Suggestions for a patch for Spaceforce

Tue Jun 19, 2007 6:47 am by breslin

I have a little sympathy for the "matter of taste" argument. For instance, I don't really like X3, but clearly it's not a bad game. I dislike X3 for the common reason, that it feels more like a full-time job than a game. But that really just boils down to a matter of taste.

Still on X3, people almost universally agree that the main story sucks. That's not a matter of taste, but an objective fact. Now, it's possible to get some fun out of the story nevertheless, but only a 12-year-old could mistake this for a finely-crafted narrative.

Taste in games is much like taste in anything else: "taste" means personal preference, and doesn't reflect on the quality of the product. "Quality" is an entirely separate category.

The general complaints against SpaceFarce do not revolve around matters of preference, but matters of quality. In other words, it's not that I don't like the abstract design decisions involving navigation, but the implementation just wasn't well executed, leaving much to be desired. It's not that I have a person preference -- a "taste" -- in favor of another way, but that their plan was executed badly.

As I mentioned about the "Tarr Chronicles," one might have preferred different design decisions, but one cannot fault them for execution.

So if people say "the main difference is that FL was good," they might mean that FL was in line with their personal preferences, but I think it's much more likely that they mean the design was well executed.

Following this thought, there's two almost-separate concerns in game design, which break down into "theory" and "practice." In the first (theory), the goal is to theorize what sorts of game design people will be most likely to enjoy (this considered alongside what is possible given current technology). In the second (practice), the goal is to take a design goal and invent ways of implementing and presenting it.

While one can certainly fault SpaceFarce on grounds that it reflects bad design decisions from the top, this we can be sure is debatable, and if one chooses, one can dismiss the entire discussion, saying (albeit rather hyperbolically) "take 100,000 people, and you'll get 100,000 preferences."

By contrast, it's really beyond debate whether the designs were well implemented.

It's for that very reason -- the obviousness of key problems -- that an open discussion with the developers is absolutely necessary. It's no good for me or you or anyone else to make a suggestion to fix something that's obviously broken.

An open discussion with the developers is also necessary because the severity of the implementation problems obscures the higher-level intention behind them. (In other words, the severity of the problem obscures its cause.) For instance, there's clearly something wrong with combat. The guns overheat too fast, the initial default fire-style is set wrong, the very first battle is much too hard, considering it's an introductory mission and should be geared towards first-timers. But what *exactly* is wrong with combat?

There are a number of possible designs involved, and we need to know which one(s) before we can make useful suggestions. For example:

1) is combat *supposed* to be much harder than conventional space-sim games? This would be a valid decision, if that's the intention, and my suggestion would begin with the proposition that one must carefully introduce the difficulty to the player, and gradually amp it up rather than hitting them with great difficulty all at once; perhaps get some assistance from a wingman during the first battle, or nerf that one a bit in any of various ways. And so on.

2) are the guns *supposed* to overheat a lot, requiring short bursts rather than continuous fire? This is also a valid design decision, if intentional, and my suggestion would be that enemy shield strength (and perhaps enemy damage) drop at least for the first mission, and the default fire-rate should be switched. Another idea: the first guns should have a more conventional fire-rate, and the more powerful guns will have the short-burst we're aiming for. Then the player can sort-of "opt-in" to the handicap, considering it an overall gain, rather than feel immediately frustrated by it. And so on.

3) is combat supposed to be more-or-less conventional? If so, do the standard work of game balance. This will involve balancing various factors, probably starting by doing the following: decrease the player's gun's overheating rate, decrease enemy shield strength, and decrease enemy maneuverability, at least early-on. And so on.

As it stands, the developers refusal to talk belies their pretended interest in suggestions. It would seem that the whole "suggestion" thing is a ploy to divert/redirect well-deserved criticism.

----

Newsflash: the only copy of SpaceFarce up for bid on ebay recently sold for $18.50. Considering that Amazon has been alternating between out-of-stock and trying to sell them for $29.99, I'd say the game isn't selling very well at all.

Edited by - breslin on 6/19/2007 7:49:05 AM

Tue Jun 19, 2007 6:30 pm by RGHawk2000

Joystick control, please make it work better with all the usual functions of today modern joystick, please, please pretty please

If you could make a full HOTAS control work that would be a dream for me...

As for X3, Im playing it again now and is not that bad, have you read the forums lately over there?? still pretty active with new ships, massive mods, and constant updates...As OP said, it is a matter of taste...

Anyway, I was pretty critical about the SF:RU game at first but the demo was ok for me. The graphics problem is more due to your card than the game itself. Since my video card is old I have to have the minimal setting for it, otherwise objects would start to bounce around the screen.

