@Michael: You’re actually describing Pokemon’s save system. Of course, they punish you by having you lose money when you lose. But you retain everything else. This works excellently, IMO, for the type of game it is.

@Zorba: Also, regarding your last idea, it could work, but realize that sometimes people just don’t have time to play something for an extended period of time. The “pause slot” could work here, perhaps.

I’ve had an idea for a while that I may never get to make, but the basic idea is that it’s a semi-random mission-like first-person-shooter. You say “I want to play” and you’re given a 30m-to-1h-long adventure, but without knowing the details. You can’t plan for ambushes because you’ll never know where the ambush is coming from, if there even is one. The goal is to beat the adventure, but even if you fail, you’re given something you can use to get stronger for the next attempt. (And you have some control over difficulty.)

And, of course, there is absolutely no way to save in the middle of an adventure. You start, you’re in it until the end.

I think it could be marvelously suspenseful.

While I agree that a save anywhere system has significant negative sides, I feel that poorly placed save points can be far more detrimental to gameplay than a simple save anywhere system. If a game developer is not willing to put in the effort of thoroughly testing and tweaking the save points, they simply shouldn’t bother and go for a save anywhere system. At least that way it’s up to the player how much gameplay is hurt.

I’d like to point out two games with their own approach:

Bioshock (and System Shock 1 & 2 before it) had a standard save anywhere system, but combined it with respawn points. The result was that although getting killed still sucked, it didn’t result in a game over scenario. That took away the pressure and made people forget about saving (it also made the game too easy for those players which didn’t need respawn points, but that’s another discussion and can be handled through extra difficulty levels).

Prince of Persia: Sands of Time used an auto-save system, but combined it with a rewind system. Failing a jump merely meant you wound back time and tried again. No problems. It made sure you couldn’t make stupid mistakes and focused the pressure on the actual difficult parts.

Nothing’s more frustrating than having finally survived a large fight, only to misjudge a rediculously easy jump and plummet to your death right before you hit the save point.

The real issue is not how to let players save, but why they want to. I say it’s about pressure: if you’re low on health and only have three bullets left and think: “Hmm… door. Most likely there’s nothing but a big pile of ammo behind it, but there MIGHT also be an enemy”, you WILL save.

Tricks like Bioshock and Prince of Persia used take away this pressure for the player. While they both have their downsides, they do provide a way out of difficult situations other than saving. They might make a game easier, but, especially in the case of Bioshock, they also make sure the player doesn’t concern himself too much with save games.

Players worried about saves (whether its because they’re constantly doing it or because they’re hating the fact that they can’t do it when they want to) signal a lack of suspension of disbelief. That’s always bad.

One final comment: a game should be fun. Every single bit of it should be fun, no matter how often you repeat it. If a player’s thinking: “I’d better save, I don’t want to play that last bit again” it should be because they want to know where the story or gameplay will take them. Not because they found it frustratingly difficult and genuinely don’t want to repeat that experience.

To summarize: the best way of handling saves is by making players forget to do it.

I definitely agree that making the game harder after a loss is a terrible idea. That just makes the player lose again. That does not achieve anything. Ratchet and Clank rigged things so that you kept all the experience and money you’d gathered, thus letting you slowly level up and get more powerful despite failure. Brutal Legend just let you go somewhere else and do something else. But Halo just tossed you back into the fray, and your personal skills improving was the only thing that made it easier.

‘Course, now I’m also playing Borderlands, which deducts some money from you but leaves all your experience alone and even lets you run back and continue the fight from whatever amount of HP the monster had when he killed you.

Single players, incidentally, aren’t predictable, but large groups of players are reasonably predictable in this sense, and this is just what players do. Enough of them, at least, to severely cut into your userbase.

Regarding RPGs, typically, death means loading a previous save.. thereby losing all progress made since then. Perhaps a good compromise would be a system wherein you’re sent back to the last save point (whether you had used it or not) and retain items, experience, gold, whatever that had been gained. In this way, you lose progress in the game but retain the progress of your characters, and a system would develop where it would get easier each time — not just because you’re more familiar with it but also because your characters are slowly getting stronger.

Come to think of it, your earlier “rts” was much like this (I really enjoyed messing with that game’s mechanics).

Many games do work this way, or close to it. MMOs particularly. Perhaps more should. Resets are never fun.

As with a lot of difficulty-effecting mechanics it comes down to frustration management. This is hard, especially since you can’t take what you’re users say at face value – remember the X-Com bug that reset the difficulty to easy? If there are easily exploitable mechanics you have to take that into account. Some players will still have to use everything that they can to make the game winnable regardless but at least this way there is a choice. As has been mentioned earlier having to repeat content is both frustrating AND boring, and the ability to save anywhere can minimize that more than any of the alternatives I’ve seen. without save-anywhere you’re stuck blaming the developer when caught for too long without a save point. While the blame may lie with them it doesn’t provide the option for a fix. Save anywhere does.

The pause slot is kind of interesting but I think most relatively casual players (and I include myself here) would prefer not to have the risk of having to repeat themselves in the event of a crappy loss. It does solve the ‘I need to quit now’ problem elegantly. Sure, intelligent placement of save points can minimize the boredom problem.

I’m also wondering about the consequences of failure – death in most games causes you to lose time (which can hurt and cause a loss on interest) or resources. Resources tend to be trivial (in which case who cares) or make the game harder if you lose them. Making the game harder for the one having trouble with it doesn’t seem like the right choice… Not exactly an argument in favor of save-anywhere but given that having to load to a save point often leads to the same precognition coupled with the boredom of (more) repeated material I’m not sure that’s better. A better loss mechanic would be a better plan.

Hope this was somewhat coherent… I’m pretty sure I disagree with the idea that the player’s actions regarding mechanics like this are so predictable – granted I haven’t seen any studies so I’m basing this mostly on my own play experience.

A game taking a long time to complete (say, a day).
Whatever happens in the game, you cannot reload.
The save feature works exactly like the “pause slot” Zorba mentioned: as a tool for taking breaks, not for cheating. There is no other kind of save.
If you die, you have to start over from the beginning. That means losing a day if you happen to lose your final battle. (That should not happen, if you manage to survive to the final battle, you should know the game well enough to be prepared for anything.)
Each time you start over, you play in a new randomly generated world. You learn the general strategies, but no detailed “walkthroughs”, as all the details will be different next time.
Randomization also keeps the game interesting even if you play your 1000th incarnation and still have not won. (Yeah, these take long to master. And that’s the matter of skill, more than luck, the best players can win like 80% of their games, and most of the remaining 20% losses are in the beginning of the game, so not much is lost.)

These conventions are used by (hardcore) roguelikes. They are not popular, but have their dedicated fans. If you want to try, many hardcore roguelike fans recommend “Dungeon Crawl Stone Soup” as the best of the kind and one of the best games in existence.

Of course getting lost is also, more often than not, a game design problem :)