It all began around last May in the XC=BASIC discussion group on Facebook, someone posted their XCB rendition of a maze demo which looked something like this:

In the source code, a book was mentioned: Compute's Programming the Commodore 64 - the Definitive Guide by Raeto Collin West.
I looked it up on the web, and I found it on archive.org! That moment I realized that archive.org doesn't archive web pages only, but basically everything Man has ever produced 😄

I found the original listing of the maze demo, which mentioned another Compute book: Compute's First Book of Commodore 64 Games. That mention threw me into a deep rabbit hole, and I ended up hoarding a half-quintal of scanned book PDFs, not only about the C64 but also the VIC-20 and the Texas Instruments TI99-4/A (which was my first computer ever, I was about 9 years old when my father introduced it to me).

So, what I wanted was trying to convert one of the games that were in all those old books to XCB, but I avoided the obvious choice of picking one of the C64 games, so I went to the TI.

One game immediately caught my attention: Superchase. It's the first game shown under the "Chase Games" category. It's also the first game shown on the back cover, and I was fascinated by the maze generation: it doesn't produce the typical "simply-connected" or "perfect" mazes.

The mazes are basically generated by placing random 5-space segments both across and down the screen. The purple diamonds (the "treasures") are placed so randomly that they can end up in places which are non-reachable. But still... fascinating.

So, a TI translation from some original. I had to know where this came from. So I dug even more into archive.org and I fished out the original publication on the October 1982 issue of the Compute! magazine. There I discovered that there was also an Atari version! Mamma mia...

Of course I wanted to see how they looked like. I had no easy way to see in action or play the VIC-20 version, couldn't find any video of it, so I had to dig out some PRG file from somewhere on the Net and making it run on the VICE emulator.

So ok, PETSCII chars, plain beeps varying in tone. You're the ball, the Monster is the asterisk.

As for the Atari version, it was easier to see it in action: a kind soul posted a video of it on YouTube!

Do you recognize the graphics? Ok... 😁

I ended up studying this video a lot. Because I was sooo lazy I couldn't be bothered to figure out which Atari computer this is, what's the best emulator for it, and so on and so forth...

Both those two versions are basically the same, the main addition of the Atari version is the nice graphics (and more engaging sounds! I just love early Atari sounds...). The game took the classic "perfect" maze generation algorithm (which is the one the "definitive guide" book refers to) and used it for a chasing game.

The main dirty trick is that the trail - that the Player leaves behind - is purely cosmetic. What the game actually does is recording the Player moves into a string variable, which is then consumed by the Monster from the beginning of it! It basically implements a queue with a string. 

The game doesn't even bother checking the Monster moves so that it doesn't erase walls, because it already did that with the Player! Is that cheating or what?? 😄

Now the part that fascinates me the most: Cheryl Regena - the person tasked with the TI translation - for reasons shrouded in mystery, chose to do it pretty differently. First, she came up with the method I described above to generate the maze. Then, she decided that the Monster would actually follow the trail. And that, of course, made the program more complicated.

In fact, so complicated that the program ended up being some intense example of spaghetti code, even for its time 😄 I couldn't, for the life of me, figure out how the program implemented the chasing mechanism, and in particular, how it made it so that the Monster would not get stuck in dead ends.

I then decided that the only way to understand the code was to do a quasi-literal translation of the TI code to XCB3 code, down to every single GOTO and the silly variable names.

While I was doing the translation I realized that she forgot to actually increment the level number when it was time to generate a new level, so the difficulty never increased.

But especially, after tons of test-runs including a few with the original code pasted in an online TI99-4/A emulator, I established that the game was ridden with bugs! You can make it quite easily so that the Monster actually gets stuck in a dead end!

(My C64 translation of the TI BASIC code. When I took the screenshot, the asterisk was moving back and forth, i.e. stuck)

I imagine this was a hasty translation, combined with the fact that TI BASIC is extremely slow, and so there wasn't much time for proper testing...

But still... I found this version quite interesting. The game is different in one more aspect: it plays like Superhot, the PC game where time moves only when you move, so you get a more strategic feeling from it. I considered adding that aspect in my version, in specific levels, but then changed my mind.

Then, there's the treasure aspect: you don't have to collect all the diamonds, while in the original version you have to, in order to get to the next level. This came out from a limitation of the TI program, but anyway, interesting.

So. I decided that I would rewrite this thing from scratch, and in the process taking the best from each version. 

First thing: the graphics. I went to the Atari listing and took the DATA numbers that represented the redefined characters, and I imported those into C64 Studio. I found the graphics I expected, plus...

... a door??!? Something not used in the Atari version!

That gave me an idea: I would use that door and I would make it open when enough treasures are taken (like the TI version did), and so you can escape!

Another thing that the Atari game has but doesn't use is colour for the treasures (they are all orange in the actual game). There are other DATA numbers for the colours, so I looked up what number correspond to what colour in the "8-bit Atari world" and tried to match them to my version as close as possible.

I wasn't able to take anything from the VIC-20 version. What had inspired me was the "crescendo" in the Monster beeping as it gets closer to the Player, so I wanted to raise the Monster sound volume as it approaches, but found the SID thing to do that too hard for me (this was my first experience ever with programmed sound!) and so, sadly, I had to give up on that.

From the TI version I took the maze generation, the real trail-following thing, and the basic idea for the Monster movement, like marking the map invisibly so that the Monster can avoid going into dead ends again and again. All the rest of the Monster "AI" I had to come up with, all by myself, and was pretty tough and rough.  At one point I was studying the Trémaux's algorithm, this video and this video helped me in the process. But then I threw it all away (I was really becoming crazy on it) and came up with something simpler...

From the Atari version I took the graphics, as I said, and inspiration for the Monster sound (I wanted so much to copy that sound but I wasn't able to).

All the rest - the "musical sounds", the Pengo-like coloured frame when you exit the maze, the Pico-8 inspired screen shake - were my idea.
Unbelievably - since I suck terribly with drawings in general - I was able to draw the game font all by myself! I know it's similar to the Sinclair ZX Spectrum font, but I swear, I drew it myself.

So there!