Return to “General”

Post

Re: Hey, it's an LT related question! :o

#16
I can only confirm what DD says about racing games. The rubberbanding-method was there from the dawn of times in this kind of games (and I played a lot of them, so I should know). While in the beginning the games were so simple that this, if obvious, wasn't much of a problem, from the moment in time where player performance wasn't dependable of four keys in a keyboard or two buttons and the cross in a pad anymore (as in, steering wheels being the proper way to play the games and the first simulators with physics built in the engine, proper setups, etc. started to appear) this method becomes an annoyance to most players. Even an insult. When the performance of both players and AIs is dependable of an ever increasing amount of parameters, using a simple turbo-boost to push the worst positioned and/or an inexplicable engine-debuff for the top drivers, you're throwing everything to the trash can. Why the hell do you make an engine with almost perfect physics for everything, from weather to track conditions, to tyre behaviour, to tarmac properties, realistic setup possibilities (nowadays there are games with the same telemetry apps than real reacing teams use) and so on if you can't make believable AIs with pros and cons, strong features and weak ones, with flaws as human drivers so that proper racing is the only way to play the game without fear of being boosted or capped for the sake of fun and entertainment? Well, there are such games. They're called racing sims and not racing games. Telling the ones from the others is important to not get frustrated. Of course, as concluded by Jensen and Grønbæk, the best course of action is let the player decide how much of sim and arcade you want for your experience.
I have been - and always shall be - your friend.
Post

Re: Hey, it's an LT related question! :o

#17
Lum wrote:I can only confirm what DD says about racing games. The rubberbanding-method was there from the dawn of times in this kind of games (and I played a lot of them, so I should know). While in the beginning the games were so simple that this, if obvious, wasn't much of a problem, from the moment in time where player performance wasn't dependable of four keys in a keyboard or two buttons and the cross in a pad anymore (as in, steering wheels being the proper way to play the games and the first simulators with physics built in the engine, proper setups, etc. started to appear) this method becomes an annoyance to most players. Even an insult.
That is absolutely not what I said. In fact it's almost counter to what I said.

Rubberbanding is necessary for fun between two players (either human vs. human or human vs. AI) that are anything but exactly matched in skill, and even the games that you think don't use it use it. As you've shown later in your post.

What is an annoyance to most players is having opponents that are either too easy or too hard. Which is why games use rubberbanding in the first place.
Lum wrote:When the performance of both players and AIs is dependable of an ever increasing amount of parameters, using a simple turbo-boost to push the worst positioned and/or an inexplicable engine-debuff for the top drivers, you're throwing everything to the trash can.
Woah there. Nobody mentioned using a turbo-boost or an engine-debuff. There are lots of ways to do rubberbanding.
Lum wrote:Why the hell do you make an engine with almost perfect physics for everything, from weather to track conditions, to tyre behaviour, to tarmac properties, realistic setup possibilities (nowadays there are games with the same telemetry apps than real reacing teams use) and so on if you can't make believable AIs with pros and cons, strong features and weak ones, with flaws as human drivers so that proper racing is the only way to play the game without fear of being boosted or capped for the sake of fun and entertainment?
Question: How can you tell how good the player is at the start of the game?

Answer: You can't! The best you can do is ask them, and unless you give them a hundred difficulty options they're going to struggle to find one suitable for them. So as they play, you have to improve AI performance when they're behind and reduce it when they're ahead. Otherwise, the AI cars either speed ahead of the player or lag behind them.

This is rubberbanding.
Lum wrote:Well, there are such games. They're called racing sims and not racing games. Telling the ones from the others is important to not get frustrated. Of course, as concluded by Jensen and Grønbæk, the best course of action is let the player decide how much of sim and arcade you want for your experience.
This is especially funny because racing sims typically have rubberbanding AI. Gran Turismo? Yup. rFactor? Yup.


And let's assume you've crafted this amazing AI that somehow can instantly assess a player's skill level before they even start the game, and provide the perfect competition.

... What are you going to do about multiplayer? You can't make players play better or worse.


I used to play OutRun Online Arcade. I don't anymore, because nobody else does. I typically preferred catch up on, even though it was pretty draconian (and engine-boosting), and even though I was the one typically in the lead. Why? Because otherwise it was a single player game for me. I'd drive off, and then never see any of my opponents.

And that'd be fine, except in single player I get to finish races. In multiplayer, they quit. Why do they quit? Because there is no competition. Without competition, the game is not fun. Without fun, why bother playing?

