Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[haibane]

I know this topic floats up sometimes, and I will be repeating ideas which were already voiced. But in this post I would like to focus less on how squadrons should be implemented, but why it's such an important idea for the game.

First, I'm a newer player and personally quite fond of the combat system. This style of combat featuring RTS with loose commands and AI autonomy caters to me a lot.
I don't like micro-heavy games (Starcraft, AoE and friends) and enjoy watching AI fumble around and do its own thing. Still, after playing through something like 30+ hour campaign found it... half-boring and half-annoying by the end. Why?

The reason, there is only one tactics which more-or-less consistently works: deathball. It isn't that hard to make the case. The ship combat is effectively a mob vs mob action. There is no coordination, there is no nuance, there is no planning. Simply concentrate as much firepower in one spot as you can and hope that on average your ships outgun the opponent. Because this is the only thing you actually can reliably do.


Let's talk about actual squadrons.

Forming a squadron enables individual ships to follow up on three key behaviors:

• Stay together, move together, act together.
• Assist on target. In general, this is fine for squadron to target multiple enemies at the same time, but should keep this number to minimum if possible.
• Flux rotation. Ships are stronger at low flux and weaker at high flux. Weaker ships should try to retreat behind its allies to cool down/vent, and its allies should try to cover up for weak ones, especially during venting.

Despite being "simple", those have profound consequences.

First of all, it removes a lot of complaints about combat AI because it will look a lot smarter. Why AI looks dumb in the first place? The three behaviors above are naturally picked up by the player, quite often subconsciously. Since there is little control over individual ships, there is a reasonable expectation that AI will be able to perform those as well. Except it is not.
Group tactics require coordination between ships and since ship's AI is individualistic this falls outside its scope. Which leads to a lot of frustration and desire to micro-manage the ships since they seemingly cannot handle even the basic thing, forget any complex tactics.

Staying together empowers existing command system, making it a lot more sensible and impactful for the player.

Currently, you don't want want to give too many commands at once. Random ships will pick random commands, disregarding whether there is sufficient amount of force allocated. This effectively disperses your forces slowing execution of each order and makes your ships much more vulnerable to "hero's death" syndrome. On the other hand giving commands frequently is limited by command points, potentially leaving you in a vulnerable state if you run out. All of this further emphasizes few, rare, generic commands.

This collapses fleet command into one-dimensional endeavor. For my fleet there are only two orders: "fight around here" and "kill THAT thing right now". Every other command is worthless.
Harass, Avoid - what do they even do? The Follow command is in fact harmful! Follower ships just hang behind leader apathetically watching it die without contributing at all. It seems like a good idea to use on non-frontline ships, but it often makes things worse. The assignment may cause some other random ships to join in, removing them too from battle. To add insult to injury, those ships are only concerned about keeping formation, so in case leader is surrounded and turn to a side to shoot a random frigate they readily float up into the frontline. Just because they have to stay behind the leader and that behind so happened to be right into enemy's face. Yes, I lost my Mora's and Gryphon's to this a number of times before I learned.

Squadrons change situation. You can keep the existing system where AI chooses which task it wants to undertake, except now it is groups of ships doing so. This renders every task much more impactful. More ships are involved in every order, making you consider both size, count and composition of squadrons depending on what you want them to do. It also means AI is much more likely to allocate an appropriate amount of force for a task while staying functional locally and without spreading thin, even if it means leaving some goals unattended. Which in turn implies that fewer tasks can be assigned to, so placing both good and bad orders has higher impact. Additionally, this gives player much clearer and immediate feedback on what AI is doing/working towards and how your orders impact its decision making. Put together it results in a sense of better control and improved player agency.

Also, with basic behavior out of the way, player can finally focus on higher-level tactics and use command system to direct and correct AI actions. I feel like this is what the existing system is supposed to be in the first place, but utterly fails to do due to AI incompetence.

Assisting on target has impact on fleet-building.

I encountered a problem that you cannot have specialized ships. Not because the specializing is conceptually bad, but because it doesn't work well with AI. One combo I wanted to do is full Sabot/Squall Gryphon + Dominator with anti-armor kit. Gryphon instantly shreds shields and Dominator rides in to finish off the target. It sounds reasonable on paper, it works in simulation, it is hopeless in real battles. They just never target the same ship, so Dominator ends up wasting flux into shields 90% of the time while Gryphon plays whack-a-mole shooting frigates with Squall. Eventually, I formulated a simple doctrine, every ship must be a self-sufficient fighting unit and be able to do everything as if it's alone: put pressure on shields, shred armor, do hull DPS, field PD. The only exception I have found is Harpoon missiles thanks to hard-coded logic.

Squadron because of limited target list helps multiple allies to converge fire on a single ship. This means, besides minute benefit of more efficiently focusing fire on specific targets, your ships can have complementary roles, excel at one thing and rely on others to provide for what they lack. It opens a number of new possibilities in equipment, ship outfitting and fleet composition.

Lack of flux rotation is a big cornerstone for deathball tactics. Let me explain.

What happens when a slow ship on the frontline is high on flux? It tries to retreat back. But it tends not to work:

• In big fights there are typically some allies with no flux standing behind and pressing forward so they can get to the front. They rarely consider moving out of the way so normally there is just no space to retreat to.
• Ship can't simply vent when facing many enemies.
• It has enemies in front, which it must shoot. This keeps flux consistently high as well as prevents anyone else from providing cover to lower the pressure. I can certainly understand, no one likes being shot in the back.

All in all it means a weakened ship get pinned down against a wall of fresh and combat-hungry allies while awaiting its demise. This applies in your direction as well, so how do you win? Answer is very simple: whoever has the longest and most flux-efficient stick tends to win on average. And as soon as opponent dies, its remnants provide enough cover for you to reset flux and focus neighbors, allowing to snowball further.

With flux rotations "balling" becomes more difficult, weak ships get covered for by their allies which give them a breathing room. Recklessly pushing in might finish them off but also might catch you in cross-fire against refreshed ship and its companions. Flanking and efficiently coordinating help across battlefield becomes key to bloodless victories.

Besides, it removes single biggest source of frustration for players, where they watch how their key ship gets forever stuck at high flux in most dangerous spot spamming Reapers into PD.


You possibly heard this reasoning already, so why am I to repeat it? The cause is the latest blog post on unquiefying factions: https://fractalsoftworks.com/2022/03/18/uniquifying-the-factions-part-1/.

Reading it I couldn't get rid of a feeling that devs are fighting a wrong problem. You see, the reason why factions feel the same to play against is because... they are! Every fight is the same: just one mob beating up other mob without rhyme and reason and an occasional shout from you to press harder. The only thing that possibly stands out of a mob is a ship which can level good part of enemy mob singlehandedly: battleships. And devs do recognize that.

However, I don't think designing new battleships is going to provide desired effect. Unless a ship is so OP to manhandle good part of a fleet alone, or possibly rub against general AI behavior in wrong way, it won't change general dynamics. Every fight will stay as mob vs mob, with only change is what ship I put personal kill order on. By the way I don't say more battleships is bad or more ships in general is bad. It's just we are fighting a symptom and not the disease.

