This is cool, one common thing Ive found is that having two sorta levels a-la cp_well style never really works, however places you can jump to and highgrounds like rocks and stuff seem to work very well, like process, badlands, and snake. Its kind of hard for me to put into words, but you get what Im saying.

less random shit on maps because when sunshine first came out i got like 60fps on that map

no water!

The point about straight lines is interesting, but there are other defining reasons behind these maps not being popular,
namely FPS on coldfront (warmfront did make a return this season if I'm not mistaken)
5gorge having uninteresting control points (see Badlands Spire), also relatively close-together points,
foundry being so wide open (though I think foundry is a great looking map, all it would really need is a tighter second point with less space)

You make a great point about designing points and segments of a map according to the offense/defense, namely defense should have some advantage over offense with the layout;
great example is granary last, given there is only 1 (technically 2) entrances to the high ground, which can be punished via bottle neck design to help defense stand a chance at a disadvantage, and puts it on the players to use things like uber and numbers to mitigate these disadvantages of being on offense with teamwork and tactics rather than sheer force of map advantage.

A great example of a rough lastpoint would be ashville (5cp), where the control point is all the way at the bottom of the last area while most other areas are several levels above.

Overall, the most important design I feel is giving a balance of offensive positioning and defensive positioning on maps. The reason badlands spire is so popular is because to control the point, you have to play the point, and it's not exactly the easiest thing to climb (for most classes).
"holding" second at choke puts defense at a disadvantage because of the height advantage of offense, but the spire itself, and the battlements, offer great defense for contesting the spire,
same goes for gullywash second, easy to defend and contest points from battlements/high ground, hard to hold aggressively, but just as easy for offense to find their way through the choke points, allowing them to rely on tactics to overcome the defensive barrier that is typically highground

If we're wanting to look at what's bad with certain maps, we can improve them a little better, and gain some knowledge as to how to improve current map-designers' maps

EDIT: I may be wrong, but I'm pretty sure valve has videos talking about the design and effort put into maps, and why maps were designed a certain way

You have the 2.5 entrances per point rule which is usually adhered to

Also proper terrain balance for all the classes such as rocks and correct spacings for rocketjumps and such

Symmetrical, reduced clutter and clipping, and high skyboxes

Most other things, particularly attacking/defending balance, have already been discussed.

No stairs or stupid little ledges that don't do anything except make your rockets do 0 damage.

A spire on second

I defenitely think of the old maps that have fallen out of favor, Warmfront and Yukon are the most promising to revamp.

just want to point out but granary is the map that belongs to this straight line group yet it's playable? how does that one differ from other straight maps?
what I think makes map viable in comp is some variety of rollouts for both solly and demo to perform, with things like ramps to surf, big spaces to make pogos and whatnot. another important thing is how you act after that and how map allows different strategies, you can go agressively (process, gullywash) or defensively (granary, snake) etc. So the rollout shouldn't take much time and be consistent, that's what I'm thinking

Don't make everything smooth and simple; that's very good for chokes, points and lasts but places like granary lunchbox, gully upper and behind badlands last make for much more interesting fights/flanks. One thing I've never liked about process is how cut and dry sewer or IT are. Claustrophobic flanks, dropdowns, (reasonable) hiding spots and unusual geometry help to make fights more than just initial positioning and ammomod aim.

The developer commentary for TF2 is particularly interesting because it addresses the artistic aspects of map design in terms of drawing attention to viable paths of attack. However, they also talk about why spawn gates worked for pubs, so I wouldn't read into them as gospel for competitive map design. Overall, the developer commentary is legitimately super interesting for learning about what goes into designing a game.

For the OP, I'd say that having intelligent ammo/health placement is a key product of map design: since TF2 doesn't have anything like the instantaneous effects of Quad Damage (Kritzkrieg requires building charge), health packs play a pretty important role as one of the two sole kinds of pickups in TF2. It's good to have an incentive for attacking players to take a risk to push up but also provides a safety net for defending players; Process IT's medium kit is a good example of this. However, having kits/ammo packs that are super close together can become a particularly bad idea really fast; I think an early version of Vanguard had a medium kit and medium ammo on a ground floor and a level that was a quick rocket-jump above it also had a medium kit and ammo pack.

I'd like to point out that the straight line thing isn't a must of 6s maps, tho usually a map of that structure presents something like 4 doors/passages from mid to a large and spacious second and this prevent the defending team to set a proper hold.
What makes granary still playable (but not the favourite of the average 6s player) is that the 2 yards are divided (unlike fastlane, foundry or gorge), allowing teams to keep tabs on all the entrances and all the space around them and distinguishing a combo and a flank side.

EDIT:Anyway I haven't examined cp_granary because I wouldn't define it a perfect 6s map.

2 other good articles I found (they are old, but scimming over them there are a lot of points that remain relevant)

http://tf2maps.net/threads/article-competitive-maps.17141/
http://tf2maps.net/threads/article-competitive-maps-follow-up.17220/

Also, to reiterate a point that I made (in a lengthy post) in the other thread: One of the big factors that makes a map good is experience and familiarity. Knowing what to expect, what potential threats exist, and even just being comfortable with basic holds and pushes between points is critical to "liking" a map.

