I think this is fixed by this PR: https://github.com/screeps/grunt-screeps/pull/19

@trepidimous said in Encouraging more combat at high GCL:

@tigga What if rampart decay rates (and therefore repair rates) were based on a formula that depended on the number of hits that they've been built up to? This way we have a smooth curve where players can select their safety-to-cost ratio rather than an on/off at 10 million.

Math.ceil(hits / 10e6) already does this.

You could make it smoother, the details don't really matter.

@hsoj said in Encouraging more combat at high GCL:

I know we have novice and respawn areas, but making defense harder just makes it more difficult to set up a stable, long-term colony. Personally I think this would be rough on players with less developed defense code.

I don't think any of the ideas outlined make defense significantly harder for new players. The vast majority of new players will get rolled over by boosted squads anyway should they be attacked by an experienced player.

EDIT: I thought I'd expand on this.

1. Decreasing repair effectiveness would mean more energy is spent on building up ramparts. Again, it's going to slow you down a bit economically, but not really make you easier to kill unless you can perform an effective repair spam defense. If you can do an effective repair spam defense, you're no longer a new player IMO.

2. Increasing rampart decay as proposed would only kick in at >10m ramparts, and would be a fairly insignificant cost. Sure, you'd get a little less into your controller. Really doing this means that ramparts become more of a trade-off. Build them high and be safe, or keep them low and get more free energy.

3. You get 1 safe mode per RCL. That's plenty enough for a new player to build rooms up IMO. Most new players don't have auto-safe mode.

4. Decreasing the effectiveness of burst damage. On the whole new players will not be able to kill T3 boosted squads attacking their rooms. If they can, they're not new players any more.

On the whole the world is very peaceful and most new players won't get killed by established players. Fundamentally even with the best code in the world you can't really grow up as after a respawn next to an established player if they don't want you to. Bases can't really defend themselves against RCL 8 attacks until they're at least RCL 7, and you'll be dead well before that. The reason new players don't have massive problems is because established players aren't trying to kill them, not because they couldn't.

I'd participate and would enjoy it. Ocasionally people propose custom rulesets for botarena and SWC and I've never really been a fan, but they're two week events rather than seasonal.

Ticks per season is going to be an interesting thing to converge on. The ScreepsPlus seasonal server is on ~2.25 million ticks and has two GCL 10 players, one GCL 8, three GCL 7.

The other non-standard thing to think about is server size. More cramped servers certainly alter Screeps experience - combat is a lot more effective as you need space to harvest. Bigger servers (like the main public one) favour the economy game more.

Changes today are good. Bring level 4s up to a point where you can't just snipe them, brings level 5s down a bit so that if you get lucky they're only "hard" rather than "very hard" and the double fortifier means "fixed function" strategies such as corner nibbling may just fail due to lack of damage.

Nukes are still an issue IMO.

@xenofix said in Broken Pathfinder:

Thanks Tigga. That's really interesting. It shouldn't fail like this, should it? I guess my problem is solved.

However the problem is still there for other players: I believe that it should always be possible to find a path inside of one room without any extra options. Most newbies rely on it. Maybe someone increases the default maxOps?

The other thing you can do if you're sure there is a path within a room is to set maxRooms to 1. That constrains the pathfinder somewhat and will prevent it exploring into other rooms.

As for defaults - it's not easy. What if there is no path? How long should it carry on searching until it can prove there is no path? Often incomplete paths will actually do a pretty good job. Increasing maxOpts will likely have an impact on CPU, and will increase the cost of every impossible or surprisingly long path.

IMO there comes a point where every Screeps player has issues with pathfinding, and has to dig in and figure out how they want to solve it. There's plenty of other issues that you'll bump into with moveTo as you keep going. Fortunately the game provides a lot of options for interacting with pathfinding - via options, via cost matrices and via direct access to Pathfinder.search. Many people have thousands of lines of code dedicated to going from one place to another, and still have issues from time to time. Getting it right is part of playing the game and things like moveTo are really just to get you started IMO.

Which is the wall it passes through in E54S18?

The E51S18 looks like it's just an issue with ops - it's a complicated room and moveTo with default options will sometimes fail. That's because the pathfinder has a default maximum operations of 2000. You can override this by passing maxOpts in the options arguement.

EDIT: Same with the E54S19 one. The default arguments will do fine for most rooms, but anything requiring a lot of twists and turns or going back on yourself tends to be more expensive and will hit the ops limit sooner.

@nicle said in On the defensive power of strongholds:

@tigga I think strongholds don't shoot power creeps is just a bug.

I'm sure it is, yes.

I wonder if level 4 strongholds shouldn't use tower repair when there are no creeps to shoot.

@orlet said in On the defensive power of strongholds:

Agreed on the overly steep difference between the Lv4 and Lv5 stronkholds. But I also do have a feeling that Lv5 only seems strong now, until someone can figure out how to cheese it...

Maybe. There are some known ways to "cheese" level 5s, but for the most part they just make your life easier. For example you can just copy-paste their defense code and get perfect pre-healing. That makes tough more effective and means you can save on a lot of parts and instead turn them toward damage. The fact that the defense code doesn't have any randomness is a big weakness IMO. If it just multiplied it's damage calculation for each target by 0.8 + 0.4 * Math.random() it'd be a lot more dangerous. Personally I've not abused this yet as I'm aware that they intend to improve the way strongholds work, and if that makes the defense make more effective use of their resources code to exploit these issues will become redundent.

In Slack we've talked about a couple of other methods - nibbling away at corners where only one ranged creep can reach you (though repair creep causes problems, as do strongholds without exposed cornes), and using fortify+sheild (helped by the fact strongholds don't shoot power creeps) to create a indestructable cover to shoot at the stronghold from. They're quite a bit more complex than the cheese for lower levels 4s and aren't so clearly cheese.

https://screeps.com/a/#!/room/shard3/E45N34 (replay inactive as nukes...)

I brought this up with @o4kapuk on slack last week, but thought it was worth popping in the thread. Right now it is very easy to take down a L5 stronghold with two nukes and loot 3/4 containers. Effort: tiny, risk: tiny, cost: tiny, reward: huge. This is only one way of doing it. An alternative method would be to nuke above and below the core so that everything dies except the core and two containers. That alternative method would work if you changed things to make the container loot somehow be destroyed on nuke impact as you'd still get all the core loot.

IMO at the very least the repair creep needs to be prioritized depth first (repair core to strength, then towers, then the rest), rather than breadth-first. That'd mean a lot of empty ramparts died, making a ground assault easier, but you'd need a much greater nuke count to actually clear it with nukes.

The real "fix" for me would be blasting away all the loot on any nuke impact. Maybe make the containers tombstones or ruins (that'd also fix the >5 containers in a room issue) then add additional logic for the core to clear the store if any nuke lands?