"Weird ROMs" launching

The #41 #42 at offset #0000 can be a header or just data (41 42 43 44...). The MSX standard does not say anything about that.
Some games, like Corageous perseus or MSX Shogi has a "dummy" header at #0000 and the real one at #4000.
Those 32K ROM could be located at page 0 or page 1 (it could be deduced from the second header). But the fact is that in the ROM file itself there is no enough information to be 100% sure how the real cartridge works.

So, this is why I said:

Courageous Perseus and MSX Shogi Game also work fine. The "dummy" header is "41 42 00 00 00 00 00 00 00 00 00 00...". Until proven otherwise ROM2MMM can execute all existing Roms.

My method for 32 KB Roms is :

• If a Basic program is present the rom is inevitably placed on banks 1 and 2 with mirror of bank 1 on bank 3 and mirror of bank 2 on bank 0.
• If the adresse from first header is between 0001h~3FFFh the rom is inevitably placed between on banks 0 and 1 with mirrors on Banks 2 and 3.
• If the adresse from first header is between 4000h~7FFFh the rom is inevitably placed between on banks 1 and 2 with mirror of bank 1 on bank 3 and mirror of bank 2 on bank 0.
• If the adresse from first header is between 8000h~bFFFh the rom is inevitably placed on banks 2 and 3 with mirror of bank 0 on bank 1. (This configuration exist?)

You can see the source code of ROM2MMM to see if I have forgotten nothing. I cannot easily explain in English.

You say my method is insufficient to execute all games. I take this example because I used this method for ROM2MMM and all roms work without exception.

There is the distinction between theoretically correct detection and practically functional detection; in theory you do not have sufficient information to make the perfect decision (which is what Guillian is saying), in practice there are methods which work well enough for all the existing published software (which is what gdx is saying).

The latter will be sufficient in practice, but it can be broken if desired and done intentionally, or in a super-unlucky case of ROM contents. At least the detection situation is much better than mapper type.

So, you're both right .

Can someone explain me the problem with this ROM? I tried it with ROM loader I made for SVI-328 and it worked out of the box. AFAIK my code has no anything special related to loading of the ROM, so I'm really puzzled why other loaders have problem with it? (In my loader I always expect that /CE of the ROM is connected to /CS12)