-314
Level 12-4
Level 13-1264
Level 14-6
Total-2983

Emulator

Bizhawk adds 1 frame to most of the loading screens, but is otherwise overshadowed by all the improvements found in the run.

Boss Battles

A boss's 100 frame invincibility timer only decreases while they're on screen, and doesn't tick if they're not currently visible. Special care was taken to make sure that we always have the boss in our sight, to remove any extra waiting which the previous TAS did not do. Note that this is does apply to the level end sequence, and so we're free to jump around once the boss is dead.

Mystery Bonus

Every 4th enemy killed drops a special bonus in the form of a pair of false teeth. Collecting these is what awards the Mystery Bonus at the end of the level, for 500 points per pickup, and annoying ~70 frames to give it to us. This was completely overlooked in the previous TAS, collecting a lot of unnecessary bonuses. Proper effort was undertook to avoid all of these, which also includes controlling which enemies actual get killed in order to make sure the bonuses never spawn in our way.

HP Management

HP management was otherwise fine on the previous TAS, ending with only 2 HP by the end, compared with potentially 3 HP for this TAS. It's still beneficial to lose that spare HP however, since it means 3 fewer frames on each score screen. The only chance that we have to get hurt without losing time was during the on-rails section in level 11, which ended up saving 9 frames over the next 3 levels.

Movement Optimisation.

Speed Control.

Basic speed is 3, with no subpixel value to worry about. The next best thing is the subspeed value, which gets truncated but nevertheless is constantly flickering and is still a crucial variable. The subvalue will be affected by how we choose to handle slopes, and we'll want to keep it as high as possible.

Landing Boosts

When falling from a significant height, Pond will gain a quick speed boost, which can increase our speed for anywhere between 1-3 frames according to our subspeed value. More ledges have been utilised in this run, and more effort was taken to make sure we get the maximum boost possible by landing when our subvalue is highest. Note that height difference is based our initial/final height, and so constant jumping doesn't work for constant boosts.

Pathfinding

Don't be fooled by the 'flat' ground, it's anything but, especially the yellow ground. Pond reacts to all the individual nooks and crannies formed by the pixels on the floor which results in constant flickering to acceleration and produces horribly inconsistent velocities across each section. This is always annoying to predict and to optimise to try to find paths which utilise as many of these potential floor boosts as possible in a given area, causing tonnes of strategies to test.

Preserved Y velocity.

Landing inside tight corridors doesn't reset Y velocity, which preserves it until falling off the next ledge. This can also be achieved by trying to 'fall faster' when in vertical ascenders, which can be seen on level 05.

Level improvements

Castle Movement.

Slightly better movement when going between entrances which are located on the left hand side, and the occasional window ledge can be used to give us extra landing boosts.

Level 01

Faster to jump onto the shoes and take advantage of the bumpy ground.

Level 02

Better initial jumps down the thin descents, and used Y Velocity conservation before the final end for a faster drop.

Level 03

Better final hit to the boss, which is the only one which matters.

Level 04

Generic improvements everywhere.

Level 05

Preserved Y velocity in area 4's tight corridors, to make each drop faster. Most noticeable just before the trampoline, where we can use the vertical ascender to 'fall', and keep velocity for when we drop off the next edge.

Level 06

Boss kept on screen at all times not delay invincibility countdown.

Level 07

Ejected from the bathtub earlier, to take advantage of the slope acceleration

Level 08

Generic improvements across all areas. Archanfel also discovered that the wheels can be delayed through camera manipulation, allowing one to be in prime position to bounce on to get to the next tier, saving over 200 frames.

Level 09

Boss kept on screen at all times not delay invincibility countdown.

Level 10

Special thanks to Archanfel for discovering it's possible to jump directly to the first exit, in addition to showing it's faster to reroute the bathtub through a bonus zone for a faster exit to the area.

Level 11

More speed boosts across bumpy things, involving a lot more effort put into testing combinations to get across all the various slopes. For the yellow area, killing the first bird was skipped, and after death respawn it's faster to just jump into the next corridor. When going across the piano, using HP for invincibility proved more useful at the start compared to when going down the slope. Afterwards, spare HP was then lost during the on-rails section, to reduce each subsequent level's score screen by 3 frames a piece.

Level 12

Laggy boss fight, but better timing on jumps reduced the end level sequence by 1 frame.

Level 13

A slightly faster start, and using the cannon to shoot into the platform provides a faster route. Afterwards, comes the most significant improvement of the run. There's just enough room to squeeze the plane into the bottom of the platform, which means we can grab both penguins on each side with only a single trip, cutting out two of the four chambers and saving nearly 20 seconds.

Level 14

Better manipulation of the boss gives a faster start to the first hit.
Extra special thanks for Archanfel for spotting improvements which allowed me to optimise an additional 533 frames across levels 08 and 10. I would like to request recognition for his efforts on the description page for the publication.

Memory: Judging
Memory: So entertainment seems pretty good and on the surface optimization looks pretty good too.
Unfortunately as it turns out this TAS in fact clears some optional levels that are not required to beat the game. This would be acceptable as a form of all levels if not for the fact that this does not beat all of the levels in the game. What this TAS does is beat all the levels that looked required. This means this submission is in an awkward middle ground where it is neither fastest completion nor full completion. The author intends to do the secret levels now, which I feel would warrant an entirely new submission as this could have a dramatic impact on the perceived entertainment value. I asked if they would like to cancel but have not received any response.
There is one other problem as well. The discovery that a large number of levels were in fact optional has delayed this submission's judgment until December 31st. This means that we have in fact FAILED to save Christmas in time. This to me seals this submission's fate.
Rejecting for failing to save Christmas.
Last Edited by adelikat on 10/29/2023 12:25 AM
Page History Latest diff List referrers