My personal approach to SPQ skill settings is to populate the level until I find it slightly challenging, then use that as the "Hard" setting. My reasoning is that I'm a fairly average SPQ player who plays other people's levels on "Normal", but my familiarity with the level gives me an advantage.

I have varying thoughts on monster removal for lower skill settings. One idea is that the most interesting or novel encounters should be preserved for all skills, since those are part of the experience of a particular map. You'd never have a skill setting where e1m7 didn't have Cthon, for example. Another idea is its only the really tough encounters that one should remove when lowering the skill setting, becuase the less difficult encounters ones are just as straightforward for everyone. Also, one thing I do often is try to place an easier monster in the same spot as the hard monster -- an ogre for a vore, for example. Finally, there are a lot of large-scale encounters where I simply scale back the numbers, while keeping the placement and types about the same.

For health, I think it might actually be a good idea to simply leave the health the same in all skills -- I figure that the less skilled you are, the more damage you'll take.

I'm not really sure how to tune ammo placement. Obviously, there should be enough to kill everything, but beyond that, how can we tell what percentage accuracy to expect from easy, medium, and hard players? 

Without player-selectable skill settings, what can you do to make the map feel "just about right" for people of different skill levels?

1. Allow multiple solutions to a problem. Especially when these solutions require different types of skills. The easiest example is secrets -- since the item in a secret usually makes an encounter easier, people who are worse at combat but better at exploring can get past a monster encounter that would otherwise be too hard. The player who is better at combat but not at finding secrets will get through the fight on skills alone, and will not feel that it was too easy. Another example is using stealth to get past a fight, or to get an advantage when the fight does start. Players who think more tactically can use that part of their brain, while players who don't, but are crack shots, can once again brute force through the encounter.

2. Allow multiple routes to the goal. This is similar to #1, but allows more player freedom. It can be seen in games like Mario 64, where the player doesn't actually have to play all the levels to get to the end. At any time, the player has a handful of levels to complete next, and each level will unlock a new one. The player never gets stuck on a single level. To apply this to a level design, try making a door that opens only after finding X buttons, but put X+2 buttons in the level. Each player might think that different hiding places are the "unfairly hard" ones, but it doesn't matter becuase those ones become optional. And by thinking that they had to find "X" buttons, they will not think that they've missed out on content when they stop searching after finding the necessary amount.

Another example of multiple routes is Mega Man 2 (or any Mega Man game, probably.) Players have to complete all stages, but in the order of their choosing. Each stage gives the player a new weapon, so he can play the levels he considers easier first, and then tackle the harder levels once he has more firepower. To apply this to level design, try making a level with multiple keys to retrieve, but no set order to retrieve them (like Quake's e1m5.)

3. I'm not sure what the general principle would be called for this, but there's a cool effect I noticed in Half-Life 2. When your health is up, and you're doing well, the natural instinct is to charge ahead into the next encounter. When your health is low, you start looking around for health packs. Basically they had health packs hidden all over the place, in houses, behind pipes, etc. etc. In fact, if you had to walk past all of them, you'd probably pass most of them up, and conclude that the game was way too easy. But since players doing well don't look as hard for health packs, they don't see how many they passed up. As a result the player gets the sense that they are playing a game that is just about the right challenge level for them. 

1. Leveling up. RPGs have a nice feature where if an obstacle is too difficult, the player can go kill orcs in the woods until he gains a level, and then go try again with a stronger character. This is a really elegant way to design around difficulty problems, becuase everybody can eventually beat the boss when they're level 50, but good players can beat him at level 30 and feel like they really accomplished something.

2. Unfamiliar control schemes. Any game that uses a well-established control scheme, like WASD+mlook, must be harder simply to satisfy genre veterans who have played all previous similar games. This means that the game will be unplayable for the newbie to the genre. The difference in ability between a top player (Thresh or whoever) and a total noob (your mom) is going to be huge. A game with a relatively novel control scheme will be equally unfamiliar and awkward to most people. The difference between the ability of a hardcore gamer and a casual gamer when playing this game will be smaller becuase they ALL have to learn the controls -- the hardcore gamers has less of an advantage.