Wed Jun 20, 2007 12:39 am by SWAT_OP-R8R

hey RG if you want good joystick controls we still could get SL: Sol wars out ^^
even with ffb

Wed Jun 20, 2007 11:46 am by Chips

*wishes wishes wishes* I hope we can stay on topic

Would hate to have to delete posts that aren't O_O

Thu Jun 21, 2007 12:05 pm by breslin

(Apologies to Chips, as the following couple points will be a bit off-topic.)

As soon as I heard the "over 2000 missions" claim, I suspected something was a little fishy.

As a programmer, I can tell you that one would never write "by hand" 2000 missions, but would write a script to auto-generate them. This being the case, it seems really weird that Provox would auto-generate 2000 missions (or at least, the larger portion of them), but they wouldn't auto-generate missions at runtime, like space-sims conventionally do. Then, instead of having 2000 pre-auto-generated missions, you can have an infinite number of auto-generated missions. For the life of me, I cannot imagine the reason for this. So I'd like an answer to that.

I have heard a few soundbytes about AI. The only substantial point I've heard is that NPCs have a preference to fly between their enemy and the sun. As a programmer who specializes in AI, I am not terribly impressed by this alone.

There are a number of ways to set up NPC-AI flight-control, but they all essentially involve consulting a heuristic which determines the preferred trajectory. Adding a couple lines 1) to calculate the area of "sun-blind" space, and 2) to weigh this preference against the other considerations -- that's a fine idea, but it's no more impressive than any of the other considerations of the AI. (An AI programmer will be impressed by an ingenious system, not by a single rule or behavior within a system, however ingenious that single behavior may be.)

Now, it may yet be that the AI is well-written -- this is the kind of thing that's difficult to discern by experiment, principally because it's easy to "cheat," or in other words, to do things to make the AI appear more impressive than it actually is on a programmatic level. I just haven't heard any real evidence of this.

The current "Company Line" includes the following two lines about the AI:

Since A.I. plays a crucial part in SpaceForce, it will create a unique playing experience throughout the game. [...

Unique NPC personalities, each designed to make their own decisions based on circumstances.

The first point is basically advertising-language, akin to "Try Widget, for a fresher, more enjoyable experience."

The second point suggests that there's some variation in behavior between different characters or character-types. In other words, perhaps something like this: pirates fly more aggressively, going for a shot even if it means facing incoming gunfire, while UF fly more defensively, dodging fire until they have a "clear" shot. But again, this variation is nothing new. Making NPCs vary their combat-behavior is not significantly more difficult to set up than making ships vary their maneuverability.

I have heard a number of players complain that the AI is too hard, but I think what is meant is that the game-balance is badly adjusted. Increasing the numerical difficulty of enemies (their shield-strength, hull-strength, maneuverability, and so on) has nothing to do with the AI.

So if this is actually a remarkable AI system, I'd like to hear more about it. My suspicion is that this implication is just a marketing ploy. (As for example Derek Smart, in an attempt to market his game, falsely claimed that Battlecruiser 3000AD used "neural nets".)

==

Now of course this does in fact boil down to a concrete "suggestion": Provox should stop describing the game using vague celebratory/marketing language, and discuss the matter itself as plainly as possible. There's precious little hope for useful communication between parties, so long as one side is attempting to delude the other.

Thu Jun 21, 2007 12:06 pm by kuvacz

I like to keep it short (even when not possible all the time),

the whole problem about the much too early release of the game is
the chosen publisher JooWood (Austria).

They already messed up severeal games by pushing the developer
to release it earlier then planed, so many games published by
JooWood are nothing else than a beta version.

All who know the RPG "Gothic (3)" and the developer Piranha Bytes
know what I'm talking about, it was released about 1 year before
it actually would've been what it was supposed to be like.

Fact is most copies of a game are sold within the first two weeks
of its release, and as long as it's somehow playable and the publisher
will not have to refund customers for releasing a not playable game
they will continue this as time is money and who cares about tomorrow
strategy.

Actually it works, Gothic 3 was sold many times more then all previous
versions (which had a quite good support and great story) together.

So why should JooWood make a change ? I f it's a lot more lucrative to
have a game released 1 year earlier then it actually should be. In this one
year they can finance the production of another new game.

And there will allways be enough (new) players that just don't think about
the publishers part in the production of a game and will still buy the
games that are released by a publisher known for it's carelessness.

The ones that suffer from this politics are the developers and the customers
while the publisher makes big money in little time.

