Splinter Cell veteran says realistic modern lighting has screwed up stealth game

But you can punch your way out of various situations - it's "crouch near stuff until you're close enough to hit them" rather than pure stealth.

Splinter Cell was good here because you could make enemies visible and you not visible by virtue of the fact your character was wearing night vision goggles.

"him wot worked on" is just British English slang for "the guy who worked on"

To clarify the lighting point in the article: there are two basic parts to realistic lighting, direct illumination and global illumination.

Direct illumination displays only the "first hit" light that comes directly from light emmiting sources, e.g. the sun or a lamp.

Global illumination (indirect illumination) then adds bounce light that is reflected from the directly lit environment back onto other parts of the environment, e.g. from directly lit walls to other walls. This can also include multiple bounces (indirect to indirect).

When you don't have global illumination, only direct illumination, you get very deep shadows. Anything that isn't directly lit appears perfectly black. Like famously in Doom 3 in 2004. This is why you get so deep shadows in space: you only have the sun as main light source but almost no bounce light from somewhere else.

But even if a game doesn't have any global illumination (bounce lighting), developers in the past were able to avoid unrealistic, perfectly black shadows by always rendering the normal texture colors at a certain minimum brightness. Then the dark shadows merely lower this base brightness to a certain degree rather than appearing completely black.

This was great for a stealth game, because the developers had very clearly defined direct show casting (from emissive lights only) while being able to exactly specify how dark they are compared to the rest of the environment which is directly lit.

But if you introduce realistic global illumination (usually through baked or real-time ray tracing), these tricks don't work anymore. Bounce lighting is very diffuse and it erases the clear distinction between shaded and unshaded areas. Everything lights everything.

I don't think there is a way around this: Either you have clearly defined dark shadows, or your lighting looks nice and realistic. You can't have both.

Possibly a solution is to make the game more stylized, so the missing bounce lighting doesn't stick out negatively. In the past stylization wasn't necessary (Splinter Cell wasn't particularly stylized), because the rest of the graphics weren't very realistic either, so the lack of global illumination wasn't as noticeable.

One of the biggest challenges with computer UIs is how you convey information to a user, and that goes especially for games where players often can't take much time to analyze complexity, and it gets boiled down into a light gem indicator. One of the things I really love about the Thief games is how intuitive they are because they're mostly relying on sight/sound senses, you can relate to how the guards/defenses are going to sense you because it's how you would detect someone. If you're noisy you know what will happen next, and the guards are extremely vocal in telling you their state.

Going beyond that simplicity to account for other factors you could technically improve the simulation with is where I'm not sure it makes it more fun. Ultimately a mission needs to be conquerable, how far can you go making it more challenging and leave space for the player to push through while remaining plausible. Silhouette, different areas of your body being lit/unlit, whether movement speed of a lit/partially lit person affects detection makes a difference, guards having long term memory and adapting to half of them - they all sound good but I'm not sure they'd actually be rewarding to players and the development studio that implemented them. How do you 'tell the story' of a guard that spots your shape, knows you're sneaking around acts accordingly to take you by surprise and ends your game.

Similar with armor systems in a lot of games, we can probably simulate a lot of coverage/protection and the impact on mobility, that characters ability to fight with various weaponry because of what they wear, injuries, and so on, but for most games it gets abstracted into categories or points. Even if computation challenges of physically simulating that were overcome for dozens of characters in a fight, how do you convey all the consequences to the player to suggest how they can change things.