/sci/ - Science & Math

Is it rational to use decision theory/game theory for guiding personal investment/business decisions?

>>6763897
>I can't really see Game Theory having any systematic application in any science.
It's a social/soft science thing. It has tons of application in biology/ecology as well as psychology, economics etc. It is a very heavily mathematical subject and was originally grouped under math, it's just in practice the mathematical rigor is so far removed from reality that it only predicted anything useful in economics, it is very useful in dealing with simple animal or other nonhuman species behavior particularly in the aggregate as well though.

>I would suggest heading over to /biz/ but that would quite possibly be a waste of time.
No, just because it's related to economics doesn't mean it's a business topic, this is definitely, definitely more under math, also the people on /biz/ are way too dumb.

OP, I took a college class on game theory, it was a 4000 level class and considered one of the harder and more quantitative ones in the department, and it was a total fucking joke, we didn't even get close to learning anything useful and everything we covered was painfully simplistic babby tier shit. I've done a small bit of further reading as I am interested in the subject and it gets very math heavy once you get past the simplistic discrete models and quite complex. Personally I believe it has a lot of future application though as it should be able to be computationally simulated and applied to statistics or aggregate information in many different applications, as it does deal with decision making though there's no reason it would apply to any higher order science than biology.

Should a good video game developer know game theory? Is game theory taught in computer science courses? What are some ways in which game theory could be applied to video games?

INB4 OP doesn't know what game theory is:
I know what it is, and I'm aware of the fact that it isn't just used for recreational games, but also for economics, decision making, etc.

>>6669707
Well, I figure we're in a game theoretic context because ethics is about rational decision-making given that you want to maximize utility everywhere over all time.

So then it's a game theory question. It's about reasoning rationally about games. You start with symmetric games and work up to the general case, which includes asymmetric games.

For a symmetric game, you're looking for equilibria. You have Nash equilibria, but Nash-rationality isn't really rational at all, because it neglects that fact that one player's behaviour can correlate with another's. When you take that into account, you get superrationality. It generalizes to asymmetric games as shown in that post.

I guess I don't really understand your qualms.

Okay, so here's the thing. "Game theory", depending on who says it, may mean different things:

- Some people use it to mean "theory about strategies in a given (video) game", in which case it's usually random Sun Tzu quotes filled with theorycrafting here and there and nothing mathematically complicated at all (maybe just maximizing simple functions here or there, but that's it).
- In finance and economics, it's used to describe a branch of game theory which is as far as I know mostly about Nash equilibriums, with each player trying to maximize a function, etc. It's game theory for sure, but restricted to contexts that can be applied to finance, economics etc.
- Then there's game theory in general, which to give an example or something that you wouldn't see in finance, can be on a finite arena (the arena is the set of states the game can be in, and the list of the possible transitions from one state to another), and also includes weird winning conditions like Büchi ("Player 1 wins if state X of the arena is visited infinitely many times"), which obviously is not really useful in games that model "real" things because then the sequence of states is finite.

So what good is it to know things about winning strategies for games that have infinite sequences of states if those games don't model real things? Well they can model other useful mathematical objects. For instance, you may be able to prove something like "Proving this theorem is equivalent to proving that this weird game has a winning strategy for player 1" and then just prove the game has a winning strategy using game theory rather than proving the theorem. You may also do something like "We can map the inputs to that decision problem to games using this set of rules, and the answer to this decision problem is 'true' if the game we obtain has a winning strategy for player 1". That kind of things.

>>6385523
Combines Game Theory, Economics, and Neuroscience to provide a model for human decision making. Challenges notions in psychology to rely on shit like the reflex and instead replaces several interactions with oscillations and re-afference. All of it is presented in a common sense way with a lot of exposition and is easy to approach for the layman. Has many experimental examples to support the arguments and analogies for exposition.

Never once mentions or relies on garbage like Qualia. Challenges dualism and models relying on non-physical "souls" or other ridiculous shit.

>>6276863

Superior.

Uses game theory and economics models to explain stochastic decision making.

>>6198017

Statistics is hands down the most powerful branch of mathematics for real world decision making.
Rare things rarely happen,
but if you do something over and over again, they eventually will.
Describing and predicting group events allows us to predict and eliminate the special causes of variation inherent in everything we do.
It is a natural progression in the sciences and mathematics where you first learn these imaginary deterministic relationships (that in fact exist nowhere) then have to try and replicate them in the real world and you run into noise - that stuff that doesn't fit the story in your head.
Later, when you mature, you realize just how flimsy those deterministic relationships really are - materials fail, unaccounted for variables soon take over - you start to see the world as the march towards entropy that it really is.

Describing this march allows us to control our decent into hell.