Besides immediate improvements to combat system I believe squadrons also might have a lot higher chance at distinguishing factions. Let's use a bit of imagination:

• Tri-Tech: squadrons are formed around single powerful ship with a number of smaller utility escorts + "rogue" phase ships to flank and distract you
• Hegemony: roughly evenly sized squadrons, each formed around 2-3 destroyers/cruisers or capital, typically rounded up for a head-on shootout with rare highly specialized groups (e.g. full fighter, missile barrage, hit-and-run, etc.)
• Pirates: few unevenly sized squadrons formed randomly from bigger hulls, everyone else acts on its own as in Seek and Destroy. It's pirates, what do you expect?
• Independent: 2-3 big squadrons formed around bigger hulls, a number of "elite" ships acting alone (e.g. better hull/weapons/officer)

I didn't include Persean and Sindrian because I still don't know what their selling point is. Also I'm convinced that both Ludd's are just green pirates who like explosions, so they are out too.

Now, try to answer few questions for yourself. Do those compositions above have a specific impression on you? Imagine how does it feel to run into one? How do you approach the fight? Can you recognize it just by glancing on the battlefield? I bet most players will think that pirate fight is free until they forget to put escorts on the capital and see it go down in flames.

With those in mind we can go further and introduce new ships to enhance ideas and give more flavor. And not only battleships. For example, let's create the following ship for Hegemony: destroyer, wide oval shape, low hull, high flux capacity but low flux dissipation, multiple front- and side-facing small ballistic turrets, 2 medium missile hardpoints facing off to the sides. Ability: dash similar to Shrike, but short, 3 charges and in any direction. What do you think about this?

In current game I might look at it because of unusual shape, see weird stats + bad weapon points and forget it ever existed.

Let's add it to squadron: fit it with PDs, some kind of guided missile (Harpoon, Sabot, Breach, Pilum), -15% shield damage mod. If guns is Hegemony's sword, then this ship is Hegemony's shield. The ship fits really well into faction's idea. As a military state it loves efficiency and doesn't like paying death reparations. It is also organized enough to have people agree to step into the line of fire to protect slower comrades in moment of weakness. All of it feeds into the doctrine of "we will fight you head-on and we will win that fight".

Will the player notice the ship? Most likely yes. It is incredibly annoying since it prevents you from quickly finishing off wounded ships: it darts in front, wide profile with good shield capacity protects against normal attacks and good PDs deny easy Harpoons. Also, what is worse than a giant ship full of guns? A giant ship full of guns which can reset flux unpunished.

How do you deal with one? Player may notice that it has 180-degree frontal shield, no back-facing mounts and low maneuverability, then remember pirates and their own - now scrapped - capital ship. Yes, this is the same way you are intended to deal with Hegemony in general: flanking and mobility. It also further drives the idea that overwhelming firepower is not  everything. Being able to efficiently move and apply firepower can be powerful too. Which, I can dream, might result in a more varied tactical landscape in the future.


Sorry for the longpost. I know it takes effort and won't be perfect the first time, but I believe *any* form of organization on the battlefield is better than none. I really love the game so far, and I just hope it becomes only better given time <3

[haibane]

I honestly didn't want to follow up on this, because from this point it's less about fundamental issues and more about my personal vision of the game, but it's probably worth a discussion.

One thing I don't like is combat tree. The problem, it represent direction of ship scaling orthogonal to everything else (hull, OP, mods) and follows "more=better" mentality. As a result we have a number of clearly artificial restrictions - officer count, levels and elite skills - and full three(!) different skills designated specifically to upping those limits. Officer skills are typically used to "break" builds and create individual strong ships which further amplify the blob-oriented warfare. Why cooperate when you can dominate alone? This feels so weird to me.

I should note that I think that's still OK for the player to have it. There are probably people wanting to roleplay combat hero/veteran/junkie, so let them. It comes at the cost of other skills, it's all fair.

Actually, there is something I didn't mention in the post. Squadron have its own big problem: AI. Whatever you do, you won't be able to satisfy everyone. There is just too much variety in ship combinations, equipment and tactics. AI sometimes can guess what player means, but it is likely to be wrong half of the time. Without any help there is a good chance that individual squadron tactics will devolve into the same blob-on-blob warfare like now just on smaller scale. We need some way to customize behavior and pass hints from player to AI on what is expected from it.

Transferring officers to command over squadrons instead of individual ships neatly solves both issues. Their skills can instead affect how ships behave within the group. One minor hiccup is how to relate skill effects to individual ships, but it is easily solved: we already can put officers on ships, so let's retain that. Ship with officer becomes squadron leader, and skills can be centered around that, for example:

• Swarm: Empty ships focus on protecting leader's sides and back. Retaliate with progressively more aggression against anyone who tries to approach or target the leader until they retreat.
• Cannon fodder: Empty ships, starting with smallest DP, try to position in front and sides of leader and protect it, unless it is actively engaging.

I bet Ludd's officers would love the last one.

To express more complex tactics we can take advantage of other officers slots. An officer brings with him a number of lieutenants (possibly spawned by a skill). Lieutenants are free, have no their own skills and can only be assigned to ships within the squadron of their officer. Their purpose is to mark important ships which you want to have interactions with certain skills. Those can get quite intricate, for example:

• Flying wedge: Your lieutenants line up to the left or right behind you in two wings to form a wedge. They prefer to keep formation when possible.
• Wolf pack: Every empty ship follows up on the target of either officer or lieutenant to the best of its ability. Officers and lieutenants prefer choosing different targets unless heavily outmatched.
• Coordinated pair: You can assign at most 1 lieutenant. Officer and lieutenant always share the target. One of the two is chosen as a lead. Lead actively engages with the target, while the other recedes to rest/vent or provides light assistance. As soon as lead accumulates high total flux, lead is transferred to the other ship.

We can go beyond just nudging AI around and have skills that affect multiple ships. For example:

• Shield link:
  Whenever a ship in formation generates hard flux from shields it only takes 30%, other 70% is distributed evenly between all other linked ships.
  
  Downsides:
  
  • Hard flux only dissipates during venting.
  • Venting now has a cooldown.
  • Ship is unlinked during venting and overload (i.e. excluded from damage sharing).
  • Ship receives 2x hard flux from link if it has no soft flux.

The last clause is to punish "flux tanks" sitting in the back, albeit I'm afraid the skill is still OP and requires 2-3 extra nerfs, but this is not the point.

What I really want to illustrate is that besides just "+25% damage", "+15% range" and "+100% missile capacity" we can also have abilities which change battlefield and ship interactions in interesting, significant and noticeable way by allowing skills to apply to multiple ships at once.

[Drazan]

Every secound topic nowadays is abot Ai incompetence and pleaads for better command. Always ignored by devs and driven down by oldtimers.
This is literrally the only thing (maybe exept some UI problems) that is holding back the game and so far looks like they are refusing to adress it. It's a shame really.

[Megas]

If nothing else, having AI reverted to pre-0.8a would be an improvement - much less backpedaling and cowardice.

[haibane]

Actually, I think even though AI is 100% incompetent with respect to cooperation, individual behavior is quite good. Most poor piloting observed by the player is caused by lack of interaction between allied ships. I noticed only 2 real issues with ship AI: poor engagement range choice and questionable use of certain weapons. But just like with squadron AI I don't think it is possible (or worth) to create smart enough AI to second-guess what player wants from a specific loadout. It is much easier to add hints.

