I just wanted to elaborate on some points made in the above article. For one, I really like Mark's advice about how to solve some problems in adventure games--such as making many puzzles optional to increase replay value. I want to add an idea to the mix--not only optional puzzles, but an achievement system. Nancy Drew already does this to some extent; you not only "win" the game, but can receive bonus awards for doing certain things well, doing them more than necessary, or being particularly thorough. It's a good idea, but unfortunately players don't learn about these achievements until the game is over; it'd be more fun if they could learn about at least some of them in-game, to give them the incentive to do that puzzle once more faster, etc.

Here's an example of what I mean. Let's take the typical sliding block puzzle (ones where you must slide blocks around to make a picture). Let's say the game randomized the puzzle each time. Okay, so the first time you do it, you may be just getting the hang of it; if you manage to solve it after 15 minutes, you might get to advance to the next plot point, but not receive the achievement you'd get it if you solved it under 5 minutes. Or maybe you could get an extra special achievement if you solved it in the minimum number of moves.

What needs to happen is more unlockable content to make such achievements worth it. It's a bit tricky in the context of ND, but these could consist of special items, decorative changes to the interface, etc. They really don't have to affect the gameplay directly to be effective.

I recently played the game (Luckless John) that was very heavy handed, though. If you didn't solve a puzzle quickly, the narrator would jump in, call you stupid, and solve it for you. I didn't like that at all. What would have been nicer is an option to skip the puzzle, but a tangible reward for actually solving it. Many modern games let you "buy hints," for instance, though it's a bit of a scam since some of the puzzles are so poorly designed you need the hints just to figure out what to do next.

Also, one thing I'd add to the interface discussion is context-specific menus. For instance, in many modern games, the interface changes if you're zoomed in on a puzzle or the like. This is particularly handy with the Myst-style games, though I like what he said about how those puzzles were challenging (since you had to make inferences and could never rely on pure trial and error).