Would be nice as a shared Google docs spreadsheet, much like the "Digital DOS games Retro PC Compatibility Sheet". In fact, it would be nice if all the shared reference material like that was available as a shared doc.

@vvbee: really nice work! It would be a shame if you have to take this offline. This is way easier to navigate and search through than the original. I could see this being an exceptionally useful tool for gear shopping / browsing.

I always wondered why this matrix only had one radeon card. Does the result apply to all radeon agp and pci cards?
Only problem in this version is that things that are explained at the bottom of the original matrix are not explained in this version at all like for example if it says "crt only", it doesn't make much sense if you don't have the explanation.
Another thing is that I have a laptop that has Savage S3/IX chip and it doesn't do "every 5th line are doubled" thing in dos games. I went through commander keen games and I was not able to find any double lines. It would be much higher in the list if it doesn't do that.

Glanceability on the go is one of the use cases I had in mind, although I'm not sure how good the load times are. Spreadsheets are good if you want access to the underlying raw data, but they're more cognitively demanding, so having a lighter interface is useful.

Some of the footnotes are included and some are not. I think in general the descriptions could benefit from structuring, like I've already separated fps information into its own thing. But managing the contents of the database isn't my concern here so much, just how it's laid out.

The project can be adopted from https://github.com/leikareipa/dataredesign1, just look at the sample json and add your data accordingly. I'll put in some basic documentation later.

It looks really nice, but I'm going to be totally honest and say that I don't think this is the direction we should be going. Compatibility isn't a matter of percentage, especially with the relatively small selection of games used for these tests originally. If someone never intends to play a certain game or series of games which are pushing a particular card to the top of the "most compatible" list, treating the list as an overall "score" may be a bit misleading. The green and red shading on the original chart also gives certain cards a very good or very bad reputation right from the start, whether or not a certain big red box actually matters to anyone or not. I mean, really, who here has actually even thought of playing Mario Shareware on DOS? And games like Boulderoid and Duff (which isn't even listed on Mobygames)... are such small obscure games that using them as a basis for overall compatibility seems odd.

Then there's the fact that half the chart is dedicated to the same few games at different resolutions. If I have no intention of playing Quake, Duke3D or Prehistorik 2, then the overall "score" of compatibility is almost completely irrelevant. Its also worth noting that the results on the chart aren't always indicative of the results with every card of that type. A defective card or system incompatibility could easily have caused at least a few of the bad results... there's just too many for it to be 100% accurate. With so many oddball games on the list, the software may actually be at fault and only affect specific hardware configurations.

I appreciate all the time that goes into making charts like these (I've done a tiny bit of benchmarking and testing myself and its a ton of work), but I think we need to be careful not to turn any of these sources of information into "userbenchmark.com" or "geekbench.com" for retro computers. It can have a profound impact on the availability and value of hardware. Absolute worst case scenario: people start scrapping all the cards that are at the bottom of the list, even if they'll never play any of the games that put them there. 😵

Perhaps it should be that kind of web page where user can put check marks on games that are important to the user and it calculates compatibility based on that?
I would put my check marks to doom, quake, duke nukem 3d, commander keens, epic pinball and jazz jackrabbit.

You'll never be able to sample the whole domain, so of course you need to put up with approximations. I'd say gona's selection of tests is enough to separate the grossly incompatible from the very compatible. There's several variables at play that muddle the waters, but you can either squint or be frozen in a shrug, until more data is available.

I did plan a feature in the search box where you can put in the name of a game and it'll show only the entries where that game ran successfully. This still doesn't mean it'll run on your exact board/bios/whatever combination, but you get the gist. Also, per baoran's suggestion, a feature where you influence the selection of tests from which the score is calculated could be thing.

Know I'm replying to a topic from last year, but might want to include the SVGA res for Pinball Illusions. Reason I suggest that, is to test how cards handle non-VESA SVGA modes. I can attest that from my personal usage (and getting tests from friends with other ones in this group), that none of the AGP Voodoo cards are compatible with that game in SVGA.