Super Mario 3d
Abra Pabst
Really cool demake, but still needs work...
the physics and controls feel really off, as Mario falls at a steady pace immediately after releasing the jump button, and his jump is much higher (for example, getting to the top of 1-2 is much harder because I keep on hitting the bricks.)... I'm curious as to what's taking up all the tokens, as it's clearly not levels, and the code could probably be shortened even more...
In principle the controls in the original are probably delayed. But I think the controls in the original are a lot more diffucult , especially without gamepad... At least my girlfriend found this easier.
Nice work! I love how you used the pallete. Also I am not feeling the jump is a big issue.
There is a concern about Nintendo bringing this down. They are known to take out assets using their IP so be careful on publising it in the open.
What is the map data being used for? I see that you define maps in code, but the top half of map editor is still filled with data. Are those just maps you were building prior to taking the code approach?
Title screen is using map data. (the first 16x16 blocks).
More ist not used anymore... I created a custom level editor, which exports strings. But if I had known, that there is only enough memory for a few levels, I had not taken the trouble...
But this 8k token limit makes no sense at all. You can make a limit, but at least 16k+16kb user memory whould make sense.
Currently it is simple impossible to create games like mario, turrican or other 8bit titles. Which is sad..
Really great remake. The backward scrolling is a welcome addition to the original game as well. Other additions such as starting the game as Super Mario and the lenient fire-flower-counts-as-a-hurt-buffer are also fun tweaks, for starters.
It would be pretty easy to use that 4k of MAP that isn't used to store levels, but unless you could squeeze all the levels down to 4k may not be worth it since then you'd need different routines to load from string vs map.
I read the code. It sure looks like the first segment of letters are stored in hexadecimal... I do see where after that you switch to a different encoding. But there is quite a bit of data stored in ASCII HEX from where I sit.
As a note, it sounds like you're using a very inefficient representation of the levels in your strings. Are you storing literal tile layouts, or using a much more efficient 'sparse' layout like the original Super Mario Bros game did?
SMB1 defined the global sky/ground/background for the entire level in two bytes, then describes individual 'screens' (16-tile sections of the map, key look that matches PICO8) in two bytes per overlay, with a single overlay able to place multiple objects.
So really I think the main issue is your level format is overly huge because it's storing raw tile layouts instead of using such a 'sparse' structure. So yes you'd need to do more work to update your editor to do such a format, but that kind of 'indirect' map layout is really the key to fitting stuff into tiny systems like the PICO8 limitations.
Old games almost never stored the entire map in 'ready to render' format back in the day; they would commonly store a 'palette' of objects, and the map would define where those objects went and in what order, sorted so that the game only had to examine a few dozen bytes to update the screen.
This is why so many early games didn't let you scroll back: They didn't have structures that could be 'scrubbed' both directions cheaply, only in one direction, and the CPUs at the time couldn't brute-force-scan the entire level data to scroll backwards.
And storing raw data into the map/tile regions and then copying that out using peek() in your _init can be a LOT more efficient than packing data into strings. It avoids the cost of the byte->whatever encoding (which at a minimum will expand the data by 20% for ASCII85, or as much as 50% for hexadecimal) and then the cost of that in the 'compressed string' size limit.
Edit:
I am still hesitating because I have no idea how to deal with the token limit. using the map memory istead of strings, will propably bring some free tokens. But there are still some missing enemies/gameplay elements, that are necessary in the later levels
...format, which saves tokes. There's a lot of token optimizations that can be done, not one silver bullet, but this first step gets you well back under the compressed-string limit which was what you'd ran into first by far. :)
Sadly a lot of game design for the PICO8 can be "design a basic idea... then redesign it to actually fit the constraints of the system" at times as you learn more and more about structures and the like to use.
Hey, I played a bit, and even if it doesn't feel exactly a 1:1 reproduction, it feels pretty solid. But I have one nitpick with the game : the camera feels a bit off. When running to the right, the camera leaves Mario around 75% on the right side of the screen, leaving not enough screen space to see what's coming on you.
Since it's been a long time since this was originally posted, can you send the level editor you used to create the levels for this game? It would be good since this game has really good physics and could be a great base for some PICO-8 Mario games!
You must import gym_super_mario_bros before trying to make an environment.This is because gym environments are registered at runtime. By default,gym_super_mario_bros environments use the full NES action space of 256discrete actions. To contstrain this, gym_super_mario_bros.actions providesthree actions lists (RIGHT_ONLY, SIMPLE_MOVEMENT, and COMPLEX_MOVEMENT)for the nes_py.wrappers.JoypadSpace wrapper. Seegym_super_mario_bros/actions.py for abreakdown of the legal actions in each of these three lists.
These environments allow 3 attempts (lives) to make it through the 32 stagesin the game. The environments only send reward-able game-play frames toagents; No cut-scenes, loading screens, etc. are sent from the NES emulatorto an agent nor can an agent perform actions during these instances. If acut-scene is not able to be skipped by hacking the NES's RAM, the environmentwill lock the Python process until the emulator is ready for the next action.
The random stage selection environment randomly selects a stage and allows asingle attempt to clear it. Upon a death and subsequent call to reset theenvironment randomly selects a new stage. This is only available for thestandard Super Mario Bros. game, not Lost Levels (at the moment). To usethese environments, append RandomStages to the SuperMarioBros id. Forexample, to use the standard ROM with random stage selection useSuperMarioBrosRandomStages-v0. To seed the random stage selection use theseed method of the env, i.e., env.seed(222), before any calls to reset.Alternatively pass the seed keyword argument to the reset method directlylike reset(seed=222).
In addition to randomly selecting any of the 32 original stages, a subset ofuser-defined stages can be specified to limit the random choice of stages to aspecific subset. For example, the stage selector could be limited to onlysample castle stages, water levels, underground, and more.
The reward function assumes the objective of the game is to move as far rightas possible (increase the agent's x value), as fast as possible, withoutdying. To model this game, three separate variables compose the reward:
59fb9ae87f