Game Theory Discussion - Today's topic: skill ceiling

[YCobb]

that's something I've wondered before, too. I don't know how familiar you are with how computers work, but generally anything random is generated based on the system clock at any given point. This is close enough to random because it's difficult to predict what the smaller components of that number will be - they just change too quickly to predict. This wouldn't work for Tetris because, to the extent of my knowledge, the Game Boy did not have an internal clock.

Personally, though, I'd wager that it simply starts up its own clock when you boot the game. I'm not sure if the Game Boy's clock cycles are fast enough this to be suitable, though, as mashing buttons as quickly as possible to enter the game might hypothetically allow you to play the exact same game every time. If the processor is fast enough, infinitesimal delays would factor in, but at 4.19 MhZ (according to wikipedia) I'm not sure it'd work. The buttons wouldn't be polled anywhere near every time the processor ticks, probably more in tune with the screen's updates.

[Syobon]

Huh? Tetris has a set number of shapes, each assigned a value. It then randomly generates one of those values to determine which shape to give to the player (in reality it's slightly more complex to avoid too painful unlucky streaks).

Games that randomly generate complex 3D shapes like Minecraft (though that's still fairly simple since it's blocks) use much more complex algorithms but the basic principle is the same. A game is purely digital, which means everything is or can be assigned a number. Numbers can be (imperfectly) randomly generated by complex mathematical formulas. You can randomise anything you'd like that way.

Unless you're asking what Tetris uses as a seed for it's random number generating, in which case I haven't been able to find a real answer, but I did find out it generates a different seed every frame and it's probably dependant on player inputs and how many frames the game has been running.

[YCobb]

I know they're predetermined, yeah. Just wondering about the seed like you said. The problem with frames/input is that with the gameboy's speed, I suspect you could still pretty easily engineer identical games, at least in the beginning.

((Not that it would be too difficult for the game to generate tetrominoes piece-by-piece, since they only have so many possibilities, all of which are already present in the game))

[DoNotDelete]

I knew that the shapes had pre-determined values too; It's more that I'm perplexed by the game occasionally dealing me instances of five or more identical shapes in a row - I just played a game of the original Tetris where I was dealt an instance of 'S' shapes which almost filled up my entire field of play.

Other times I'll get four 'I' (long) shapes in quick succession - which is equally irritiating when I've just experienced a chain of 20-30 shapes without one showing up.

[Galaxy Man]

I have once gotten a tetris game comprised of all S bends

it was the worst experience of my life

[Brekkjern]

This is how you make a random number while coding in Lua.

math.randomseed(seed)
math.random(lower, upper)

While not a common language to create games in and strictly speaking not a programming language, the procedure is similar in all other languages I know of.

It is pseudo random, meaning that it is based on an algorithm and that algorithm can be predicted, but for all intents and purposes, it is random. You set the seed with math.randomseed() and math.random() returns a value using that seed in an algorithm. With the same seed, math.random() will return the same sequence of numbers.

In the case of the GameBoy with a processor of 4.19MHz it does 4 190 000 cycles per second. It probably records the cycle you press start as the random seed.

[YCobb]

But on the Game Boy, I expect the buttons are polled only on each frame drawn. (this is true in modern systems, at least) It doesn't have 4.19 million possible values per second, it has sixty. (Of course, it's a tad more complex than that based on how you navigate the menus)

Also... the bit about lua isn't really relevant here? The semantics for telling the computer to make a random number are very different from the internal method. We had already jumped to talking about seeds, anyway.
As an aside, I don't know about Lua, but most languages let you skip assigning a seed and just use the system's date/time value. It changes so quickly that to predict its output would require absurd precision.