I'm going to quote one of the papers cited in the previous article, although I no longer have the original to hand (only the quote). It used "heart rate scaling" as the catch up mechanic in a competitive cycling scenario:
Heart rate scaling should increase competitiveness of players with radically different fitness levels while not negatively impacting the competitiveness of players of similar fitness levels. When examining the results of all races, we saw no significant difference between the conditions. When we separately analyze close versus “blowout” races, however, a more interesting picture emerges. In blowout races, HR scaling enormously improves competitiveness, while in close races, it actually worsens competitiveness. This implies that HR scaling might be applicable when there is a large difference in players’ ability, but should perhaps be avoided in cases where abilities are similar. Interestingly, the speed difference for the HR scaling condition is similar in both the close race and blowout race clusters (7.8% versus 8.0%).

Together, these results suggest that HR scaling is limited in how close it can bring players together, but is a highly promising technique when players’ capabilities are significantly different. It is interesting to note that races fell into either close (up to 4% speed difference) or “blowout” (over 15% difference) categories with no races in-between. We conjecture that when races are close, players are motivated to try harder to close the gap. This is shown by the statements made by some of the participants. For example, “it (the game) was demanding but it depends on the person and how much they want to win” and “it (the game) is powered by competition, so there is no limit. It depends on your competitor, because if they are more competitive than you, then you are going to go and try to get above him.” It is possible than when the speed differences are higher, players perceive the situation as hopeless, and therefore do not make an extra effort to catch up.

Engagement scores were similar between the two cases. No significant difference was found. (Power was low, indicating the possibility of a type 2 error; however, the difference observed between the two cases is very small.) This suggests that participants did not find HR scaling to be jarringly unnatural. This is a positive result, given that (as is suggested in figure 3) HR scaling does behave quite differently from speed computations based on pure player power. This is reflected by player comments, such as “I didn’t really notice a difference,” “I didn’t really see a difference,” and “they were both ok, no difference.”

However, in some instances of “blowout” races, where a player lost by a significant amount in the standard case, the losing player identified a preference for the HR scaled condition. For example “I like the second one (the HR scaled condition) better, because it was more competitive so I had a better work out” and “I think the second one (the HR scaled condition). I’m not entirely sure what the difference is; it seemed to be responding a bit more.”
TL;DR: Games between players with different skill levels are provably more fun for both players when catch-up mechanics are enabled within the game; certainly in the case when HR scaling is used as the catch-up mechanic (which is unfeasible for most consumer games). The paper I cited first makes the case for other forms of catch up. Most racing games have catch up; very few racing games have players complain about this catch up.
Games I like, in order of how much I like them. (Now permanent and updated regularly!)
Post

Re: Hey, it's an LT related question! :o

#18
So, how does all this relate to Limit Theory? Do we really want there to be these systems operating in the background while we play various aspects of LT?

In what cases should this system be used in LT and when would it be better not to use this system?

Perhaps it is possible to have technology and ship designs vary based on what the player has designed, while actual combat or placement of structures would not depend on a player's skill in this respect.
Image
Post

Re: Hey, it's an LT related question! :o

#19
None of this should relate to LT. One of the emerging pillars of the game was that the AI players would have the same tools and opportunities as the player. AI players would differ from each other in their goals, and in their skill in meeting those goals. There shouldn't be any "artificial" weakening or strengthening the AI because the player is too far ahead or behind. You're in an infinite universe; if the competition is too fierce in one region with particularly savvy AI opponents the player can always just migrate away and find a new region to build in.
Post

Re: Hey, it's an LT related question! :o

#21
I think this whole rubber-banding thing is STUPID. :D When I first played I-War 2, the AI absolutely murdered me on the first combat mission. The sheer speed and ease with which my ship was separated into its component molecules left my jaw hanging. I immediately fired up the "Instant Action" mode and spent a couple months :monkey: practicing combat. I was then able to beat that first combat mission, and quite a few after it. Did that mean that I never died again? Heck no, I must have died a hundred times. :mrgreen: What I loved was that the AI showed no mercy whatsoever. If you weren't good enough to beat it, then you got beat. The game hooked me, and is still my favorite game ten years later.