Realistically, all you need is to add engagement modes to ships. It is a combination of target shield status (low flux/high flux/no shield) cross armor (low/high), desired engagement range (just pick a weapon group) and which weapon groups should/should not be used. Just don't make it into hard orders. Make AI simply weigh against every mode and gravitate towards most appropriate one. This way there is still enough leeway so it can act according to the situation, but also player will observe desired behavior often enough.

I wanted to write about it too, but I think lack of cooperative behaviors is a much bugger issue than poor usage of certain loadouts.

[Linnis]

You guys want the AI to have some real sentience *** going on.

The AI is part of the difficulty, learn to wrangle it into submission.

The only complaint is that some orders on the tactical map is context driven and picks *** for you.

[Drazan]

You guys want the AI to have some real sentience *** going on.

No. For the last goddamn time. No. Nobody wants that. What people want is:
A: To have meaningfull orders. If i say to a ship to go there and die then it beter go there and fckn die.
If A is not possible then, B: Ships to have just a bit of sense when engaging thing. Not ramming stations, not chasing frigates twice their speed etc. It's truly not sentience level stuff.

[SCC]

It's worth noting that there's no coordination on part of the AI. It treats each ship as out for itself, with only orders forcing them to act differently (defence on ship keeps them together and escort changes behaviour the most, since the escorting ship will try to stick to the back side of the escorted ship).

[Hiruma Kai]

You guys want the AI to have some real sentience *** going on.

No. For the last goddamn time. No. Nobody wants that. What people want is:
A: To have meaningfull orders. If i say to a ship to go there and die then it beter go there and fckn die.
If A is not possible then, B: Ships to have just a bit of sense when engaging thing. Not ramming stations, not chasing frigates twice their speed etc. It's truly not sentience level stuff.

Personally, I can tell the difference between an exterminate order and an engage order.  The end behavior does look different for me, so that feels meaningful to me.  As for a hold position with the same stickiness as exterminate, that seems like a reasonable request.  It is at least implementable, although the wisdom of its use would be debatable.  Battle lines tend to ebb and flow, and I've had many a fight where I use waypoints only to find my self at the top of the map in an Odyssey while the rest of my fleet is at the mid point, apparently eating popcorn and watching while I unintentionally solo the enemy fleet.

I don't have ships ram stations too often.  The last time I can remember was a poorly timed burn drive (back before they could be cancelled) running into a rotating strut that technically wasn't in the way at the start of the burn.  On the other hand, I don't typically issue many orders on station assaults, so the base ship AI typically works well enough for my fleet compositions in that kind of situation.  Although I can imagine a stronger weight to the "don't come closer logic" that clearly already exists if other people are seeing issues with this regularly.

However, your third request I would argue does require sentience (or at least a lot of algorithmic considerations) to implement properly, or at least doesn't result in just as annoying behavior as we have now, since it's context dependent.  Enemy fleet has mixed slow and fast ships.  Player fleet is made up of SO Hyperions only.  Does half the enemy fleet simply not bother going anywhere and sit at the top of the map (or just grab objectives).  Perhaps you meant to say slow ships should prioritize other slower ships, and if there is nothing left, only then go after the faster ships.  Although even then, having an Apogee decide to go after the 120 speed Lasher across the map instead of the 150 speed Omen adjacent to it might look funny (and equally likely to fail to catch).

Take it to an extreme, and have the player deploy a Paragon from reserve into that all SO Hyperion fleet.  Do all the slow ships stop engaging the frigates right next to them, and make a beeline for the Paragon across the map that just entered (even if they had cornered a frigate on the edge of the map?).  Does the slow ship press through a line of frigates, ignoring their flux levels, towards that capital, or do they have to engage with the frigates in between, despite them being twice as fast?  Do they attempt to go around and fail because the frigates are faster?  I could imagine some interesting degenerate fleet strategies for the Paragon + fast frigates setup assuming such an AI.

What if the slow ship has carrier support or long range missile support from an ally.  What if the ships is equipped with a pile of ECCM Harpoons?  Is it enough to drive the flux up on said frigate, which will be driven higher by pursuing since 120-60 = 60 relative retreat speed instead of 12).  Now that ship needs to be aware of ally weapon and range considerations to properly make an intelligent decision on whether to continue pursuing.

In general should an isolated capital just let a frigate constantly shoot it's engines if there's a distant, slow target?

The fundamental issue is there's no guarantee of any given type of fleet configuration, on either side.  The absolute statement "not chasing frigates twice their speed" only makes sense in a classic spread of ship types versus a similar spread, and assuming the slowest ships die last.  The AI has to be general enough to handle a very wide range of situations, including situations where the "obvious" behavior that should be hard coded no longer works.  Not to mention as the fleets take losses, the AI still needs to do something vaguely sensible.  Capitals chase frigates because sometimes frigates are the only ship around it, and the ship AI has a very local view.  Making a decision of whether it's better for a really slow ship to ignore the really close fast frigate and instead pick from 5 different potentially slower target ships across the map that is going to take 2 minutes to reach any of them is a sentient level strategic decision weighing many considerations, assuming you want it done right.

As for the OP, it would require an entire layer of AI in between the current orders and low level ship AI, that would be just as complicated (perhaps more so) than the ship AI.  With just as many places to result in dumb behavior in slightly unusual situations. 

For example, how does a squadron, focused on an enemy front line, interact with a player Hyperion jumping around inside their squadron formation.  Or a flanking player Aurora pushing ships into each other.  Do they try to maintain formation, rather than engaging the interloper if they are already engaging the main line of enemy ships, so that they're already at some nebulous max number of targets?  Do they just let extra ships above and beyond focus attack their engines unimpeded or do they increase the number of targets they are willing to unfocus fire on?  Does the entire squadron turn to target that new ship, despite not necessarily having clear firing lines and expose their flank to the main line of enemy ships?  Perhaps these could eventually ironed out, with some kinda of relatively quick algorithmic criteria, but it would be a non-trivial effort, likely requiring a year or more of effort (based on the time to get the ship AI to where it is these days).  And I'm not sure how much better it would be than some of the currently organic behavior that individual ship AI produces at the fleet level (i.e. make sure allies are to my flanks and rear, get the extra enemies to be on the opposite side of my target, and flank my target).

Also, as a side note if you weren't aware, if you don't want random ships picking up commands as losses happen, there's the search and destroy (or it's called something like that) that has the ordered ship NOT get assigned to any other orders, and just uses the base ship AI.

[Morrokain]

I've been thinking about this a lot recently and obviously this isn't the first or likely last time its come up, but-

However, your third request I would argue does require sentience (or at least a lot of algorithmic considerations) to implement properly, or at least doesn't result in just as annoying behavior as we have now, since it's context dependent.  Enemy fleet has mixed slow and fast ships.  Player fleet is made up of SO Hyperions only.  Does half the enemy fleet simply not bother going anywhere and sit at the top of the map (or just grab objectives). 

Yes. It's better that the ships stay together in almost every scenario I can think of. If they can't catch anything anyway, spreading out is exactly what the enemy fleet wants so that they can be easily picked off by the concentrating faster ships on isolated ones and leveraging numbers. The slow ships can move towards the center or something after objectives but the onus should be on the player to engage them not for them to engage the player. And that is only because the player has the faster ships. If the situation was reversed, that is where orders like Search and Destroy and the various override orders come into play. But it shouldn't be the default behavior imo.

