Top 3 worst ideas in Amiga history?

[MskoDestny]

3. Stayed with 68K.

Really? 68K was reasonably competitive to x86 during the time Commodore was around. Apple didn't move to PowerPC until 94. There were high end workstations on RISC architectures at the time, but those were largely competing for a different market than Commodore was in. The cost of an architecture switch would have been high and Commodore had made too many mistakes already to survive long enough to gain any benefit IMHO.

My top 3:
1. Not investing in the necessary R&D to stay ahead. When the A1000 was released in 1985, it was superior to the IBM compatibles and Macs that were available at the time. The A2000 had an edge over the Mac II in some respects (namely hardware acceleration), but had a much slower CPU. Later on, AGA was both too late and too limited to compete as others have already pointed out.

2. Trying to compete directly with the game consoles. A blitter makes sense for a personal computer, but the tile-based hardware common in game consoles of the time generally produced better results at lower cost. Combine this with the pricing advantages that charging developers licensing fees brings and this clearly wasn't going to work out. Further, the attempts solidified the perception that the Amiga was merely a gaming machine.

3. Not doing more to pursue "professional" markets. Apple survived because of their dominance in the desktop publishing market. The Amiga did well in video production, but that alone wasn't enough (I imagine desktop publishing was a much larger market at the time, lots of companies had internal art departments for print. I can't imagine too many did video production internally).

[MskoDestny]

The Amiga did PRIMARILY well in the game/homecomputer market ("we bought it to help with your homework")

And that is fundamentally the problem. It didn't make much sense to buy an Amiga strictly as a games machine. Too expensive compared to consoles and once the Genesis/Megadrive and SNES got on the scene, it was graphically outclassed. Now if there's something else useful you want to do with the computer besides playing games, you can justify the extra cost. If you're looking to do useful work you're likely to want the same platform as you use at your job. This gave computers that had success in professional markets a leg up in the home market even if they weren't as good for playing games.

Further, consumers tend to be stingy and buy low margin products. Professional users tend to buy higher-end stuff and will tolerate higher margins.

And personally I think there's still potential in the game/homecomputer hybrid principle.

To an extent I agree with you. Lots of people play games on their computers, not just hardcore PC gamers. However, for most of those people, gaming is a secondary function. They don't buy the computer to play games, but once they have it they end up playing games on it.

[MskoDestny]

I've got both a Genesis and an Amiga, and I must say the games that get the most out of the Amiga hardware (like Elfmania, Fighting Spirit and Kid Chaos) deliver better gfx,

Kid Chaos is impressive given the limitations of the hardware, but I don't think it can match the better looking platformers on the Genesis like Sonic 3, Sonic & Knuckles or Vectorman. Elfmania is more colorful than your average Genesis game and the animation is quite smooth. They did a good job of pulling off some parallax with only a single playfield. So I'll give you that one. Fightin' Spirit doesn't seem particularly impressive though.

Really the only things the OCS video hardware had going for it for 2D games over the Genesis were color selection and the amount of RAM directly accessible by the video hardware. Any pixel can be any of the 32 colors in the palette and palette entries are 12-bit whereas on the Genesis a given tile/sprite is limited to using a single palette of 15 colors and the palette entries are only 9-bit. However, there are 4 such palettes and you choose which palette to use on a per-tile and per-sprite basis (so 61 colors onscreen unless you resort to "tricks").

Now if you want two independent playfields (for parallax scrolling for instance), the color selection advantage of OCS is greatly reduced since each playfield is limited to a 7 color palette (+1 background color). The Genesis hardware doesn't have this problem. Also, the OCS sprite hardware is pretty lame compared to the Genesis. OCS supports 8 3-color  or 4 15-color 16-pixel wide sprites per line. The Genesis supports up to 20 sprites or 320 pixels worth of sprites per line whichever comes first in sizes of 8,16 or 32 pixels wide with a limit of 80 sprites per frame. Genesis sprites always use a 15 color palette. You can work around these limitations to an extent using the blitter, but on OCS/ECS you'll probably run into bandwidth limits before you get into the neighborhood of what the Genesis hardware can do.

Fighting games are probably one area where OCS fairs pretty well as parallax isn't as important and you don't need many sprites. Further, all the RAM available to the video hardware allows you to do smooth animations on relatively large sprites.

Now the OCS hardware was a lot more flexible than the Genesis hardware making it much more appropriate for a general purpose computer. Making a GUI system work on the Genesis hardware would require some sacrifices and even then would probably perform poorly in comparison.

BUT many Genesis games are better looking because they were well developed and were put on a cartridge (which saves A LOT of processing power and memory).

Well it definitely does wonders for load times. I doubt it made much of a difference for per-frame processing though. The A500 had a reasonable amount of RAM. Enough that you can load all your data for a single level up front and then just work out of RAM. As for using less memory, that's definitely true, but the Genesis had a lot less of that (64K for the 68K, 64K for the VDP and 8K for the Z80) so it's a bit moot for comparing the two.