- SNES9x 1.51 v6 svn113
- Abuses glitches
- Abuses physics
Join Mario, Yoshi and all of his friends in the new racing game destined to sweep the world! Super Mario Kart! Now for your SNES!
Super Mario Kart was released in North America September 1, 1992. Along with F-zero used Mode 7 graphics to give the appearance of 3D viewing, a marvel of it's time. It was also the first of it's kind to allow two players to play simultaneously and, it has Mario!
There are multiple modes in this game. Grand Prix, where you race all 8 characters in 4 different cups. There's Battle, which pits you head to head against another player. Then, last but not least, there is Time Trial.
I chose for this run, a Time Trial attempt of Mario Circuit 1. Anyone that has played this game for any amount of time knows how much of a tease Mario Circuit 1 really is. You could spend hours, days, maybe even months trying to top your score and barely miss the 1'00"00 mark. My current record on console for this course is 1'00"29...come on!
This current run is after the result of multiple attempts, six to be exact. At first, I only wanted to prove that this course could be done in under 1'00"00, then I found myself approaching the 58 second mark and kept working at it. As of the last run I fell below that mark, making it by a single frame to 0'57"98.
This run would not be possible on console, there's a few things at work here. Firstly, the most obvious trick when watching this is the second turn. There is a glitch in this game, when bouncing off a pipe sometimes you can be slingshot forward. Furthermore, you can only make use of this glitch to it's extent if you release and continue acceleration at just the right times.
Second, there is some physics abuse. I don't want to call it a glitch, but it is possible for most of the turns, upon landing, to drive in gravel for a single frame. This gains far more than it loses, in most cases. The only exception was the third turn, where I found it possible on the right frame to go the distance without treading gravel, which saved a hair more. It's a very sharp, nearly impossible turn for the shortcut that allows this.
The third thing at work here is abuse of the way the game measures how far you've traveled on the course. Frame position is not consistent, sometimes you'll travel 2 pixels in a frame, sometimes you'll vault ahead 6 or more pixels, and all of this changes depending on how many pixels toward the next frame you are. I tried to time it so that I could make the most of the inconsistency, for example timing a jump so that I would lunge ahead 6 pixels before jumping into the next turn. This often resulted in pixel savings turning into entire frame savings.
I will compare my previous run to this run, as that is what I used to help me obtain my current score.
Lap 1: 11"76 - 11"73 (2 frames)
Lap 2: 11"57 - 11"54 (2 frames)
Lap 3: 11"54 - 11"50 (2 frames)
Lap 4: 11"57 - 11"54 (2 frames)
Lap 5: 11"54 - 11"50 (2 frames)
Total: 57"98 - 57"81 (10 frames)
The frame rules in this game seem very complicated. The game can only calculate 60 times a second, while there are 100 increments of each second. The time rule is fairly simple. 2,2,1,2,2,1 etc. That is two frames in a cycle equal "02 and one frame per cycle equal "01.
The first thing I'm sure most of you will notice is that my lap times are inconsistent. This problem has made itself quite profound as of my last attempt at this run. I could feel it when I was trying to get everything to work, the frame rules just seem to make the course different depending on which lap you're on. I can get away with more on laps 3 and 5, than I can laps 2 and 4. In fact, the motivation for trying this run again came from this occurring. I suspect that "loss" comes from not being able to jump as far over the turns without going into gravel for 2 frames, where on laps 3 and 5 I can make the same jump and land in the gravel for only 1 frame. I did my best to correct this, and I am confident that I did so with as little loss as possible.
Old run: http://www.youtube.com/watch?v=AGAcjnt05Eg (0'57"98)
New run: http://www.youtube.com/watch?v=8wWiucH4dHs (0'57"81)
I know this run does not follow the usual standard for making it to the end credits, but I felt that it was still relevant given the circumstances. I tried experimenting with left+right and up+down but did not find anything meaningful.
_EDIT:_ Oops. Well, there I go looking silly again. I thought I'd done all of my research. I optimized a run that isn't even fastest! Oh well.
Noxxa: Claiming for judgment.
Noxxa: Rejecting for being slower than the published run of this game, being slower than real-time records in the same category, using a deprecated emulator, and doing a category that has been deprecated and disallowed for about 7 years.