3. Adaptive difficulty. This just means that if you get low on health, the bad guys shoot less accurately, or the breakable crates give health packs at a higher rate, or, have the other race cars magically teleport right behind you when you have a ridiculous lead, or some other sneaky way to make the game go easy on you when you're losing, and get harder when you're doing well. I think the big flaw with this idea is that it's dishonest, becuase if the player knows about it, he loses a certain amount of the satisfaction of playing. And if he doesn't know about it, he's being tricked.

3b. Pretend danger. This is like 3 in that it's dishonest, except instead of tricking the player by changing the rules, you just trick the player by making it seem like they were in more danger than they really are. Example: in Prince of Persia 4, there was a section where you had to jump from stalagtite to stalagtite in a sequence. Just as you jumped, the previous stalagtite would break and fall into the bottomless pit, making it seem that you had just barely made it. In fact, you could wait there for 5 minutes and it would still break right after you jumped. Scripted battle chaos in Medal of Honor type games also falls into this category.

4. Fixed-length games. One general problem I see with too-challenging games is that they are very frustrating to casual gamers. When you read a book or watch a movie, there's a reliable sense of how much time you are committing, and whether you will get to experience the entire thing, including the end. A game can en-trance and engross a player, and then halfway through simpy end up being too hard, thereby never giving the player the rest of the story or experience. Indeed, part of the experience of playing a game is the sense of accomplishment, which would not be present if you could see the end no matter what. However, I think it should be possible to design games which take a known amount of time to complete, and yet still provide a sense of "winning" and "losing" or at least "doing well" and "doing poorly." Most sports simulation games take the same amount of time whether you win or lose. And some interactive fiction is set up to last a set number of turns regardless of player performance. In the end, I think the types of games that can be designed to have a fixed length might be limited, but it's probably more than we think.

5. Hint systems. In any game that involves puzzle solving, a system that gradually gives better hints (for example after the player is stuck for a couple minutes) will help anyone that gets stumped on a puzzle for a while, while allowing the puzzle to feel challenging enough for people who solve the original puzzle quickly. This is sort of a fuzzy idea so i'm not sure of any good examples, but i suppose just having characters in adventure games you can ask for help is a decent example (any recent Zelda game.) 

Reaper bots had an option to be adaptable to player performance. I had a lot of fun putting four bots in a level with this option enabled and fragging them for 20 minutes. Maybe it was acceptable because there was no victory condition. I'm not sure. 

You've pretty much said everything there is to say yourself. Wow.

I don't like grinding RPGs, I don't think there is much satisfaction in constantly battling the same shitty monsters for hours just so you can do it all over again with some slightly harder monsters. I know this sounds like most games, and not just RPGs, but there is something I find that is rather tedious about many RPGs. It is nice when you reach a new level, and can tackle the next area, but sometimes the time commitment asked of the player is just stupid. Then they boast about how many hours the game takes to finish. Huh?

I hate the computer cheating on harder skills. It's ok if the harder skill levels make more baddies appear, makes them smarter, or, as in the forthcoming SiN: episodes, has the baddies wise up to commonly used tactics, but I can't stand the AI cheating in Mario Kart. There is also the powerup selection, which kinda pisses me off (mostly because of the blue shell in the recent versions). Oddly, I don't like the way Half-life makes the enemies physically tougher on hard difficulty. I think this is shit because the game is balanced for normal, and if the enemies take too many hits to kill, it make the game less satisfying. It's shit if you have to unload your whole clip into a guy's face to drop him, imho.

As long as the computer doesn't seem to be cheating, it think difficulty auto-adjust is ok. It's just when it's really obvious it is no fun. If the player is being lied to and the player is buying the lie, but having fun, so what?