Now let's take World of Warships. Their method is to increase or decrease the amount of damage your shells will do upon hitting a ship, in order to "balance" everything, and keep it "competitive" for every skill level. And I HATE IT. How am I supposed to get better if I can't fight against the super-pro-league-mad-skillz players and get murdered by them? If the game takes their normal shell penetration damage (say 4450 for the USS Iowa) and drop it to a third of that so that I can compete with my poorer skills... where's the fun in that? When I were a wee little lad (:ghost:) and wanted to play a chess game with my dad, I'd get REALLY cranky if he deliberately did anything to permit me to win. I told him that losing because I was outwitted was more fun that winning because he let me. I apply the same rule to games.

On a side note, I've died a few hundred times in Ori and the Blind Forest and gotten nowhere near the finish. :squirrel: It really inspires you to MASTER the game. And finally getting past a spot that's instagibbed you ten times in one minute is immensely satisfying.

--IronDuke

P.S. Actually, WoWs will also play around with the dispersion of your shells, and one pal of mine will repeatedly hit 100mm armor with 406mm shells that have 867mm of penetration - and the shells bounce off. The only way that I'm able to do well in the game is by noting the specific areas where the "balance" rules do not apply. (Overpenetration does precisely 1/10 of max damage, so that's a guaranteed method of some damage; the overmatch mechanic means that hitting thin enough armor guarantees high damage returns, blah blah blah.)
Knowledge is Power, and Power goes in Cars.
I-War 2 thread
Epic Limit Theory Limerick
Post

Re: Hey, it's an LT related question! :o

#22
Rubberbanding came up, and I'd say it's relevant to "how fair will the LT economy be?" in that it's the computer tweaking some game results to try to maximize player fun.

In that respect, both of these share a couple of similar characteristics: 1) It's hard to do well, and 2) No matter what, someone will complain about it. ;)

Let's look in particular at this statement:
bkdevil wrote:There shouldn't be any "artificial" weakening or strengthening the AI because the player is too far ahead or behind.
I'm not picking on you, bkdevil; this is just a good entry point to the question: how do you know when a simulation of real-world processes is good enough as game content?

Is a faucet/drain economy, which just creates resources on the fly as needed by the player character and destroys them when traded to an NPC, good enough for Limit Theory? That's a popular way to build in-game economies because it satisfies most of the action gamers while being relatively quick to develop and hard to break. But gamers who enjoy exploring systems, who like poking them with a stick to see what happens, are bored by faucet/drain economies and want something that simulates more real-world economic principles.

So if Josh built the LT economy using the faucet/drain model, might some players consider that an artificial strengthening of the economic AI of NPCs? After all, it would basically let them "win" an economic game as needed to provide different levels of challenge for the player.

Or would most of the likely players of LT be satisfied with a highly-simulated economy that does not give the player's character any special treatment compared to NPCs? What if that occasionally means any character, including NPCs, and including your character, may sometimes completely lose everything and have to start over from nothing?

In other words, is any amount of economic rubberbanding unacceptable in Limit Theory?
Post

Re: Hey, it's an LT related question! :o

#23
OK I'll bite.

Having read the various links on the theory of game economic modelling, I've come to the conclusion that perhaps I don't want a fully simulated economy. In that I mean I like the idea of Dwarven Fortress, the extent to which the smallest detail is modelled, and I enjoy reading other people's account of their progress through the game, but I've found that even after downloading and tinkering with the game, I've never actually played it.

I can see why a faucet / drain economic model is attractive. I would argue that our reality can quite accurately be described in precisely that manner. The faucet is the Sun (the Monty Python guys weren't kidding when they said it's the source of all our power). We may be burning fossil fuels at the moment, but fossil fuels are merely solidified solar energy. The drain is the Laws of Thermodynamics. Everything we do produces waste heat that is unrecoverable.

(Terrestrial nuclear power may upset this model, especially if / when fusion gets off the ground, but I'll choose to ignore that for now. I won't even think of glancing at the more exotic energy sources that are not even a twinkling in the mad theoretical physicist's eye.)

The modelling of our economy then becomes a question of choosing the level of simulation (probably not adding much to this discussion but it's sometimes handy to take an assumption to its logical conclusion).

As for rubberbanding - in an infinite (or near infinite) playspace, if the heat gets too hot, get out of the kitchen as they say. If an NPC is too powerful, then migrate away. So long as the player has the option of retreating to safety, and has the capability to harvest and directly use raw materials, then there is always a chance to outlast the most formidable NPC.

There was a time when the Roman Empire must have seemed eternal, and that was in a time when the full extent of the Earth was unknown. But the Roman Empire eventually got too big to manage, and it crumbled. It would be tricky to build such emergent behaviour into simple economic rules, but that would be a nice thing to see. That's about as much rubberbanding that would be needed.