Perhaps you meant to say slow ships should prioritize other slower ships, and if there is nothing left, only then go after the faster ships.  Although even then, having an Apogee decide to go after the 120 speed Lasher across the map instead of the 150 speed Omen adjacent to it might look funny (and equally likely to fail to catch).

Yeah this is something that would have to be weighed and prioritized or it would indeed look funny. This is the most complicated part I think. I could definitely be missing more edge cases too in my thought process. Unfortunately, actual implementation would probably be required to flush them all out.

Take it to an extreme, and have the player deploy a Paragon from reserve into that all SO Hyperion fleet.  Do all the slow ships stop engaging the frigates right next to them, and make a beeline for the Paragon across the map that just entered (even if they had cornered a frigate on the edge of the map?). 

Yes. The reasoning is simple. There is no point to engaging the frigates. If they can't be bursted down (and they already likely would have been) then nothing is actually being accomplished by engaging them in the first place. Flux damage isn't actually damage when the ship is a lot faster. If they can be bursted down then not engaging isn't an issue because when the frigates naturally decide to harass they get bursted down. Er, does that make sense?

Does the slow ship press through a line of frigates, ignoring their flux levels, towards that capital, or do they have to engage with the frigates in between, despite them being twice as fast?  Do they attempt to go around and fail because the frigates are faster?  I could imagine some interesting degenerate fleet strategies for the Paragon + fast frigates setup assuming such an AI.

The first thing. Barrel through the frigate line. Same reasons as above. If the capital ship can't ignore frigates it can't catch then is it really a capital? If it loses a battle with another capital because that frigate line put too much flux pressure on it while breaking through then all the better! Now there is a better reason for escorts and the important thing is that its very intuitive! Chasing a frigate while the enemy capital mows through everything isn't intuitive at all to me.

What if the slow ship has carrier support or long range missile support from an ally.  What if the ships is equipped with a pile of ECCM Harpoons?  Is it enough to drive the flux up on said frigate, which will be driven higher by pursuing since 120-60 = 60 relative retreat speed instead of 12).  Now that ship needs to be aware of ally weapon and range considerations to properly make an intelligent decision on whether to continue pursuing.

I would just ignore that kind of complexity entirely to be honest. The choice of whether to pursue is basically a binary decision based upon relative speeds with the wrinkle already mentioned above. I don't think it actually needs much more nuance than that in the majority of cases. I'll further explain my reasoning below but I've already touched on it a bit.

In general should an isolated capital just let a frigate constantly shoot it's engines if there's a distant, slow target?

Either that is a feature of the ship and it really needs allied escorts, or it has rear guns exactly for that reason. As stated above, taking the frigates into account is a tactical error because that isn't what the capital ship is really supposed to do. If there was a concrete benefit to turning and engaging the frigates then ok but I honestly don't think that there is in over 90% of cases. It just slows things down a lot. Probably more so than the engines getting taken out a bunch because the frigates can always vent and close the gap and the capital can't actually get anywhere it really needs to be because its always concerned with the presence of ships it cant catch but could beat if it could. Hopefully this makes sense as I'm not sure I'm explaining it all that great.

The fundamental issue is there's no guarantee of any given type of fleet configuration, on either side.  The absolute statement "not chasing frigates twice their speed" only makes sense in a classic spread of ship types versus a similar spread, and assuming the slowest ships die last.  The AI has to be general enough to handle a very wide range of situations, including situations where the "obvious" behavior that should be hard coded no longer works.  Not to mention as the fleets take losses, the AI still needs to do something vaguely sensible.  Capitals chase frigates because sometimes frigates are the only ship around it, and the ship AI has a very local view.  Making a decision of whether it's better for a really slow ship to ignore the really close fast frigate and instead pick from 5 different potentially slower target ships across the map that is going to take 2 minutes to reach any of them is a sentient level strategic decision weighing many considerations, assuming you want it done right.

I think you are sort of right here, but I also think we have a tendency to over think things as well or otherwise expect more complexity than is really needed. Why do I feel this way? Thinking about it, the painful thing about a capital chasing a frigate is that it will never, ever, catch it in a reasonable time frame that brings value to the battle even if it eventually does catch it and kill it. The big thing here is the "value" part and of course that's high subjective. A capital breaking the loose "formation" to go on a wild goose chase will always be a negative thing imo and so it doesn't bring value to me. Honestly there isn't a situation in mind where its beneficial to do this off the top of my head, but I'm sure others could come up with them.

Anyway, to me I think the best course of action for the AI would be to not ever pursue it and let the ship come to it instead. That is, unless an order to pursue/engage/eliminate, etc was given. Capital ships are not obligated to engage anything but stations, really, or anything they can catch. They are the anvil around which the hammers of the battle operate. 

Taking my earlier comment into account, why should capitals brute force their way through enemy frigate lines? Because they can a lot of the time, and that's a lot better than stagnating the battle by getting out of position. If the capital gets destroyed because it over extends itself, that is a player problem in my mind because its really easy to avoid with a defend order or a target redirection or even taking manual control. But what it certainly isn't in this case is an AI problem. The AI is doing what it should be doing and have the ships engage targets it can realistically destroy. Anything in between is part of the tactical layer of an engagement.

To put it simply, I think battles should be far more decisive than they are right now and it would make for better combat both visually and in a tactical sense if ships operated this way. What we have right now is a lot of stagnation by ships being overly cautious and spreading out too much. The really painful thing here is that despite all that work put into these edge cases to try and make sure the player doesn't lose ships in silly ways, it still happens from time to time. So what is really being accomplished here when you think about cautious AI? It isn't fulfilling its primary directive, and the cost of that directive brings a host of other problems along with it that are equally if not even more painful.

Alex has stated that part of the reason the AI is cautious is so that the player gets to do the heavy lifting in a battle. That does indeed make a lot of sense don't get me wrong - especially considering the narrow view of the player outside of the tactical map. However, when you consider scale it gets a bit messier. I don't think the player wants to personally destroy every enemy ship if that makes sense - and I think that's what we are close to at the moment.

I think cautious AI was a mistake that was a well-meaning reaction to player frustration over ship losses. I was talking with my brother in detail about this and I think the better approach to that problem was to tackle it at the campaign level. For instance, tweaking the recovery system to always allow player ships to be recovered without an OP tax (because why does it really need that?) and something like templates that make restoring a built ship with as little button presses as possible.

Our conclusion was that the real reason that losing ships is annoying has little to do with the money lost or the supplies, etc. That is manageable as long as there are ways to make money without combat - and there are plenty. The real reason is the tedium of rebuilding your lost ship and by the fact that in the current system RNG is such a large factor. If it takes you half of a campaign to find the ship you want and then it gets lost in a battle because of AI derp - that's where you are going to see angry posts on the forums. If its just a matter of pressing a few buttons and maybe waiting a bit (we discussed a mothball sort of situation that takes docking with a station to repair or something along those lines) as well as what essentially boils down to lost money, its a lot less painful to lose ships.

So to really simplify this, iirc the line went something like: "You should really only have to build a ship once. After that, its just some sort of minor penalty for losing it that makes it unavailable for a bit but you don't have to find it or its weapons again."

To summarize: The AI should be more decisive for the sake of both tactics and battle stagnation with the implied expectation that ship losses are more common. Because ship losses are currently really unfun, that needs to be addressed at the campaign layer with a variety of QOL improvements which I've detailed a bit.