[Syobon]

New topic: Mindgames vs Random Guesses

I'm sure you're all familiar with Rock-Paper-Scissors, where you without knowledge of your opponent's decision have to pick the option that counters it. Often times this will lead to players trying to get into each other's head after repeat games "he just picked scissors so he probably won't pick it again". This is mindgames in it's purest form, where there are no other methods to make the right choice than the psychological one. The most famous game with a strong psychological foundation is probably poker (*). To win one usually has to guess what cards the other player has in his hand. The big difference with RPS however is that there are factors besides the psychological one. Because certain cards have an advantage over others, they influence the player's playstyle, which makes it possible to determine a player's cards by analysing their playstyle. What makes poker fun for most people is that this results in the mathematically optimal playstyle not being the actual optimal one. If you always play in the most mathematically optimal way your opponent can predict your moves and cards and exploit this to win, even if he had worse cards (ie. bluffing)! Players will still generally have a thought process and play according to risk/rewards probabilities though, so the game becomes about discerning your opponent's thought progress and applying it to the 'situation' which the cards that you can see create, to predict their moves and cards. What's important here is that each round the cards create a different 'situation' which provides clues to your opponent's thought process in contrast to RPS where there are no extenuating factors and every round is the same 'situation' which holds no clues.

The point I want to make is the difference between the mindgames in RPS and poker. In poker you have clues in the form of the cards and the variety of playstyles with down- and upsides, where as in RPS you basically have to be a mindreader and guess between three equal choices (unless your opponent for some reason follows a really predictable pattern like alternating rock and scissor). Also worth noting is that it becomes more plausible to make an accurate prediction after multiple rounds.

Now let's look at some examples in actual videogames. For starters, something relatively simple: Team Aerial Combos in UMvC. You have three choices with different advantages. Your opponent has to predict which choice you make to prevent you from gaining those advantages. This is a good implementation of mindgames as the optimal choice depends on the situation (eg. stealing meter from Morrigan), thus there are clues upon which to base your predictions. However, there is a glitch that basically changes the situation into two choices for the attacking player, neither with specific advantages or disadvantages. Thus the defending player doesn't have any real clues to base his prediction off and has to randomly guess, which would be a bad implementation of a mindgame.

Finally, a more complex example. Opening moves (or build orders) in Starcraft 2. In SC2 the optimal strategy is highly dependant on your opponents' strategy. However determining that requires you to scout him first, which always has a cost (lost mining time+unit cost). Thus not only is your opening move 'blind', how long you stay 'blind' is also part of your opening move (how long do you build up economy before sending a scout). The following factors provide clues to determine opening moves however: the map, metagame, viability of the strategy itself, your opponent's playstyle (most starcraft players have at least distinct preferences in strategy), what your opponent thinks your playstyle is, previous matches (it's less desirable to repeat the same opening too much since your opponent will have had time to analyse it), situation in a tournament (do you really dare to pull off a risky opening when it's the deciding match for a 100$k prize?) and your experience with executing a certain opening. Also interesting is that it's possible to mislead your opponent by showing him misleading information (cancelling or hiding buildings). All these factors contribute to making SC2 openings fascinating mindgames. As you might have noted some of these factors aren't a direct consequence of game design, but the game design does in a way allow them to be important.

In conclusion I'd posit that mindgames in videogames are better when there are more clues to add depth to the decisionmaking, and that variety to the decisionmaking can be introduced due to having a multitude of situations that affect the mindgame. Mindgames are also a way to make it harder for games to be 'solved'.




(*):I'll be using 2-card texas hold'em no limit for my example to avoid misunderstandings

[Syobon]

Article about the ethics of cheating, it's implementation in microtransaction and some conclusions we can draw from this about what makes a game fun. Interesting read, the top comments are definitely also worth checking out.

[wordNumber]

I don't think that things like Rocket jumping or exploiting level geometry really count as "Cheating," per say. They are still exploits unintended by the developer, but cheating in my mind always had more to do with entering a code or hacking/using some device to create unfair situations. And so many things along the lines of RJ have evolved into gameplay mechanics. I mean, combos in fighting games were born from a glitch, and now combos are a big part of a fair amount of genres. It just seems like there should be a separation between "Cheating" and exploits.

