I make a specific study of game design. This goes far beyond "I really liked this game", and into the the specific balance of the games, the various features, and how they impact the overall experience. I look at existing games of examples of how various elements can interact and try to discern what makes good game design and what is bad.
Today, I will be analysing randomness in games. The degree of randomness in a game occurs on a sliding scale. On one end, you have no randomness. Everything is perfectly predictable. You take an action, you can predict the result. The interest in such games comes from the actions of another entity, be it another player or the game world itself.
Chess is a great example of a chanceless game. Everything is visible. Every action predicatable. You move a pawn, you know exactly where it will end up. This allows you to formulate and execute elaborate strategies, and the only determiner of who wins is the players. This is a very solid game design.
On the oppisite extreme, you have games that are pure chance. This is gambling. A slot machine. You put in your coin, pull the lever, and the result is pure chance. A roullete wheel, or Bingo, you have no influence over the outcome, you can only wait for the result. Whether it is a series of random events, as is the case with bingo, or a singular event, as with a roullette wheel, the result is the same, one just teases it out longer. This is a poor game, as there is no real involvement from the player, but it can play off some of the same triggers that a game does- the desire to win, the feeling of risk, etc.
A variation of that is the guided game of chance. Most card games fall into this category. You have a strong element of chance- what hand you are dealt, what cards you draw, etc. However, that chance is not completely out of your control. You can still win with a poor hand if you play your cards right. Skill can work to counter chance, so a skilled person with a poor hand can beat an unskilled person with a good hand. You can again enact strategies. Only now, strategies are not deterministic, but based on chance. You develop strategies to increase the odds of your own victory. This generally does not occur on an explicit level; you typically don't think that playing a certain card has a 5% greater chance of winning over another card. But you still make choices as to the best card to play.
However, even with all of the strategies in play, you don't win every hand, even with a perfect stategy. That element of chance is still there, and still matters significantly. However, there is an important property of card games; they don't rely on a single attempt. You don't play a single hand, you play a series of hands. You don't rely on the outocme of a single draw, you work off of a series of draws. If you have any understanding of probability, you know that this heavily normalizes the average outcome. Flipping 2 coins, the chance of various number of heads is very wild. 25% of 0, 50% chance of 1, and 25% chance of 2. That is a very wide spread, because there are only 2 samples. If you flip it ten times, then it balances out. There is .097% chance of 0 heads, but a ~25% chance of getting 5 heads. There is a 65% chance of getting 4,5,or 6 heads. The likelyhood of getting an normal result is much, much higher than the chance of getting abnormal results. The more iterations, the smaller the standad deviation. This means that the alterations you do to the probability have a much more profound impact. If you have a 75% chance of gettign heads on any flip, then you end up with a nearly 30% chance of rolling an 8, an you will get 8+ heads over 50% of the time.with 100 flips, you have practicalyl no oflanding outside the 60-80 heads range.
So, multiple iterative chances has the impact of heavily normalizing the results. Chance still matters moment to moment, which helps keep the flow of the game exciting, but the final outcome is still mostly dependant on how well you play, and shift the probabilities in your favor. In our coind flip example, 100 flips will favor heads 50% of the time. if you can shift the probabiltty by just 5%, the chance of winning shoots up to 80-90%. So, with iterative chances, swaying the probablity, even slighlty, can be a huge advantage. Hence, your skill at playing becomes very important, and the chance involved ceases to have a strong influence on the final result. Yet, moment-to-moment, chance does matter. The unpredicatability is still there, the tension of whether the card is a 7 or a 9, and the course of the game is interesting.
This is even more true when the quality of a hand is bounded. Take hearts, for example. Basic idea of the game is you play cards from your hand, and whoever plays the highest card of the right suit takes all of the played cards. Certain cards are wroth poitns, which you want to avoid. Hence, a good hand allows you to not take many points if played well. However, there is an additional rule. If you take all of the points in a round, you shoot the moon, and get no points, whilst everyone else takes the max number of points. This is a tricky thing to pull off, but is a huge advantage in winning the overall game. This also means that a hand that will take many cards does not have to be an awful hand; it can instead be a good hand. This helps bound the quality of your hand, and hence limits the probablity from outright dooming you.
In addition, you have accumulating value. Be it the cards within a hand or the chips you earn between hands, you get to accumulate the value of your successes, building towards a goal. This acts as a overall measurement of the probabilities. Without it, you are eternally on the influence of the last chance, and the benefit of iterative chances is lost, if not inverted. If every hand is double or nothing, it does not matter how good you do overall. A failure on the current hand becomes a complete failure, and hence you are back to the singular chance situation. In fact, multiple iterations make this worse, as it is no longer the chance that you have accumulated more, but the chance that you have never failed. Even if you have a 99% chance of winning, you will have a 64% chance to have failed by 100 iterations. a 90% chance of success becomes a .00026% chance of sucess after 100 iterations. Any chance of failure is quickly magnified. Of course, just becuase you are not accumulating value doesn't mean you risk losing everything on a single failure. As shown, a single point of failure with multiple iterations is a bad idea. If instead you are iterating until success, those probabilities flip, and even small chance of success will quickly build to a near-certainty. Which makes the entire excercise a waste if time and resources. This is why instant-death in a game is a bad idea. Even if its rare or easily avoidable, eventually it will get them, and becomes the single point of failure over repeated iterations.
The magnitudes of the probabilities involved influence how many iterations are needed to smooth it out. If you are talking about 50% chances of sucess, even a hundred iterations clmaps down on the variance by quite a bit. If you are talking about .5%, you need many more iterations. It is important to match these up properly.