Now I could definitely be wrong here and I don't want to make it seem like I 100% know the answer to these issues, but after a lot of thought and some discourse on game design this is what I came up with. Hopefully it can fuel some friendly discussion at the very least.

[Megas]

Re: Morrokain's post
This is why I write that bounties underpay.  The rewards are alright only if the player can win flawlessly repeatedly, which means the player's fleet is overpowered (relative to the enemy).  If the player loses a big enough ship, there goes the bounty reward.  Assuming casualties remain as costly as they are, named bounties should pay at least twice as much like contact bounties, and jobs (like warehouse raids) that spawn revenge fleets should match pay of contact bounties because they persist and chase down the player like the Terminator.

Aside from Reinforced Bulkheads or Hull Restoration, an officer or s-mod on the ship grants near guaranteed recovery.

Ships can be recovered, but unless the player has Hull Restoration, getting more and more d-mods on the ship is as good as losing the ship (unless player has Derelict Operations to exploit d-mods).

The biggest problem with restoring ships is the cost.  Unless player has Hull Restoration, restoring most ships are very expensive.

AI (at least on the enemy side) should retreat off the field if it can or become suicidally berserk if it cannot after its PPT times out and CR decays.  It gets annoying that the AI maintains its cowardly and stalling behavior even while CR is ticking down until it literally cannot anymore because of engine flameout at zero CR.  If the enemy is that cowardly, it should try to flee instead and let the player auto-resolve the ensuing pursuit.

I would not mind more casualties, more aggressive AI (at least to pre-0.8a behavior), easier ship acquisition and/or recovery, and bigger rewards to offset increased lethality.

[intrinsic_parity]

The first thing. Barrel through the frigate line. Same reasons as above. If the capital ship can't ignore frigates it can't catch then is it really a capital? If it loses a battle with another capital because that frigate line put too much flux pressure on it while breaking through then all the better! Now there is a better reason for escorts and the important thing is that its very intuitive! Chasing a frigate while the enemy capital mows through everything isn't intuitive at all to me.

Really don't agree that this should be default behavior. IMO The AI should be able to behave reasonably regardless of the decisions that the player makes (loadouts/orders). Barreling through a line of enemies (of any size) has a huge amount of risk associated. All it takes is one reaper, or a big burst of ion damage to the engines, and the ship is dead. There are no ships that can be safely ignored IMO, every ship poses a threat and needs to be addressed (not chased, just addressed). Of course you can give orders or design loadouts to try and mitigate that, but the AI shouldn't assume that the player has done that, and in fact the AI will look quite stupid if the player fails to do that. If the player wants the capital ships to charge like that, it should be on them to give the order.

You seem to be suggesting that the AI should default to taking risks and the player should be responsible for mitigating the risks with orders. That seems like a recipe for disaster to me, since it relies on the AI's ability to asses the risk and value of actions which it is not good at IMO (and is a really really difficult task, even for a human). I think the current default behavior is correct: the AI defaults to playing it safe and the player is responsible for telling it where it should take risks via eliminate/engage/search and destroy orders. If there is a problem with the games current approach to AI risk taking, it is that the player doesn't have enough/effective enough tools to give the AI these orders (and maybe that the enemy admiral AI is not agressive enough).


I also don't think the problem is the local/tactical decisions of the AI. The behavior of 'keep guns/shields pointed towards nearby enemies while moving in the direction of my objective' that is already in place is perfectly reasonable IMO, and I think a lot of the tactical movements that the AI uses (pivoting around ships to move to the flanks/spread out to engage the enemy rather than bunching up) work quite well. The issue IMO is in large scale/strategic understanding of the battle and knowing where a ship needs to go to maximize value, i.e. 'what is my objective'. I really don't think trying to make an individual ship AI that can understand that sort of stuff will be a productive undertaking. If that sort of decision making needs to be automated (you could argue that is the role of the player not the AI), that definitely should be the domain of an 'admiral' AI which makes decisions accounting for the entire fleet, rather than something each ship is doing individually IMO.

I think maybe what is needed is more abstract strategic orders. Something as simple as a 'try to engage ships of the same class' order could be helpful. But I think a more general 'try to move towards the center of the enemy formation' order vs a 'try to flank' order would be more useful (and also a 'retreat towards allies' order). In practice that could mean doing some sort of DP weighted clustering (or DP density calculation) to find centers of combat and basing some of the positioning logic on that. That also might achieve some of what the OP was thinking of: ships with the same orders (e.g. flanking) will tend to move towards the same areas and thus stick together. It also could be helpful to have a sense of where allies are concentrated using the same DP density/clustering to identify where safer areas of the battle are for retreating even without orders.


I think the campaign level risk/reward is an entirely separate issue. If the campaign doesn't punish you for losing ships/taking damage as much, that doesn't mean that the AI needlessly taking damage or making bad decisions in combat is less problematic. I think those combat issues still need to be addressed, even if they are not so harshly punished on the campaign layer.

I do kind of like the idea of changing the campaign recovery system though. Something like making every friendly ship always recoverable with much lower rates of d-mods, but hugely slowing down CR recovery rates so that ships which are lost or take lots of damage are also 'out of commission' for a long time. You could add much higher chances for d-mods if you fight on low CR too to discourage the player from using ships on low CR. The idea is that the penalty for loses is in combat uptime vs a direct monetary penalty of repairing and restoring. That also gives more value to having backup ships in a fleet where right now there is heavy incentive to not have more than one deployment worth of ships due to skill DP scaling and bonus XP. That actually reminds me of the Xcom injury system. But maybe that should be a separate discussion/thread.

[Morrokain]

Megas:

I agree with that assessment yeah.

intrinsic_parity:

You make some really good points, but I don't agree with everything. To start, the original premise:

IMO The AI should be able to behave reasonably regardless of the decisions that the player makes (loadouts/orders). Barreling through a line of enemies (of any size) has a huge amount of risk associated. All it takes is one reaper, or a big burst of ion damage to the engines, and the ship is dead. There are no ships that can be safely ignored IMO, every ship poses a threat and needs to be addressed (not chased, just addressed). Of course you can give orders or design loadouts to try and mitigate that, but the AI shouldn't assume that the player has done that, and in fact the AI will look quite stupid if the player fails to do that. If the player wants the capital ships to charge like that, it should be on them to give the order.

I think this is the heart of what we are actually discussing here. The AI isn't really behaving reasonably when it chases something it can never catch across the map. And "addressed" is pretty vague, but I'm assuming you just mean facing with shields up or something. The other things you are describing are perfectly fine, but chasing isn't imo. To me the difference between a ship getting destroyed because it was kited out of position and not contributing and one that was destroyed because it overextended while pushing into the battle line is massive. If we are talking about reasonable expectations then the second scenario feels both a lot better to me when it happens and more easily mitigated than the first.

Is that an admiral AI kind of thing? Sure. But, you can certainly argue that to the player this is unnecessary babysitting and relies upon more orders than are currently possible if they are supposed to mitigate that behavior with orders. That's why builds must be heavily optimized for the harder fights and why the tempest and phase ships are meta. (Not the only reason of course but a large factor.) They won't have this problem because they are generally fast and they will pull larger ships into positions where they can be overwhelmed under default AI without the player needing to babysit them as much. That's a pretty huge deal. Its also why the enemy AI seems less stupid than if the player doesn't give any orders. It issues escort commands to larger ships.