So rubberbanding that is internally consistent is totally fine. All I ask is for fairness (this is a game after all, not real life).
Post

Re: Hey, it's an LT related question! :o

#24
2) No matter what, someone will complain about it.
I couldn't agree more with that! :D
So if Josh built the LT economy using the faucet/drain model, might some players consider that an artificial strengthening of the economic AI of NPCs? After all, it would basically let them "win" an economic game as needed to provide different levels of challenge for the player.
I'm sure some people would consider the whole faucet/drain type model as unacceptable and the reason you state would be amongst the complaints. Personally I don't have a problem with that system, and since Josh has stated that the asteroids will repopulate over time I think we can assume the system will be something like faucet/drain. I wouldn't consider it a type of rubberbanding. As long as the player and AI have the exact same options inside whatever economic system is in place I think it is all above board. The player can always travel around and find easier markets or more plentiful resources, but so can the AI.

I don't think I fully understand what you mean by the second part of that quote, flat. Maybe you could elaborate?

*edited formatting
Last edited by bkdevil on Wed Dec 21, 2016 2:34 pm, edited 1 time in total.
Post

Re: Hey, it's an LT related question! :o

#25
As for rubberbanding - in an infinite (or near infinite) playspace, if the heat gets too hot, get out of the kitchen as they say. If an NPC is too powerful, then migrate away. So long as the player has the option of retreating to safety, and has the capability to harvest and directly use raw materials, then there is always a chance to outlast the most formidable NPC.
Totally agree.
Post

Re: Hey, it's an LT related question! :o

#26
Great responses -- thank you for an enjoyable conversation.

bkdevil, as a quick answer to your question, a definition: a fair game is one that any player is theoretically capable of winning.

That's not too hard a condition to satisfy in a well-designed board game, where all rules are enforced by the players themselves. But what happens when "the game" is a very complex software simulation? How do you know it's playing fairly? If you can't win at the highest difficulty levels, is that game "fair?"

And the point I was making is that, if you program a game to watch the player's performance and dynamically adjust the difficulty so that the human player has a fair (or better) chance of winning, this inevitably opens up the game's developer of being "unfair" -- even if the whole point of fiddling with the level of challenge on the fly is to allow the player a fair chance to win.

But at the same time, if the developer decides to make the game absolutely fair, completely ignoring whether you-the-player are getting clobbered Every Single Time, on the theory that players want to be assured that the game isn't cheating by dynamically altering its difficulty, is that fun? If that somehow is fun for some people, do those people constitute the majority of a game's players? Really?

So, back to Limit Theory, we can see that this is actually a very old question in game design: which do you favor when they're in conflict, simulation or mechanics? In other words, when there's only one way to implement a feature, will you tend to favor simulation even if it means many players will lose, or will you favor mechanics that insure a good chance of winning even if it means many players will quickly be bored?

Bugbear, I think this is a response to your comments as well. ;)
Post

Re: Hey, it's an LT related question! :o

#27
Re faucet-drain economy:
I see the appeal, but i cant see any way to make it work with the fully dynamic world of NPC stations being built and destroyed witb the prices at least somewhat adapting to changed circumstances.

With a fixed faucet-drain economy thats not exactly doable.

Also: how does it react to new devices being developed and built?
How to include those in an at least minimally immersion preserving way?
Post

Re: Hey, it's an LT related question! :o

#28
Flatfingers wrote:which do you favor when they're in conflict, simulation or mechanics?
My hope for LT would be that in a universe that is infinite this conflict shouldn't arise. I would not expect to be able to jump into any system in the universe and "win" in the sense of my business ventures succeeding in that particular location. However, I do expect to be able to jump into some system and be successful. I say, make the simulation as good as possible, and give the AI the potential to be as crafty as possible. In a heavily populated system I may run into opponents that are far beyond my skill level to handle, but I should be able to travel, and at any skill level, find a system either sparsely populated, or controlled by less proficient NPCs, where I can "win".
Cornflakes_91 wrote:I see the appeal, but i cant see any way to make it work with the fully dynamic world of NPC stations being built and destroyed witb the prices at least somewhat adapting to changed circumstances.
Maybe I'm not entirely understanding the definition of a faucet-drain economy. I thought in LT the most basic resources would regenerate (although I realise nothing is certain), and that would constitute a faucet-drain. How would this break down in what you are talking about above cornflakes? Perhaps I'm totally ignorant about what we're talking about. :)
Post

Re: Hey, it's an LT related question! :o

#29
bkdevil wrote: Maybe I'm not entirely understanding the definition of a faucet-drain economy. I thought in LT the most basic resources would regenerate (although I realise nothing is certain), and that would constitute a faucet-drain. How would this break down in what you are talking about above cornflakes? Perhaps I'm totally ignorant about what we're talking about. :)
Well, Flats usage of "faucet-drain" strongly implied it being the polar opposite of a dynamic market.
With prices being (mostly) fixed/hardcoded whereas in a dynamic economy prices are only defined by supply and demand.