Fourier analysis, DSP, Complex analysis, all are under the umbrella of statistics. Statistical Quality Control, Statistical Mechanics, Game theory, Queuing theory, Reliability theory, The theory of constraints: practically all real decision making relies on statistics and its progenitor probability.

Take a math based, probability and statistics course - leave that watered down social sciences shit behind - and you'll see both the power and the beauty of the branch.

Still convinced game theory is too dependent on assuming human decision making for it to be well grounded math. Feel like it's more of Psychology

I thought the grounds for game theory are based on predicting human decisions, how can something so complex and abstract be assumed/predicted? If human decisions can be predicted, why isn't there a formula that predicts stock market behavior? Psychology is beginning to be taught to business students and those working with stock trading for this reason. Again, cam somebody explain to me how game theory is math if it assumes human decision making will lean one way instead of the other?

Any online grinders here that apply mathematics and game theory to win online?
I was always curious as to how many people play here.

I have just started to delve deeper into the math side of the game, so far I have only studied EV but I can already tell how much easier having this knowledge improves decision making.

>>6117179
This is what is typically known as computer science.

You won't need a whole lot of math but the small amount you will need you will need to know well.


>CSC463H1 Computational Complexity and Computability[24L/12P]

This class for example and any data structures/algorithms course is going to rape you unless you either have a really good intuition for discrete math or know combinatorics for some reason.

All of the graphics courses will need some background in linear algebra (depending on the course, OpenGL will require transformational matrix stuff, for example).

I also just noticed at the bottom of the page it says.

>Students wishing to study scientific computing should have a strong background in mathematics, in particular calculus of several variables, linear algebra and statistics, be fluent in programming, and have a good understanding of data structures and algorithm design.

It's worth noting that the probability that comp sci people use isn't the same as the probability that statisticians use but due to backwards teaching methodologies you'll probably have to take a probability course geared at stats majors. It will cover a lot of shit you won't need to know like a bunch of distributions of well behaved random variables, but it won't cover shit you actually need to know. Bayesian networks, Markov chains, Markov Decision Processes, Partially Observable Markov Decision Processes, Random variables that are totally not well behaved (Some AI stuff will have you writing algorithms where you'll construct a random variable out of noise and manipulate it with probability to give you information about stuff), game theory, etc..

If these things all sound like things you're prepared to learn, then go for it. It's not what a mathematician would call a lot of math (barely any by mathfag standards), but it might be more than you're expecting to learn.

> > Quantum Groups - (Algebra)
> > Topological groups
> > Lie groups - (Lie Algebras)
> > Representation theory
> > Special functions - (Number theory)
> > Differential equations - (Lie Groups)
> > Partial Differential equations - (Functional analysis)
> > Ordinary Differential equations
> > Harmonic analysis
> > Fourier analysis
> > Potential theory
> >
> > Cartan geometry - (Fibre Bundle, Lie Groups)
> > Riemannian geometry - (Differential Geometry)
> > Kähler manifolds
> > Pseudo-Riemann geometry
> > Klein geometry
> > Euclidean geometry
> > Non-Euclidean geometry
> > Elliptic geometry
> > Hyperbolic geometry
> > Lie sphere geometry
> > Affine geometry
> > Convex geometry
> > Conformal geometry
> > Plane geometry
> > Trigonometry
> > Discrete Geometry - (Graph Theory)
> >
> > Fractal geometry
> >
> > Dynamical systems - (Mechanics, Differential equations)
> > Ergodic theory
> > Bifurcation theory
> > Optimal control theory
> > Optimization
> > Calculus of vations - (Differential Equations)
> >
> > Metric spaces - (Topology, Riemannian geometry)
> > Measure theory
> > Ergodic theory
> > Probability theory
> > Stochastics
> > Statistics
> > Estimation theory
> > Game theory
> > Decision theory

>>6058848
Why would you want it?

If you want to learn to make good choices, go read up on game theory or decision theory.

Free will is a social construct. It is about instinctive game theory ingrained in our brain architectures.

Anticipating blame and praise changes decisions.

"you could have acted differently" is a recognition that the decision would have been made differently if the actor had put more priority on the reaction of others etc.

All of that is formalizable by game theory and implemented of course in physics.

>>5957594
Not really. It should be treated with the same respect we assign to other technologies.

But by all means, have it crunch the numbers and figure out computationally frustrating problems. Just don't give it the nuclear launch codes.

I approach this problem from game theory. Mistrust, which may later prove to be unnecessary, is almost always a wise decision for first contact.

>>5926872
game theory is a study of strategic decision making. How will you design good AI without studying the best response dynamics of your game? You'll all need to make a game that doesn't have poor nash equilibria

did you thnk that your dumbed down econ course covered everything in one semester?