The current set up of "choose your partner and do-see-do" across the map has some valid problems right now. There isn't a sense of any battle cohesion and requires the player to A) understand that the AI is going to do that and B) make sure it doesn't. Hence lots of suggestions for formations, etc. Let's look at B in a vacuum. You have to "fix" the AI with an order of which (unmodded) you have very little of to begin with. That's not really ideal right? So I don't think that we can ignore that this is a problem. Even if the issue is player perception rather than strictly true, it's still valid because that's the player perception if that makes sense. The real trick is finding the root cause. In this instance, that is how I analyze the OP's suggestion. Its less that formations are strictly needed and more that the AI either has to have a cursory understanding of general position and a more refined strategic layer OR the system must rely upon some strategic assumptions to work properly even considering the current good behavior of trying to flank, etc. One of these assumptions would be that slower ships act as anvils and faster ships act as hammers and so slower ships can realistically pursue targets it can catch instead of getting kited into a bad position. And I'm not even saying that there shouldn't be a threat analysis there but rather that the threat analysis currently overcompensates and causes stagnation. The main difference between the two scenarios in question is what we have now requires repeated adjustments and is generally more annoying and assigning an escort order on a slow ship that is barreling towards something it can hit is something that only needs more adjustments if the overall attack fails. And if it looks like it will the player can pull it back. Will it cause more losses? Absolutely. But that's only bad if it causes heavy setbacks for the player. It doesn't necessarily have to. Whereas losing a battle should have some long-term consequences, losing a ship shouldn't set progress as far back as it does now. They are sort of separate topics, but they all tie together in the overall feel of game.

And that is before we get into the can of worms that is the current order system in general. I've stated this before, but orders need to be more concrete. If there actually existed an order that would do what you are describing then that would be a fair assessment (so taking the opposite approach to what I proposed but still having the behavior possible) but in my experience that is not currently the case. Far too many things can disrupt the order at large and the AI is making these decisions against the player's wishes. That still doesn't address battle stagnation either unless the admiral AI actually gives those orders.

You seem to be suggesting that the AI should default to taking risks and the player should be responsible for mitigating the risks with orders. That seems like a recipe for disaster to me, since it relies on the AI's ability to asses the risk and value of actions which it is not good at IMO (and is a really really difficult task, even for a human). I think the current default behavior is correct: the AI defaults to playing it safe and the player is responsible for telling it where it should take risks via eliminate/engage/search and destroy orders. If there is a problem with the games current approach to AI risk taking, it is that the player doesn't have enough/effective enough tools to give the AI these orders (and maybe that the enemy admiral AI is not agressive enough).

To split hairs a little bit, I think you got the wrong impression when you say that I'm suggesting that we have the AI assess all risks in a sophisticated way. I'm actually arguing the opposite. I don't think it needs to in all cases because of baseline assumptions of what ships can or cannot do. So:

I think maybe what is needed is more abstract strategic orders. Something as simple as a 'try to engage ships of the same class' order could be helpful. But I think a more general 'try to move towards the center of the enemy formation' order vs a 'try to flank' order would be more useful (and also a 'retreat towards allies' order). In practice that could mean doing some sort of DP weighted clustering (or DP density calculation) to find centers of combat and basing some of the positioning logic on that. That also might achieve some of what the OP was thinking of: ships with the same orders (e.g. flanking) will tend to move towards the same areas and thus stick together. It also could be helpful to have a sense of where allies are concentrated using the same DP density/clustering to identify where safer areas of the battle are for retreating even without orders.

These would certainly help, but they don't really fix the key issue which is the default behavior. In order for this sort of thing to work, you'd definitely need more command points. And then you get closer and closer to RTS micro as you implement more and more of these sorts of things for edge cases. I'm not sayings its a bad idea or anything, just stating that these things aren't exactly the stated design goal of "orders are for special occasions and the AI is pretty autonomous" which I think is still the current intention.

What I meant by "tactics" isn't that the AI needs to necessarily be more tactical, but having a capital ignore a frigate line creates tactics because it creates a risk that has to be mitigated. And the important things is its decisive in nature. Its really no different than the current system as far as necessitating orders except that I think its more intuitive and creates more interesting scenarios at the cost of heavier losses. And even then, that is only before the player learns to properly escort the ships that act this way.

Basically I disagree that the behavior is a "bad idea" by the AI because the "good idea" of battle stagnation is pretty boring at the extreme levels like what Megas was referring to.

[intrinsic_parity]

To me the difference between a ship getting destroyed because it was kited out of position and not contributing and one that was destroyed because it overextended while pushing into the battle line is massive. If we are talking about reasonable expectations then the second scenario feels both a lot better to me when it happens and more easily mitigated than the first.

Just to be clear, I don't want capital ships chasing frigates (and agree that the AI should be changed in some way to reduce that), but there's a massive gap between not chasing frigates and yeeting yourself through a bunch of enemy frigates to get to the enemy capital ships. I'm not even against trying to restrict chasing of faster ships, I just think blanket ignoring faster ships is a terrible terrible idea that will result in way more issues than it could possibly fix. I prefer the scenario where a ship gets kited, because at the end of the day it's not dead, and I can just give it an eliminate order to get it back into the fight (I think it's actually quite easy to mitigate kiting with orders, just a bit annoying). A ship that suicides is dead for the rest of combat, even if campaign concerns are ignored. And the margin for error (in terms of how quickly you have to notice/react to recover) is much lower for a ship suiciding than a ship getting kited so I think the babysitting issues are much worse in that case. You still need to constantly check if your ships are going in the wrong direction, it's just that the wrong direction is now into the middle of the enemy formation instead of towards a random frigate. I don't think you really fix anything. In fact, I think you make a lot of issues worse.

And even then, that is only before the player learns to properly escort the ships that act this way.

I also think you're extremely optimistic if you think that players will learn to use escorts/orders instead of coming to the forum/dev whining about how their ships are suiciding into the enemy formation . No matter what agency the player is given, people will still not understand how to use it, and complain that the AI is bad lol. IMO that's what is happening right now. People have a bunch of expectations about how the game should function, and don't spend time figuring out how it actually functions and how they can best utilize the tools they have. That's not going to go away, nothing will be intuitive for everyone.

FWIW, it's much more intuitive to me that I need to give high level orders like 'go over there and engage those ships', than orders like 'go help/escort the capital ship that is currently ignoring the frigate unloading a heavy blaster into it'....

With the current order system, in the scenario where you have a capital ship with a few smaller friendly ships near a group of enemy frigates, and an enemy capital far away, all you need to do to get desirable behavior (IMO) is to order the smaller ships to defend/escort the capital, and then order the capital ship to eliminate the enemy capital ship. That level of player decision making doesn't seem unreasonable to me, and both of those orders seem very natural. The escort order is likely not even necessary IMO.

RE: stagnation
Personally, I don't really have the same experience with stagnation being a major issue. When it does feel like progress isn't being made in combat, it's usually because the battle has become very condensed and ships are concentrating fire to the point where it's actually not wise to push into the enemy formation. The solution is to flank, which causes enemies to split focus and creates opportunities for aggression, and the AI does a decent job of flanking in my experience, it just needs some improvements on knowing how to not overextend and when it needs to play more safe because it has no allies around to assist (where I think some DP density understanding could help). The AI does sometimes need eliminate orders to effectively take advantage of opening, but I like that I am making the decision to take a risk so I can own the consequences, rather than the AI randomly making those decisions without having enough information to make them well.

