Return to “Suggestions”

Post

Twitch Plays Limit Theory

#1
This is a topic in response to an idea I posted earlier.
JoshParnell wrote:
AbhChallenger wrote:You know. That might be an interesting way to market LT. You know about "Twitch Play's Pokemon" Perhaps you could write a mod for your own game that has the "player" run as an AI. But twitch chat serves as its decision process. If most of the chat says "Mine epic ore #2 in system alpha centari" The AI character will do so. The goal would be to get this AI character as rich and powerful as possible before trolls eventually lead the character to its death.

It might actually work. Eventually Twitch beat Pokemon Red. Apparently Twitch plays Dark Souls managed to actually beat the first boss.
That sounds like a pretty fantastic idea actually. Crowd-playing! Definitely after release though :) And since the AI in LT is already supposed to act like a player...I don't imagine there would be much technicality behind it, other than injecting twitch votes into the decision process of the AI.
Basically the idea is sort of similar to that of https://en.wikipedia.org/wiki/Twitch_Plays_Pok%C3%A9mon However, the basic goal of the chat should be simply to guide the AI player to riches and power for as long as possible before the trolls or random LT events lead to the character's demise.

Personally I think this is a great way to market LT. People would watch simply to see what kind of craziness would happen from the chat. (Oh wow the trolls managed to jump the player into a pirate system with his shiny new battleship lol)

So if Josh happens to decide to do this after LT 1 is out.
  • * What features or limitations do you think should be part of it?
    * What kind of "Start" should the AI character have?
    * Should the universe be redone after every defeat?
    * Can you think of any special events that could be fun?
Post

Re: Twitch Plays Limit Theory

#2
As I threw at the IRC, I think each vote should help assign weight to specific AI preferences.
eg, Aggressive, Passive, Pirate, Trader, Fly into the Sun...

Each vote survives for one hour, so people have to keep spamming the thing they want to get the AI to do things. :V
This way the AI will do it's own thing.

I would like to see it instead of a video update or equivalent, while we wait for Beta. :D
°˖◝(ಠ‸ಠ)◜˖°
WebGL Spaceships and Trails
<Cuisinart8> apparently without the demon driving him around Silver has the intelligence of a botched lobotomy patient ~ Mar 04 2020
console.log(`What's all ${this} ${Date.now()}`);
Post

Re: Twitch Plays Limit Theory

#3
AbhChallenger wrote:* What kind of "Start" should the AI character have?
Randomised. Procedural generation is one of the selling points, and a randomised start would show this off to its fullest.
AbhChallenger wrote:* Should the universe be redone after every defeat?
Absolutely. Procedural generation etc.
AbhChallenger wrote:* Can you think of any special events that could be fun?
Not particularly; if anything that should be up to the "players". Keeping rare events rare will make it all the more exciting when it happens on stream, after all.

Now, I've saved your first question for last because it's a big one.
AbhChallenger wrote:* What features or limitations do you think should be part of it?
Here's the problem: if you're just giving fairly abstract orders to an AI, it's not going to hold anyone's attention. If you say, "Mine epic ore #2 in system Alpha Centauri" you're looking at a long wait while it does it, while in the meantime thousands of other people are shouting out orders too. You can't immediately cancel an order when a new one is given, because nothing would ever get done. And you can't wait for a voting process, because you'd lose viewer interest with the lack of visible feedback.

There's nothing wrong with long, slow streams. I do them all the time. But if you're fishing for meaningful interaction on a crowd level, it has to be instant (well, you can't get instant with livestreaming but it has to be fast), it has to be noticeable, it has to be small on the individual level. TPP worked because viewers were literally providing the controller inputs (barring the mode system).

Limit Theory, on the other hand, isn't particularly small on any level. You can't use controller inputs for example - it's just not going to work. Democratic decisions of abstract orders isn't going to hold viewers either.

Is there a solution?

Well, the fact that voting for orders is not going to appeal to the impatient everyman doesn't mean it can't or shouldn't be implemented - I believe it absolutely should. Keep track of all orders made in the past (say) two minutes, take the most common one and execute it. Only cancel the order when a majority of people are continuing to give different orders (maybe don't allow cancelling of the order until after two minutes).