I certainly don't mind when a game decreases the difficulty of some part of the game when I've been stuck there for ages. FPS games don't seem to do this often, and rely on the player having a previous save to fall back to. This is the wrong way to do things in this case. I would have LOVED it if the shitty ninja zombie marines in Doom 3 became temporarily less accurate after a couple of reloads from the same point. I don't want to have to replay a chunk of the game just because I now have 3% health and because of the high accuracy of the marines, it's very unlikely I will survive and encounter with them without getting some more.

Hint systems are ok, but only if they aren't forced on the player (like a dialogue popping up or something). What's to say the player isn't about to have a eureka moment, and then they have it spoiled by the game telling them what to do because some designer decided that if the player hasn't finished the puzzle after x minutes, they are dumb and need help.

Having a sidekick you can ask for help is cool though. I haven't played it in ages, but in Indy and the Fate of Atlantis, I'm sure Sophia was quite helpful sometimes, but only if you asked her.

Bleh, I can't be bothered to write any more. I think that different games require different methods of making things manageable for players of all skills. Some ways work, some ways don't. 

Thanks for this thread. That is what I'm working at the moment in my map.
I agree with everything you said in "Response to #1". This propably is the way most mappers go when implementing skill settings. One thought is that I want to leave some surprises on higher skills for people who replay the level starting form easier skill. Thus I don't always want to place an easier monster in the same spot as the hard monster (or easier battle in the same place as a harder batle).
I do it similar to you. First implementing hard skill, then removing/replacing monsters for lower skills. Also some architectural changes can be done to enhance skill settings like removing/adding a wall or lift.
I also came to the conclusion that health packs shouldn't be removed on any skill. Maybe they should be only replaced to the weaker ones (hp25 to hp15 etc.)
Ammo seems a bigger problem for me. In my map there will be 90 monsters on skill 2 and 50 on skill 0. So there will be plenty of unused ammo on easy skill, but I also tend to leave it this way. Easy skill is for fun I think. Lot's of ammo should enhance it.
One way of implementing skill settings I use, is to trigger monster on higher skills to wake them up earlier than in easy skill for example. This way I make it a bit tougher and can surprise players replaying the level on higher skill.
What do you people think about changing weapon/artifact position between skills? I have done it in my first map, but now I think it isn't the best solution. On the other hand I somehow remember that Id used to do it in the original levels. 

1: I don't really have an opinion on how difficultly modification should be done, but I do think that easy should be easy and hard should be hard, which is fatuous but I guess I mean that there should be really quite a lot of variation between the skills.

2/3: I think the idea of having a hard but 'proper' way and a slow and less satisfying way to get past a given point is a good idea. I remember Jesse's Runners Delight maps which were designed for speed running and so had masses of opportunites for tricking but were completable by more usual running as well -- a bridge slowly extending, so you could rj if you wanted, etc. 

I think it is ok to add extra ammo and health on hard, since it can help the gameplay if the played doesn't have too much stuff on easy and too little on hard.

With apsp1, I made small changes, the main difference between easy and hard being the size of ammo boxes and health/armour contained in pickups (and number/strength of monsters, obviously). I don't know if this was completely neccessary in all cases, but I didn't find the gameplay frustrating at all, and didn't get any negative feedback regarding difficulty, so I guess it was ok.

I also had a fair few people test it. Probably 10 testers or so. They didn't all get the map at the same time though, so I was able to test the couple of iterations that I made before release, each time with new testers, so comments regarding difficulty would be valid.