>>5883677
>sociology and psychology combined.
no, no it's not. If fact in doesn't have elements of either. It's more about understanding decision making and optimizing decisions in the work place. It's more about economics and mathematics and game theory.
>Where does the evidence come from?
By evidence if you mean observable decisions, then the evidence comes from people and history.
>What is falsified?
Decisions are falsified to be strictly inferior to other decisions.
Too lazy to talk about it more, so...
>https://en.wikipedia.org/wiki/Operations_research

I was playing 'Poker Night at the Inventory' for a while last night, learning the rules of the game for the first time, and I realized that there's a lot of game theory to it. A fair bit of strategy involved in the decision-making, much of it coming down to predicting and reacting to player behavior. Unlike board games where everything is known, you don't know the state of the other player's hands and can only make predictions based on their betting patterns.

I figured some sort of competitive evolutionary algorithm would ideally make for a competitive set of players which would perform optimally against a varied set of human players.

I've used GAs in the past to evolve Sudoku solutions and predict horse races, but never did any cross-competitive stuff. There's a fair bit of papers on the topic using neural networks, and ever since last night, I've been trying to come up with some sort of decision tree with parameters which could be represented as individuals in a gene pool. Perhaps even make it a mutable tree which can represent more complex decision paths.

Anyone have any experience with this sort of stuff, or try this directly?

>complexity, semantics, satisfiability, decision theory, statistics, machine learning, computer vision, game theory
>"CS not a science"

Stay mad.

That there are tons of SoftEng fags running around doesn't take anything away from CS. Computer Scientist != Code Monkey. Pic related.

>>5801722

I am a math fag too as in, I am double majoring in economics and math.

> In mathematics we deal with absolute certainty and complete rigor.

I dont get what you are saying. No one deals with certainty. 'Certainty' is a quality of people, not academic fields. Plenty of non-mathematicians get to certain. I recognize that mathematicians strive for rigor. Naturally when one uses mathematics in a different field (say, physics), one has to rely on mathematical rigor as well.

A lot of good economic is just mathematicians. They are just doing mathematics. Like decision theory, auction theory, game theory, I would call pure math.

for weeks now i have been fantasizing about a computer game with an infinite layer of complexity. let me explain, and while i do so try to not start drifting into "that's simulation theory", because that's really not what i would like this thread to be about.

imagine a game where there is no real hud- interface. say you sit into a fighter spacecraft, and your view does not have data on your screen, but the data is REALLY on different screens inside the cockpit, and so are the buttons that you can manipulate. also the ships steering does not follow some simple algorithm but a set of highly complex algorithms that simulate the reactions of the engines etc etc.

am i making sense? it's hard to describe what i want without getting into tldr territory...

the basic gist of my question is, when will we start approaching computation so powerful that games with these vast layers of complexity will start becoming a reality?

and once they do, how long until we can realistically simulate complex organisms inside a simulated virtual world... say simulate a fight between 4 gorillas and 20 hyenas just for the lulz... however with such a vast complexity that every biochemical decision making process, every anatomical and physical aspect is accurately calculated taking slight random variations and possible mutations (or even defects in some form or another) into consideration...

>>5650860

Ohh, learning.

See, I was lying in bed trying to sleep and thinking about math (specifically game mechanics for tabletop RPGs, I'm from /tg/) when the coin game occured to me, and now I have both the mathematical answer, the practical answer, and I know more about decision theory.

Thanks, /sci/. You're a good board.

From what I've read it is pretty hard to increase raw intelligence. There are subcultures on the internet that are fairly into this sort of thing. n-backers and the dual-n-back community mostly on google groups who discuss various ways to improve working memory, which seems to be associated with raw intelligence. There is also a growing community of people into nootropics; drugs which improve various aspects of cognition. There is also those people zap their brain with electricity. I can't remember the specific name of the community. There are also the various online brain game sites like lumosity. I went through most of these (apart from the brain zapping stuff). I think the one associated with the most evidence is exercise and nutrition, which is common sense, but people seem to look for little tricks and ignore the common sense strategies.

Also, you may want to read up on the psychologist Keith Stanovich (who I found on the website lesswrong, which you should also check out). He has written a fair amount on intelligence and rationality, and what exactly people can improve on. Pic related is a taxonomy of rationality from his book Rationality and the Reflective Mind. His conclusions are that you can improve a bunch of areas associated with system 2 style thinking (slow, deliberate thinking, as opposed to the other half of the dual-process: system 1 style thinking, which is off-the-cuff instinctive thinking that tends to dominate our minds, and we don't know it). You need more than just logic though to improve this aspect of your thinking (stuff like decision theory, inductive logic, staying away from contaminated mindware, debiasing strategies, etc). See this post that adds more areas: http://lesswrong.com/lw/7e5/the_cognitive_science_of_rationality/