BUT

There also needs to be a subset of orders that are faster, more concrete, and simpler. Commands like "attack", "defend", "retreat" are probably time-sensitive and although you'd expect the AI to do it themselves, you need to give control to the viewers too. I'd recommend taking these and requiring a lower voting threshold (or a lower time limit - execute if it's the most common order in the last 30 seconds).

Also, if the AI isn't doing anything particularly action-oriented, allow commands like "DO A BARREL ROLL" actionable instantly to keep viewers happy that they are actually interacting.
Games I like, in order of how much I like them. (Now permanent and updated regularly!)
Post

Re: Twitch Plays Limit Theory

#4
Maybe just have one universe running, and when an AI dies pick a random list of AI's and let twitch chat vote which one. Or just pick a random AI. That way you can see the repercussions of the last AI's actions.
Scytale wrote: Look, I know people suck. That's the whole point! My question is, are you going to be as horrible as the mob, or are you going to work to be better?
Image
Post

Re: Twitch Plays Limit Theory

#6
Yes, Cornflakes, I would much prefer the valuable time of the coding warrior to be invested in the game itself. If Josh has the time available for this peripheral stuff then he should let the graphics monkey out of his cage to provide a super shiny coating to LT. :thumbup: :thumbup: :angel:

Don't give him any ideas about how to avoid the less interesting (for him) aspects of the game. :P

:shifty: Perhaps graphics isn't the best example of what he could or should be doing. :oops:
Post

Re: Twitch Plays Limit Theory

#9
Cornflakes_91 wrote:i'd say scratch the "plays" in "twitch plays limit theory".

it would take time to design and set up a "fair" and reasonable control system and i'd rather have that time spent on the game itself.

perpetual twitch stream, yes, "twitch plays" , no
My idea is about marketing. Not development of LT 1.0

Yes obviously we want tons of new fancy stuff for LT 1.01 and beyond. However, To fund such fancy stuff the game still has to sell. Josh may indeed have to spend a little time away from the game code to get the game in the eyes of the public after the surge of game media attention fades.

Not saying my idea is the best idea for such marketing. It is just something to think about.
Post

Re: Twitch Plays Limit Theory

#11
That's like saying, "how is creating an advert for a game that airs on TV better for PR than just letting them buy the game itself?"

Twitch Plays Limit Theory would bring in a different crowd of people. And even at its worst, it's an extra LPer.

Issuing orders is already in the game; the extra work required is fairly minimal.
Games I like, in order of how much I like them. (Now permanent and updated regularly!)
Post

Re: Twitch Plays Limit Theory

#12
DigitalDuck wrote:That's like saying, "how is creating an advert for a game that airs on TV better for PR than just letting them buy the game itself?"
im not disagreeing with the perpetual commercial of the twitch stream.

im questioning the purpose in doing something thats functionally identical to rigging a random button smasher to the game.
which will lead to 99% of the time either aimless spiralling, confused paralysis or having the object of interest just dying in perpeuity.
DigitalDuck wrote: Issuing orders is already in the game; the extra work required is fairly minimal.
with the assumption that the orders arent coming in 200ms intervals and flatly distributed through the possibility space.

twitch plays pokemon worked at all because of the very limited moment-to-moment action space in the game.
and that was circular and death riddled for 99% of its time.
Post

Re: Twitch Plays Limit Theory

#13
Cornflakes_91 wrote:and.. where is developing a fork for letting people playing it on twitch better for PR than having people actually playing the game itself and producing lets plays and whatnot?
It is not an either/or situation. The part with twitch playing it is simply an addition. People enjoyed the challenge of Twitch Plays Pokemon and trying to get the character to victory to beat the trolls. That is why for a time. Only the international tournaments of MOBAs were beating it on twitch.

The struggle was interesting to watch. While of course LT is unlikely to generate the kinds of viewers as a well known game like pokemon. You have to admit it is far more entertaining then watching an AI slowly work his way up.

There will be plenty of let's players. Plenty of Twitch streamers. However, most of them will only upload and stream during the early days.

Let me put it another way. Even if I gifted LT to most of my contacts. I doubt many would even install it. However, if I showed them the Twitch Plays LT Stream. They might actually "play" it.

Online Now

Users browsing this forum: No registered users and 30 guests

cron