Actually, I think always using the same testers is a bad idea. It is fairly common in the games industry to just have the same people testing the same game for months, but I believe this leads to them overlooking problems a fresh tester might spot. I don't know if larger companies cycle their testers between projects, but I think that's I would do it were I in charge of EA or something :) 

and some random ideas, mainly concerning quake:

given the fact that everybody who's still playing quake has so many years of experience, new maps should have a higher, more challenging, difficulty than, for instance, the id maps. like id hard equals today's normal.
i don't know what i would change for easier skills, maybe some weaker/fewer monsters, less ammo/health on harder skills..
personally, i like challenging maps (in a healthy relation), so if i run out of ammo - which rarely happens - i try to continue like this, even if it means i have to axe a shambler. i reckon it's annoying for most players to run out of ammo completely or e.g. have only grenades left, but this can a nice twist as well, for it will force them to use a different (more thoughtful) tactic - the grenade thing happened to me a few times and i was pleasantly surprised how nicely it influenced the way i played the map. such things can also be used deliberately by the designers, though it has to be carefully tested, e.g. axing a shambler is ok if there is enough armor/health around, but it sucks in situations were regular combat was intended.

metlslime's 'more buttons than needed' idea is interesting, but i'm not sure how it could be realised appropriately. at any rate, the map has to be fairly large for such features (in most cases maps/paths are more or less self-explanatory anyway) and the extra buttons would have to be 'presented' in a way that makes sense (ok, likely not that much of problem). all of this has to be considered when planing the map's layout, and it would lengthen the creation process even more. of course, this would add greatly to a nonlinear gameplay and replay value - something that has been lost in the (fps) games of the last ~6 years.

this is also somehow related to the megaman thing metl mentioned. hub-based games/levels - quake can only be vaguely counted to them.
there is more to megaman than just the possibility of choosing another level if one is too hard, though: one can play them in any order, but there is always one weapon that works best against one level-boss, and there are other weapons that are utterly uneffective. this could be applied to other games, too. quake is probably not the best example, but i could imagine a hub where you get a different weapons in each level, like the rocket launcher in one, and it's possible to play the shambler map afterwards, but it will be damn hard. so the player might want to finish the supernailgun level first. or something like that... 

Infocom games (text adventures like zork and planetfall) had a hint system called "invisiclues" which had a series of clues about each puzzle. The clues were set up so that you could see each clue one at a time, and the clues got progressively more blatant. It was entirely up to the player's discretion, but it was designed to make it possible to get just a little hint if you needed one, but a bigger hint if you needed that.

More info and examples:
http://www.csd.uwo.ca/Infocom/Invisiclues/zork1/

The strength of this idea is that if players want/need to cheat to get past a tough spot, they want to be able to cheat in a way that has the smallest impact on the rest of the game. 

i could imagine a hub where you get a different weapons in each level, like the rocket launcher in one, and it's possible to play the shambler map afterwards, but it will be damn hard. so the player might want to finish the supernailgun level first. or something like that...

Sounds like a cool sm theme 

...I'm in the process of populating qte2m2 at the moment, so I'll have a go at "1".

During the construction of my levels I tend to only include enemy if I'm wanting to test a specific set piece or pieces. If the set ups are successful then these enemy typically stay in the level till alpha. qte2m2 had 10 enemy only for a long time. Once the brushwork is finished I'll then start populating the level from the "start" so to speak. I don't specifically keep skill 2 in mind when populating, rather I ask myself "What enemy or combination thereof (type and number) best turns this particular place into a hard place to survive?" Of course, the end product of using this method is generally that I end up with skill 2 as a starting point.

Obviously, progression is an issue so I don't tend to stick three shamblers in the start room :)

I'll work "forward" into the level in stages, adding ammo and health as I test each stage. I find working in stages that seem to fall naturally from the level design means that the player will end up having some moments during the level to catch their breath.

I've only implemented one case so far where I want the player to HAVE to use the axe to kill an enemy :)

So once scraggy sends through a few demos and a test report I can then go back and adjust the level to be skill 2, just so. Then I use a combination of subtraction and substitution of enemy to get skill 0 & 1. Also, if a shambler is the best enemy for a particular set piece then I might keep it so but add some health (or maybe armour, or both) for lower skill. I tend to not adjust ammo downwards, but rather leave it the same for all skill. The rationale being that the monsters might be easier to kill but the easy player might waste more through lack of skill at aiming etc.

Now I have a beta, and the real testing and tuning can begin.