Why there are few native Linux games compared to Windows or even Mac?

I'll give you my point of view as game developer.

Disclaimer first: I work as a coder, everything I say about publisher interaction is second-hand knowledge.

We have made one Linux game. It was the first one of our two "indie" titles (quotation marks, because both of them ended up being partially funded by a publisher, so they weren't really indie in the end), where we had promised a Linux build on Kickstarter, long before a publisher got involved.

The main reason why we did not do native Linux in our publisher-funded games is quite simple: Our publishers didn't pay us for it.

There are actually some publishers who are very keen on getting native Linux versions for their games, but we sadly have not released a game with any of them yet...

The publishers we released games with did not agree to the buget that we think is needed to do a Linux port of sufficient quality. If we would lower the price for doing a Linux port to the point where our publishers would agree to it, we would take on a lot of financial risk ourselves, so this is sadly not an option.

If everything worked as it is advertised by engine developers, making a Linux version would be quite cheap: Just click a few buttons and ship it. This is, sadly, not the case in real-life, as there are always platform specific bugs in game-engines. Our one Linux game was made with Unity, and we had quite a few Linux-only bugs that we forwarded to the Unity devs (we didn't have engine source code access), and had to wait for them to fix... For the engine we mainly use nowadays, Unreal, we have a rule-of-thumb: "Engine features that are used by Fortnite are usually well maintained." There is no native Linux version of Fortnite... (We did try Unreal's Vulkan RHI in Unreal 4.26 for Steam Deck support in one of our games. Let me put it this way: The game in question still uses Direct3D on Steam Deck.)

So, from experience we expect that the chance that we would have to find and fix Linux-specific engine bugs is quite high. Therefore we have to budget for this, what makes offering a native Linux version relatively costly compared to the platform's market share. Costly enough to make our publishers say "no".

This, by the way, also answers the question why publishers are willing to pay for the way more expensive console ports. There are also way more console players, and therefore potential customers out there...

(I can only guess, but I would expect publishers to be even more reluctant to pay for native Linux, now that WINE works so well that getting a game running on Linux needs typically zero extra work.)

Market share and yes, Proton/WINE ultimately lessens the need for a native Linux port.

In a fair number of cases, even when there is a native Linux port, Proton/WINE has worked better than the native game.

If Linux gets to 5-10% of the market, we'll probably see them come back for platform specific optimization reasons. However, without a larger market share and with the translation being so good these days, there's not a lot of need.

Linux is such a tiny slice of the market compared to Windows, it doesn't make financial sense for dev studios to spend any of their budget in it, because they just won't sell enough copies to make it worth their while.

On Steam https://store.steampowered.com/hwsurvey/, all Linux operating systems combined have around ~2% users, compared to the MacOSX ~1.4%. This is only a recent trend, as for the longest run Mac had more Steam users than before. And building a native Mac game was more straight forward than on Linux.

Nowadays its completely different than before, thanks to Proton integrated into Steam. This means even though there is a higher percentage of Linux players on Steam, there is less reason to make native Linux games. That has some advantages: Windows binary through Proton has feature parity without the devs needing to understand the underlying Linux system and libraries, less work for the developers means higher probability of supporting Linux for longer time, thanks to Proton and the auto selection of Proton version for each game its also less worry for the end user. It does not matter what system libraries you have installed or what operating system you are using.

It would be nice to have, but in reality there is no real need for native Linux games from developers or for the end user / player.

market share. youre going to create a game, do you build it for the smallest market or the biggest one?

its nice that many devs keep linux in mind, but they gotta pay the bills

In just a few words to summarise a lot of these comments:

What user spends the most moneh?

As others have said, tiny market, but also that it often requires more development for the Linux port to get going, and even more development to actually make it run well. Like for instance, Civilization series usually release with Linux and Mac ports, but those are done by a third-party company which I imagine does add additional costs, and those suck regardless.

Not like it's a bad thing necessarily, the vast majority of native Linux ports I've tried were either severely out of date, had significant performance issues, crashed a lot or had some quirks that would make it not worth playing anyway. It's probably just easier if developers focused on proton compatibility instead.

It's more convenient to package things for wine than it is to support every flavour of linux. Being a cross-platform library for games is the one thing the windows API is useful for.

I was trying to keep my comment short(ish), but you're not wrong. There are other complications :)