Also the Kiln jump thing he mentions strikes me as trying to say that speed running in itself is a kind of cheating. I suppose one could argue that due to it speed running's reliance on exploits, but speed runs tend to require so much skill that I feel it's a bit invalidating to label speed runners as cheaters.

[Syobon]

I feel that the author is not trying to attach any derogatory connotations to "cheating" but uses it in it's purest definition as breaking the rules the dev intended.

However, I definitely agree that exploits like Rocket Jumping should not be looked at as cheating but as an unintentional enrichment of game mechanics, which proves the conclusion the author draws, that cheating should be encouraged. I find that conclusion only worth considering if you as you said make the separation between exploits and external cheats. Even still, exploits can be game-breaking. However, I think the blame in those instances fall solely on the developers leaving exploits in, rather than the players using them. In particular, I think it's unethical for MMOs to permaban players for using exploits.

[YCobb]

Consider the skiing mechanics in Tribes or in some TF2 servers - they both came from glitches, which the article would consider "cheating."

I think the problem with the article, (not that it's bad, I'm just criticizing) as most of the commenters pointed out, is that it seems to be ignoring the idea of emergent gameplay. The comment about how cheating give such empty rewards that it loses its appeal struck me as especially insightful. ((It's a bit different, but I've noticed the same with pirating games.))

[Syobon]

A term that is often used when talking about competitive games is skill ceiling. For those not familiar with it: skill ceiling denotes the idea that there is a point where you become so good at a game that there is no room for improvement. It has the connotation that games with a low skill ceiling are easy and games with a high skill ceiling are hard (this isn't necessarily accurate, which will hopefully become clear from my next paragraph).

I feel this term is a bit misleading though, some people seem to interpret it as meaning you've reached perfection. As long as you're human though, this is clearly impossible almost no matter how simple the game. Therefore, I feel skill asymptote much more accurately describes the phenomenon. As you reach a higher level of skill it becomes more and more difficult to progress, until you're putting in tremendous amounts of effort no noticeable progress, or even to just maintain your skill. This exists outside of gaming too, most noticeably in sports.

My opinion is that while a high skill asymptote (sa.) isn't necessary for a good game, it does increase the longevity and replay value of it tremendously. However, it shouldn't be achieved at all costs (notably arbitrarily decreasing user friendliness solely for that purpose). Large amounts of decisionmaking, deep and meaningful game mechanics, minimizing RNG and a fair playing field are all examples of good game design that will also lead to a high sa. In competitive games though, a high sa. is essential to keep a game interesting for a longer period of time.

[Brekkjern]

You forget the inverse. The skill floor.

Prime examples would be grenade/rocket launchers in Call of Duty. Weapons that are easy to hit with and strike hard which allows lower skilled players to compete with higher skilled players. A game with a low skill floor will allow more people to play, but also increase the difference between the lower and higher skilled which means that as long as the lower skilled player does not have access to their "crutch", they will not stand any sort of chance against a high skill player.

This is ultimately an issue with competitive games, but I do not believe all games should be designed to have a low skill floor and a higher skill ceiling. It all depends on what demographic you want to appeal to. Call of Duty has a lower skill floor than the Battlefield series and as such they both try to appeal to slightly different demographics while still competing over the same user base. Their skill ceilings are also so low that they are not eating into the Tribes/Counter Strike demographic.

I'm too tired to form a proper, coherent argument, but this should give you something to work with at least.

[Syobon]

Again, I think skill floor is a bit of a misleading term, as there's no real limit to how bad someone can be at videogames. "Ease of entry" is probably a more accurate description. I think "ease of entry" is always a positive thing in games, "easy to learn, hard to master" should be an ideal games strive to. However the problem comes from things like in your example where the gameplay is negatively influenced to achieve it. Grenade launchers in CoD create an uneven playing field because they almost guarantee two free kills. Because of this it also lowers the skill assymptote, like you said it acts a crutch that dimineshes the diference between a skilled and lesser skilled player.