I think maybe what is needed is more abstract strategic orders. Something as simple as a 'try to engage ships of the same class' order could be helpful. But I think a more general 'try to move towards the center of the enemy formation' order vs a 'try to flank' order would be more useful (and also a 'retreat towards allies' order). In practice that could mean doing some sort of DP weighted clustering (or DP density calculation) to find centers of combat and basing some of the positioning logic on that. That also might achieve some of what the OP was thinking of: ships with the same orders (e.g. flanking) will tend to move towards the same areas and thus stick together. It also could be helpful to have a sense of where allies are concentrated using the same DP density/clustering to identify where safer areas of the battle are for retreating even without orders.

These would certainly help, but they don't really fix the key issue which is the default behavior. In order for this sort of thing to work, you'd definitely need more command points. And then you get closer and closer to RTS micro as you implement more and more of these sorts of things for edge cases. I'm not sayings its a bad idea or anything, just stating that these things aren't exactly the stated design goal of "orders are for special occasions and the AI is pretty autonomous" which I think is still the current intention.

I think the intention of that is is exactly the opposite of what you're saying. The idea is that the concept of moving towards or around areas of high enemy DP density is applicable in many different contexts, so you don't need to give lots of orders, maybe even only one order at the beginning of combat, as opposed to right now where you need to regularly un-assign and reassign orders on specific ships to try and achieve similar behavior (which does work, but is pretty annoying). The idea is to try and abstract the idea of 'flanking' and 'engaging' away from specific ships so you don't need to worry and babysit the details of specific ships.

Even if you don't implement those orders, I think giving the AI that sort of information about DP density, can let it make better decisions autonomously. That's the sort of information that gives it some idea of the context of battle rather than just its immediate surroundings, and is a step in the direction of identifying when chasing a frigate might be taking it away from a more important area of combat etc.

[haibane]

@Hiruma Kai

About squadron AI - yes, this is the assumption that we need another overseer AI. However, I believe you view of what is required from such AI is incorrect.

Most people when imagining AI commanding multiple units default to RTS full control style. As I already mentioned, individual AI behavior is reasonable enough. The only thing we need is to push it toward cooperative behaviors. In a sense such squadron AI would have as much control as a player: it automatically put hints on things and let ship AI decide what it wants to do. Those hints change action prioritization by AI, i.e. feed into existing mechanism, but don't necessarily force it do something specific.

With respect to implementation difficulty this approach is a lot more surmountable if you limit it to three behaviors I prescribed.

• Assisting can be achieved by AI simply picking a number of nearby enemy ships as high priority targets. There are nuances (danger, current status, distance, ability to disengage etc.) but those can be worked out.
• Flux rotations can be as simple as putting "retreat, you stupid" command on high flux ships which are engaged. Such a ship will prioritize protecting itself and moving away from enemy, as well as prompt nearby ships to open space behind or move between target and the enemy.
• Staying together is probably the most problematic one, even though it sounds the simplest. The base is straightforward: there is a squadron center, so ships are disincentified to move too far. Center can stay in the vicinity of squadron leader. However, we will need to factor in size and mobility to prevent weird cases, like frigates cuddling together in the back. Another problem is which ships should form frontline. This can be guessed, but sometimes it will be wrong, so probably passing the decision to player is the right choice. We can say, by default there is no prioritization (whoever happens to be in front is the lucky one), but there can be officer skills which prompt officer/lieutenant ships to engage first or last. This also solves second half of rotation, where squadron need to decide whether rested ship should be sent back to frontline (for ex. slow ships once retreated might just get blocked out and sit in the back if not given a priority).

This is certainly a pain to implement, but far from being impossible IMO.

About player interaction. Player can be treated as an officer just like now. It means player gets its own squadron and can configure it however he likes. Few heuristics can be used to keep cooperative behaviors working when player manually pilots:

• Squadron center can be rubber-banded behind player's ship. It will cause squadron to loosely follow the player.
• Assist part of AI can work as usual. The only change is that current player's target is always focused, so other ships can provide assistance.
• Flux rotations can be based on heuristic: player moves away from enemy on high flux -> cover him up. And in general get out of the way when player moves. With respect to other ships it works as usual.

It is possible to have squadrons changes inside battle in this system. This can be useful in face of losses/reinforcements/partial squadrons. Semantics: moving a ship to new squadron submits it to new chain of command, i.e. it behaves like a ship without officer/lieutenant. We can forbid moving player's ship into other squadrons and force player to add ships to his squadron instead - which has clear semantics without causing conflict with manual piloting.

There is no interaction between ships in different squadrons, so basically all such interactions will look roughly the same as they are now.

And yes, there is nothing else to it. No complex decision making, marking battlefield objectives or such. It is player's job. Job of squadron AI is to babysit a bunch of ships and make them better work together.

Alex has stated that part of the reason the AI is cautious is so that the player gets to do the heavy lifting in a battle.

This thought did cross my head. Current battle system likely originated from attempt to combine fleet on fleet combat with manual piloting. You literally cannot control you fleet while piloting, so AI is designed to be more-or-less autonomous with long-lasting hints.

This is also the point where opinions diverge. There are players who enjoy wrangling fight in such way, and there are also players who don't care or don't want to do it like myself. I view the combat from strategic perspective where it's quite bland and want to see the improvements. But there are also other players who don't care about complex strategies because it presumably subtracts from piloting aspect. This whole topic is certainly on the fence.


About decisiveness in general. The biggest problem with aggressiveness/decisiveness is contextual awareness. It is a complex topic, and humans are naturally very good at it. Which kinda runs into this persistent issue of AI not being smart enough. There are no simple solutions. Just decisively charging in is a great and almost certain way to lose a ship. I certainly understand devs erring on safer side because randomly losing ships to AI being stupid feels way worse than dragged out fights. The common way for player to address it is making super-OP-optimized ships using officers and s-mods, so they on average can push harder, further amplifying blob-oriented warfare.

That's another big reason why I want to have cooperative behaviors. It guarantees some level of both offensive potential (focus few targets + reasonable amount of firepower concentrated in one place) as well as defensive tools (avoid isolation + flux rotations) extending beyond single ship boundary. It means in general squadrons can afford to be more aggressive at committing to player's orders without risking individual ships as much.

The real reason is the tedium of rebuilding your lost ship

Alright, this is going to be another super-long-convoluted-controversial-post.

Spoiler

Just as people are arguing that ship restoration should more accessible, I can argue the other way around too. Losing ships and getting d-mods is very few ways you can suffer permanent consequences in this game. Simply knowing that I can take a skill or pay out a pile of cash to erase d-mods cheapens their impact. Ability to always recover ships cheapens impact of taking risks, bad decisions and difficult fights. There is no punishment here except bonus grind to quickly get you back where you were. It dilutes experience, erases choices, removes high and low memorable moments. I consider this an incredibly poor direction for the game.

Does this mean we don't have a problem? Hell, no, we have a huge one.

The TL;DR: losing ships is bad because Starsector manages both difficulty and progression very, very poorly.

