MiSTer FPGA Forum

This might be a bit of a silly subject, but i've noticed that the attract mode battles (cpu vs. cpu) are much more random on original hardware. I was busy with getting the image right for the 16-bit console cores (tv + mister settings) and was using a few select games to do so. After letting the attract mode for this game run for while, i decided to check the image of a few other cores and then went back to the Genesis core again with this game. When the attract mode started again (after using other cores), i noticed the battles being same to what i saw earlier, or almost completely the same. On original hardware, the battles seem to be much more random, even when the same opponents face each other. So i did a short test on original hardware by letting the attract mode run a battle, turn the console off and let it run another Ryu battle and so on. It really seems to be much more random.

What i'm trying to say is that while on the MiSTer core the game sounds, looks and plays fine, underneath there seems to be something different, or missing. I don't know. I'm very happy with my MiSTer console, but i was curious if anyone noticed the same or something similar.

Genesis doesn't have a realtime clock, right? What would a RNG use for a seed then, especially when there's no player input?

Lots of things... Programmers tend to find hidden tricks to make random numbers.

Reading address locations that return open bus or noise, or RAM that is initialized at power up with semi-random values. Anything...

You not the first to report this on cores.
I have seen similar observations on Neogeo core to PCB comparisons where the vloger noted the attractmode changes on powercycles and they needed to reboot the PCB a few times for it to pick the right one.
Also during the CPS2 development by jotego on one of the shootem ups someone noted on a real PCB the bullet patterns were more random and on the mister core they were always the same. (this may have been addressed)

I guess what I'm saying is I have seen these things mentioned before but I'm not sure what the fix would be.

It's a bit like the cpu vs. cpu option in Fatal Fury on Genesis, seemingly being very random with who wins and how, though i haven't played that version of the game for many years now and don't own an original copy. With SFII' CE, the character being portrayed at that moment usually gets a win, though not in the exact same way, even when facing the same opponent again.

I'm not sure this is the actual cause, but it looks to me like the game uses the H counter from the VDP. As part of the overall game initialization, it reads that register and then advances some RNG-like state that many times. Although the H counter is obviously not a random process per se, I guess the value obtained at the specific time that subroutine runs might be random-ish on original hardware due to analog stuff affecting reset timing.