What we should really be trying to do is setting up 'map reviews' for new maps. Have the map maker sit in mumble with a team (or even a few teams) as they walk through a map and discuss holds and pushes. Players don't have the knowledge of map making to translate their feelings about flaws in maps to useful comments in the same way that most map makers don't have the extensive knowledge of all competitive classes to be able to see their map as each of those classes.

lots of stairs and traintracks

the more the better

In terms of control points, I believe area control is key to a good competitive map. You want teams to feel like they are able to hold an area or fall back to an area in the event part of the team gets picked off or they lose a player. I have granary in mind in this case (maybe since it's granary week) that when a scout and/or soldier dies, the rest of the team can fall back (lets say to 2nd yard) and hold back the enemy team long enough for their respawns to return, instead of losing not only middle point but possibly even 2nd. In mapping this can usually be done with one or two choke points, a wider area to push through, and a large open space for fights to occur. You want to set the pace for your map, do you want a map that gives advantage to the team that makes high risk/high reward plays? Do you want a map that rewards teams that play at a slower pace, making effective plays or coordinated pushes?

I also think variance in terrain is a key part of a map as well. I like Gullywash's 2nd point for example, its open area and obstacles gives soldiers a chance to gain a height advantage (something that is mostly missing from Granary 2nd yard) but also has flat areas and jump accessible obstacles in which scouts can attempt to gain height advantage. This is also seen in Badlands 2nd yard, with the slope from mid choke to grey bridge, with a chance for scouts to run up on pride/cliff and jump to spire, from house to flank the enemy pushing through choke. Bottom line is that you want to make a map that brings out the best in all the main classes, not just one or two.

less clusterfuck badlands and granary
more maps like snake and process.
give a few ramps for slides, don't make pushing from choke to mid impossible because of extreme height differences (logjam).

Creating maps that allow each core class to play its role effectively while not allowing these classes to be overpowered should be, if it's not already in mind, a good rule of thumb. For example, there's nothing wrong with placing props, albeit they make sense given the map theme/environment and are also in places that make at least a little sense, so that a demo could possibly utilize that prop for a sticky trap. On the flip side, you don't want this to be such reoccurring theme to the point where a trap could literally be anywhere and you spend too long checking for them. Props also help against sniper sight lines, as long as you don't pull a Valve and do whatever the hell they thought was a good idea to nerf snipers on Suijin (as opposed to nearly any other option considering any other option would have been better). And even with too few props, demos can still obviously trap walls in choke points (see Saw on Snakewater). Another example is having, say, hoodoo props or, again, sensible props that can be easily utilized by a scout to expand on their movement, like on Badlands mid. On the flip side, you don't want them to be all over the place and in such a way that makes them too difficult to combat by default.

On an aesthetic side, I wouldn't mind seeing more maps themed differently. I mean, I get the several or less reasons for every map being pretty similar, but having something different would be refreshing. Egypt and Suijin are good examples of this, actually. While I don't think they're great maps for many reasons, I do have to give them props for being different and looking nice. Well, Egypt is kind of bland and repetitive, but still.

Choke points shouldn't be a dime a dozen, but areas that aren't choke points don't need to stretch on for miles in any given direction either. Save choke points for transitional sections and have the other areas either comfortably big (Granary yard for a smaller example, Gullywash second for a bigger example) or sensibly bigger than that (Badlands second kind of pushes it, but it's not terribly huge). No stairs, unnecessary ledges or random cracks in the ground/on the wall. No one should be punished because shit doesn't work right or because they shot a perfect rocket into an area that completely absorbs it. Obviously no low risk, high reward sniper sight lines.

I don't mind water so long as it's not in a position to send you to the sky box and achieve either unnecessary fall damage or undeserving frags. Process in its alpha and beta stages were ridiculous because of the water on second. The same sort of thing applies on Freight. Speaking of trains, while by now this is common knowledge, there really shouldn't be anything that kills you because you touch it. I mean, it's hilarious and adds a fun element, but it's not a fair concept whatsoever.

Every map also needs at least 6 spires, with one maybe on top of another. cp_aggrocrag lets go boys

Do people want a new Koth map in 6v6? I'm trying to make a new Koth 6s map but I don't know if there's a demand for it, since I see people discussing only CP maps atm. And if there is, what makes a good 6s Koth map?

as long as you don't make scouts and snipers literally god on it it's fine.

also the point must not be inside a kind of dip.
Tho is very difficult to say what makes a good koth map since we play just one of the many that valve gave us.
the only way to understand it is to analyse all the other koth maps and see why they haven't been used yet and build an experience on the koth mod.
Surely if you start creating a koth map you'll encounter even more criticisms than for a 5cp map, because anyone has in his mind the idea of what makes a good king of the hill map.

high hiding places for roamers

high skyboxes for dank bombs

This kind of element trashes koth_king and Suijin, they're basically shooting galleries because everything on the point is vulnerable to every kind of spam. Enclosures around it are also a bad idea -- entering them without killing the other team is a death sentence, and they're usually large enough to split the map in half outside it.

I think the geography of Viaduct is one of the reasons it's been so lasting -- the point is uphill but not prominently so, while cliff plus the structures (plus the rock, I guess) provide some verticality for the attackers without completely overlooking the point. Compare to basically every other KotH map, where the point is usually level with or below all open areas. Lakeside is the only stock exception, and it's the only other stock KotH played competitively. (Badlands kinda as well, and it's run in 4s.) And Coalplant's buildings are to the side, so defenders can hold on them.

Big doorways

Don't make the spawn -> point distance too long. Pay very close attention to respawn wave times as well as cap time, these things can change the feel of a map drastically (lookin at you cp_standin).

Could someone move this thread in the newly made "Map Discussion" topic (and maybe pin it too) along with this other thread?

As long as there's sniper sight lines I can use to defend 2nd point and fall back to the nearest spawn/building in a reasonable amount of time then I approve of it as a good 6s map.