Currently, primary way player works through the game is by making his fleet stronger. In response, game scales up encounters and difficulty to make it challenging. This is a prime example of linear difficulty system. But here's a problem: even linear games ostensibly struggle to properly tune the difficulty. Players are different. They have different experience and skill, adjusting game to be appropriately challenging for everyone is literally impossible. This is a huge reason why multitude of difficulty sliders become so popular with games featuring short session times, especially roguelikes. Look at Slay the Spire with ascension, Hades with pact of punishment and many, many others. Game starts slow to ease you in but then let's you tune the difficulty up to the point player feel challenged but not overwhelmed, while tempting him to go higher. Unfortunately, this is the best way of handling linear difficulty that I know of. Games with much longer and slower runs are just... out there trying something. All classical RPG games struggle with this. They may offer you difficulty levels, but more often than not those are a miss. Commonly, two players playing the same difficulty have completely different experience, or player finds that "medium" too easy and "hard" is too hard. And the worst part that without actually playing it's hard to say which bits are which.

Alright, you may think, let's add difficulty sliders to SS then? No. The real conflict here is that Starsector is non-linear game which employs linear difficulty.

SS is a sandbox, and it very hard to argue against that. You are thrown into the world with no direction and expected to do something, whatever it is. There is no structure to your adventure, you just choose whatever you like to do next. However, game expects you to always grow the fleet. This creates an irreconcilable conflict, where all activities are tied into or are measured in strength of your fleet (piracy, military, bounty) or exist for cosmetic purposes without any real gameplay or progression attached to it (trading, smuggling, scavenging, exploration). This is what creates a paradoxical situation, where an extremely memorable story event like losing a ship, in fact pushes you back on progression scale, so you as a player now feel behind difficulty curve set by the game and have to grind back, because there is no other content to interact with.

For a sandbox the only solution is to provide gradient difficulty, i.e. give player the spectrum of challenges and let him make the choice. So, for example, in bounty hunting, player is offered 5k for few frigates, 5M for huge pirate armada and stuff in between at every point in the game.
There is a number of advantages to this approach:

• Player is responsible for choosing appropriate difficulty and managing risks associated with it. He has only himself to blame for taking on mission which left half of the fleet crippled.
• Progression happens at player's speed. Stronger challenges happen whenever player feels like taking on them.
• Having impossible missions accessible sounds counterintuitive, but it is also a big boon. Player will remember that pirate fleet he was forced to pay "passage fee" five times. It creates a goal ("wanna kill them"), direction ("need to make fleet stronger") and sense of accomplishment when it is finally done. Giving player excuses to create goals is incredibly important in a game which doesn't have a directed experience. It gives emotional attachment, and otherwise player might feel directionless, lost and ultimately lose interest.
• Having weaker missions accessible also helps. It is safety net and provides means of measuring progress.
• It is also really good at fleshing out the world. Challenges from higher and lower difficulty almost exist in its own facet of world, independently from player. Moreover, gradient decouples player progression from world progression, further disrupting the sense that world revolves around the player.
• There can be multiple difficulty gradients, representing different ways player can progress through the game. Fighting, trading, smuggling, scavenging, etc. can coexist without forcing player into single style of play (e.g. stronger fleet). It also helps with setbacks, as losing progress on one axis doesn't mean other activities are affected to the same extent. Meaning recovery from a loss can even become a fun process, as you have to take a break and interact with other game parts.

Progression in such difficulty system is not about modifying the world - in fact world can be entirely static! - but rather about player being able to engage with bigger part of it.

There are some downsides to it too:

• Balancing still is a must. There is more leeway as player's situation is out of the way, but risk/reward still should be at least reasonably weighted, including between difficulty levels and between different axes of progression.
• Challenges at different points of a given scale must coexist within the sandbox without breaking it.
• It might be difficult for the player to find correct challenge to interact with because there are always some which are too difficult and some which are too easy.
• Progression in such world may feel flat. You can observe future challenges so working towards them may feel like climbing stairs for the sake of climbing stairs.

The way to fight the last two is also well known. By controlling what information about the world player has access to, game can control how player interacts with the world. I already mentioned this in a neighbor thread, but there are two simple mechanisms which can achieve what we want: news and search.

News do three things:

• Expose player to world events to which other actors may react and vice versa, deepening the immersion.
• Give rare/profitable/restricted opportunities to spice up gameplay.
• Introduce player to activities he can do by giving prompts.

Newsletter is a persistent background noise which always try to creep into player's attention. It contains things we want player to be aware of: illusion of working world and what ways he can interact with the game. Latter is especially important for newbies, as this is likely to be the first point at which they are given some direction.

Search does multiple jobs:

• It decouples what player knows from what game knows. Sounds weird, but it is extremely important, as it allows the game to change things while player is not looking. The simplest example of this can be a pirate you got bounty on moving from system to system in search of a ransom victim. Without search player might get instantly lost unless game spells it out. However with search game can be sure that player always knows how to get back on track. Application potential is immense.
• Search allows more of player actions to be intentional. You move from place to place not in blind search, but because something you are looking for is there.
• Search can inform player of availability. You don't need to run across whole galaxy to know that there are no bounties you might want to take, saving player's time and effort.
• Search can hide information in plain sight. Assuming search can give you info about anything in the game there is just too much information to brute force through. Player have to ask intentional questions and make sense of answers.

And before you start arguing that I'm trying to turn the game into god-knows-what, Starsector already implements very restricted versions of both. There are already bounty, "exploration" and trade convoy notifications which are just blunter, gamey news. There are also features which substitute search in a sense that they expose information which player might need access to. For example, bounty contracts tell you where to find the victim, goods tooltips tell about highest and lowest prices. Still there plenty of cases where it is not sufficient. How many times you looked at fuel prices in tooltip to realize it isn't helpful, because travelling to any of those places will actually cost you more than spending extra 5-6k on buying overpriced fuel nearby? Or how many loops you make through core sector visiting every single planet to find (choose any): the weapon, the hull, the bar mission?

There is definitely a whole direction to further develop handling and presenting information to player.

For example, we can figure that do-illegal-things don't get to news and official job board. But it doesn't have to be given though contracts/bar exclusively. Pirates may run a DarkNet which is well, also news+search+job board just filled with illegal stuff. Obviously unmoderated, so you can pay money to shitpost news feed, just like everyone else.
Why stop there? Let every faction have their personal in-faction news + search + job board. Suddenly, being commissioned is not just about painting colors and stipend but access to another information network which can change gameplay drastically. Having contacts from different factions is now incredibly useful as they can provide limited access to their in-faction intel and jobs instead of random generic "hit-me-up"s.

Another direction we can explore is public dossiers. It just tracks your (and others) activity: trade deals, places seen at, fleet composition, latest space activity. But, it would be weird to be attributed to something you did with transponder off unless there is a really good evidence. We can imagine that transponder also keeps an automatic fleet log while on. Suddenly, transponder stops being a "smuggler mode toggle" but an important tool used to shape your relationship to the world and its inhabitants. It can naturally creep into and interact with other game parts. For example you can approach another fleet and offer to exchange fleet logs, getting access to information before it goes public at their destination port. Or have a difficult decision after seeing distress signal being logged, because you are now legally obliged to answer (even if you suspect it's pirates) or face a huge fine and blemish on the record.

We can go deeper but I'm standing by my words that even having simple well-formed public news and search-anything-button is already a huge improvement.

[close]

There is probably something else I wanted to say, but I'm tired of typing this. Thanks for coming to my TED talk.