And i cant see how prices in a faucet-drain economy can be defined.
At least without the determining algorithm being a dynamic market again.
Post

Re: Hey, it's an LT related question! :o

#30
Cornflakes_91 wrote:Flats usage of "faucet-drain" strongly implied it being the polar opposite of a dynamic market.
With prices being (mostly) fixed/hardcoded whereas in a dynamic economy prices are only defined by supply and demand.

And i cant see how prices in a faucet-drain economy can be defined.
At least without the determining algorithm being a dynamic market again.
It's not really a binary option. Even a faucet/drain economy is "dynamic" in the sense that it creates resources at particular prices in response to player demand.

What IMO really distinguishes a faucet/drain economy from a true simulated economy is that the primary measure of success in the former is not price, but total wealth.

By "success" I mean from the point of view of the operator of the game: a successful faucet/drain economy is one that's at the peak of the bell curve between not enough wealth in the world and too much wealth. You want the total amount of wealth -- primarily meaning resources, and secondarily meaning money -- to be maintained at a particular point, relative to the number and experience of the game's players, that you maximize the desire of players to keep playing the game.

Any game economy needs to do this; a faucet/drain economy does it by controlling the algorithms for magically generating new resources and by creating enough (but not too many!) good reasons for players to want to destroy wealth. Controlling the creation and destruction of wealth indirectly controls the total amount of wealth in the world. As long as the algorithms are good at keeping total wealth commensurate with the number and level of players, the game's economy becomes less of an excuse for players to stop playing.

The difference between this faucet/drain economy and a simulated economy is that, in the latter, the raw resources of the universe exist independently of character actions, and created goods require character actions. If an NPC vendor doesn't have some piece of cheap basic gear, then he doesn't have it. He won't magically generate it to be able to sell it to you at a floor price fixed by the developer. The only way he will have that item in stock is if some system in the game causes that item to be created, and the NPC vendor obtains that item and chooses to sell it to the public for a profit.

So in a simulated economy, there are multiple measures of success: 1) general price stability, 2) availability of desirable items (i.e., generally moderate prices for most goods), and, in strongly simulated economies, 3) stable labor costs. The main metric for a game with a simulated economy is price stability: neither ongoing inflation (increasing prices due to an insufficient rate of goods creation or excessive money creation) nor deflation (usually due to goods being too durable, or to not enough money being created). Avoiding inflation and deflation (the latter of which in online games is sometimes referred to as MUDflation) is important in a game for the same reason it matters IRL: both inflation and deflation reduce the actual amount of economic activity that can be realized from its theoretical maximum.

It's actually more complicated than this. In some macroeconomic theories, a little bit of inflation is desirable as it's believed to increase incentives to work and produce. But that brings me to my final point, which is that the complexity of managing an economy in which goods must actually be created by characters in order to exist, and in which the supplies of money and labor affect prices, is so much greater than that of a faucet/drain economy that virtually every developer who actually ships a game with an economy uses some version of faucet/drain. That's not to say a mostly-simulated-economy game is impossible -- it's just so hard to do well that developers who try it usually wind up going down a multi-year rabbit hole of economic theory instead of actually implementing a complete game.

Which brings me (finally!) to Cornflakes's comments: the fact that Game A can allow a few forms of "magical" resource generation and still be mostly a simulated economy, and Game B can include character production of wealth-items (crafting) and still be mostly a faucet/drain economy, just means that these are notional endpoints on an analog line. Most real games with an economy just implement faucet/drain for expedience, but they will maybe slip in one or another element of a simulated economy to keep things interesting.

From the sound of it, Josh may be trying to build Limit Theory from the other direction -- a mostly simulated economy, with some magical resource generation (e.g., mineral resources in asteroids) where a truly closed economy would be too much development work for not enough long-term player fun.

No idea whether that supposition is accurate until/unless Josh tells us so, of course, or releases Limit Theory itself.

Online Now

Users browsing this forum: No registered users and 1 guest

cron