Writing an engine for a single game isn't that hard, especially when you use the most common form of code reuse: ctrl-c ctrl-v

Writing an engine to reuse internally (like any bit of software) is twice as hard. 

Writing an engine for third parties to use is exponentially more difficult. 

If something goes wrong with your engine, we'll you have the source so you can fix it. 

But also, you have to fix it. 

If you use a third party engine, especially a popular one, almost any problem you encounter will usually have been encountered by someone else and probably fixed or worked around. 

Going to take a slightly different tack here. I've had a couple informal discussions and it appears that executives increasingly consider custom engines/tech to be a liability. This is different from years past. It used to be that an engine was considered a very large but viable investment. In other words a lot of studios used licensed tech mostly because they couldn't afford to build out custom systems, but there was generally a sense that there was indeed value in doing so. In 2017 however, it's considered a poor use of funds to even build custom tech at a studio level, regardless of whether they can afford it or not. It's only worthwhile, in management's eyes, at a publisher level where the engine can be shared across a lot of teams. That's the energy behind Frostbite, for example. There are some studios which have long term investments in custom tech, although publishers are largely working to share those codebases across their member studios.

Otherwise, execs don't want to have custom tech on hand. It's considered a problem that can be farmed out to someone else, with minimal gain to doing it yourself. Having your own engine means a permanent staff of X people there, some of which can't be replaced at any cost because of how well they know that system. Managers don't like irreplaceable staff. They also don't like long term variable costs. Training/onboarding new hires is harder. Hell, hiring engine devs has itself become difficult.

Homebrew mobile tech is so rare as to be effectively nonexistent in the marketplace.

What that really leaves us with is a rather small group of studios who are building tech as a significant differentiating factor. Games in this category include No Man's Sky and Ashes of the Singularity. 

While there are arguments to be made both ways, I personally dislike the increased homogenization that is the inevitable result of centralization to a handful of technology packages. There are consequences from architectural decisions that impact final look and feel of games, more so than people often wish to admit.

On 6/23/2017 at 6:10 PM, Hodgman said:

Unreal is also interesting that they give out the source these days, but many parts are simply too complicated to customise in a way that would typically be required, or doing so would lock you out of upgrading.... 

e.g. I've never worked on two games that have used identical rendering pipelines, but doing these changes in Unreal, while technically possible, isn't really feasible. Instead you just use what you're given.

Unreal is cheap noe for hobbyists, but it's still just as expensive as ever for big games.

I think this is partly true, but not completely. I've worked with a custom version of the engine (although I don't touch the actual engine code a great deal, so this is largely secondhand) and my experience is that, while it's difficult to do these things, it's still much easier (and cheaper) than the alternative of maintaining a custom engine.

Specifically, it's challenging to make significant changes to the engine, but there's nothing that complicated about it. I've seen people change really significant bits of the rendering engine quite successfully. It's also possible to keep the engine more or less up-to-date (albeit with some latency) after doing this. Merging features in from the official branch can be tricky, but it's still a lot easier than, say, implementing those same features from scratch instead.

The easiest path is to use the engine as-is, of course, but adapting it for a "big" game is often much less expensive than not doing that.

General purpose engines where less of a good option before because they simply were not good enough. Unity and Unreal have evolved into great tools.

Custom engines will give you better results if you are willing to put in the time, effort and *experience* to use them. Part of the reason people might use a custom engine is because the devs that wrote it  and know how to use it are still working in the company.

If you're AAA and you don't mind spending another 100K$ to use a custom engine. And you allready have 70k$ invested in a custom engine from previous games, why not reuse it?

But to counter that point: Big name companies can negotiate better royalties deals with game engine companies than you will get as an indie, hobbyist, or even mid-size studio. Don't underestimate the power of a strong finance department...