No reply needed

Cya Kuvacz

Fri Jun 22, 2007 2:23 am by SWAT_OP-R8R

but you know that jowood didnt push anything?
LR got information about the release date even before provox made the deal with jowood

it would be good if you guys get some more information before you claim something

Fri Jun 22, 2007 5:14 am by breslin

That's interesting to know.

It's a commonplace in the programming industry to underestimate the time required for "finishing" a project. -- It's so common that people make up jokes about it. "As a rule it will take ten times as long as you calculate that it will take, even if you apply this rule to the calculation." Ask any programmer: it's funny because it's true.

So if JoWood or any other external force is to blame, they're to blame for holding Provox to the deadline they set for themselves, not knowing when making the agreement that they were vastly underestimating how long it would take.

This doesn't explain some of the radically wrong choices they made (which have to do with design and so on, not execution), but it does explain the general "beta" quality of the product.

By the way, was that planned release date originally the 2nd of June (in North America)? I have a pet theory that they moved the date up as soon as they realized how bad the demo was, i.e., that it was going to significantly harm sales if they gave people enough time to play it, and form critical opinions of it. It would be nice to have this confirmed.

Fri Jun 22, 2007 5:55 am by H.Q. Victor

Any more acusations will be deleted, Period! This thread is for sugjestions of the game, not attacks on the creators.

Fri Jun 22, 2007 10:23 am by Thomas45

I would also like them to add in fly able batlleships

And mod tools....

And fix the missile bug which makes your ship and all other ships impossible to kill when you where meant to die with missiles are all around you.

Fri Jun 22, 2007 10:59 am by Chips

For the AI, isn't it possible they just have variables? How many friendlies, how many hostiles, condition of ship, strength of ship, amount of weaponry (mean missiles / ammo available) - and all these factor into the behaviour of the AI.

If it's damaged, or low shields, it's more likely to dodge. If its got full shields/good weapons and more firepower, it's going to concentrate on getting/keeping you in it's sights. If it's out-numbered, damaged or in an inferior ship - it's likely to leg it, if it's in numerical superiority, it's likely to fire until you start shooting at it, and then attempt to make you into a target for the others by leading you about?

Add to that the idea of hiding in the sun, which is neat

Anyways, isn't that likely to be the sort of thing they've done for "individual AI" - kinda "agents" basically, as in agent based systems etc?
Don't really know much about AI i have to admit

Should buy a book and read up on it more

Fri Jun 22, 2007 7:34 pm by breslin

Yes, Chips, very good explanation. Those are exactly the kind of things we'd look for in an NPC-AI. If anybody experimentally sort-of reverse-engineers the NPC flight-AI, I'd be very interested in hearing observations.

Come to think of it, I only noticed one single "dodge" behavior (and in fact I did not particularly notice a tendency towards the sun-blind technique, although I do not doubt its existence). I did notice a few interesting maneuvers, particularly the "hop" technique (for lack of a better term), but the typical enemy-flight-pattern remained pretty consistent, in appearance.

Maybe the dodge techniques increased frequency when the enemy was damaged -- it would require further study.

The way it's almost certainly set up, there's a set of specific states that an NPC can get into. (Something like: aggressive_attack, attack, evade, wild_evade, regroup, etc.) And there's an evaluation function called every once-in-a-while, maybe every second, maybe even more frequently, maybe varying depending on the current framerate, or even on the current difficulty level. The evaluation function looks at the current "variables" as Chips well puts it, along with whatever specific personality-preferences of the ship, and alters the ship's state as appropriate. So if you're trying to deconstruct the AI, you'll want to look for the identifying characteristics of the various states, and look for transitions between the states, hopefully figuring out what triggers them.

This is essentially how Freelancer works, by the way. There's a couple evasion states, which can be got into depending on a few factors, and a couple attack states. The NPC goes into heeavy evasion when the rest of his wing is destroyed; he goes into light or heavy evasion when he's being chased; he circles round and makes a run at his enemy when he exits firing range; what type of run (direct or glancing) seems to be random, by my estimation. There seems to be a couple (or more) maneuvers associated with each state. It may be slightly more complicated than this -- it's hard to say just from observation.

Anyway, I'd be interested to hear if the NPC-AI in SF is any more sophisticated than this. At present, I have no reason to suppose one way or another.

Oh, and yes NPCs are a type of "agent," for the purposes of AI programming (and terminology). If you're serious about brushing up on your knowledge of AI, the universally recommended book is: Artificial Intelligence: A Modern Approach, Second Edition, Russel and Norvig. You can check out the book's homepage.

Important Message

You are browsing the archived Lancers Reactor forums. You cannot register or login.
The content may be outdated and links may not be functional.

To get the latest in Freelancer news, mods, modding and downloads, go to
The-Starport