Applying these principles to other games, you can identify some interesting features. Take D&D, for instance. The central mechanic for combat is you roll a dice, and it determines whether you hit or miss, and if you hit, you roll dice to determine how much damage. Lots of chance involved. However, the players receive a very high degree of control over this chance. The characters get modifiers they can apply to the rolls, boosting them higher, or making the opponent need a higher result. This comes into play at multiple levels. At one level, it reqards the character creation. your ability to create a good character influences your bonuses, and hence your relative advantage. Another level is tactical. The maneuvering and strategies enacted on the battlefield can have an even more profound impact than character creation. Manuevering can give you specific bonuses to hit, or for certain builds to damage, further shifting the probabilities. Good tactics can also supress an enemies ability to attack, making it so they have fewer chacnes to hit you. Cover can increase your defenses, concealment can add an additional miss chance, and as a result, you have a very high degree of control over the probabilities involved. Couple this with the fact that you are rolling a lot of chances, and you meet these criteria. You have a large number of iterative chances to normalize the results, and the ability to shift the probabilities to your advantage.
However, there are also flaws in that system. Sometimes, you don't get a large number of iterative chances. You get one chance. You need to jump over a pit, for instance. You have a 75% chance of making it. In an iterative situation, that is a very significant advantage. However, in a single event, that is a 25% chance of straight-up failure. There is no iterative advanatage to be gained over multiple tries, such as when you are attacking an enemy and wearing down its health. You come down to a single roll. Its a gamble. The overall game design has broken down.
Now lets examine minecraft. The largest source of randomness is finding ores while mining/spelunking. Each block of stone has a chance of being an ore. Mining or spelunking, you expose a large number of blocks, hoping to find ore. Since this is random, there is a lot of chance, and you never know when you will find ore. This makes it more exciting when you do find ore than if you get it reguarly. However, since it is a highly-iterative chance, the rate of mineral gathering is fairly predictable, and does not suffer wide variance on average. You do have control over the chances. Efficient mining techniques can expose more stone per stone mined than inefficent techniques, and expose stone that is more likely to have ores, and so shifts the probabilites to your favor. This also follows the accumulation rule, as every ore found accumulates to keep track of your overall progress. And even though individually, you have a low chance of finding ores, the volume of tests makes up for it. Anyone who has mined can tell you that you see plenty of ores, without being drowned in it, so this aspect is reasoanbly balanced. So, minecraft's mining follows these principles, and at this level, is a solid mechanic.
Now lets look at its enchantments. you gather a large number of resources, and enchant an item. This is a singular event, and you either get something good, or something bad. You can't accumulate the results of successes. You can influence the probabilities, but at an ever-increasing cost, and since it is a singular check, that influence is weakened. These enchantments break from the good design of the rest of the game, standing out as a unmatching mechanic, and violating the game design principles.
In summary, chance can be a useful element in game design. You want to leave the player in control, since that interaction is what makes a game as opposed to a movie, and is the point. Chance would seem to defy that principle, but when used correctly, it can work with it. A highly-repteated chance with accumulating returns and the ability to influence the probablities can help ensure the player's control is meaningful and interesting while still allowing chance to have its influence. However, it is not a perfect design. Even with highly iterative chances to even it out, there still remains a chance that a player, not matter how skilled or how well they played, just gets screwed due to chance. This is not a desireable outcome. However, it is the trade-off for adding in the interest of randomness into the general game flow. And if that singular failure is not the end overall, the game itself can become a highly iterative chance with outcomes influenced by the player, and drive the impact of random unfair failure even lower. If the chance of unfair failure is low, then it does not take many iterations to smooth it over. But chance can be a succefully incorporated element of a game if done right.
What I'd like to see would be a multiplayer shooter with skill adjustments (No, Not quite like borderlands). Like, you can make your character along a set amount of points for certain attributes which will adjust damage etc. Of course to make the game fair leveling shouldn't do anything to level up. You're stuck with the attributes you chose for your character from the beginning. Also custom weapons and their attributes would be cool. For balance there you could make it so that recoil could depend on weapon damage/caliber. Weight could let you move and aim faster or slower.
What I'd like to see would be a multiplayer shooter with skill adjustments (No, Not quite like borderlands). Like, you can make your character along a set amount of points for certain attributes which will adjust damage etc. Of course to make the game fair leveling shouldn't do anything to level up. You're stuck with the attributes you chose for your character from the beginning. Also custom weapons and their attributes would be cool. For balance there you could make it so that recoil could depend on weapon damage/caliber. Weight could let you move and aim faster or slower.
I can continue but I don't think I need to.
Balance would be very, very important for a concept like that. You don't want one buid to be stronger than the others, yet still make different builds distinct. This is complicated by the unpredictable nature of strategy. A long-range weapon with a fast build may not end up with much damage output, but be practically unstoppable due to being too far away to be hit and able to stay that way, for instance. Its tricky to balance a game with a wide range of abilities when they are well-known, such as with team fortress. Balancing arbitrary skill sets gainst each other would be very, very tricky. Though if you can pull it off, the result would be pretty cool.
One obnoxious thing about chance (that makes some games more interesting) is hidden or privileged information. Say you're playing as the Spy in Team Fortress 2, you decide to walk through narrow door-And a Pyro just happens to enter through as soon as you go through.
With this kind of stuff, you face player-created randomness. So you have to be cautious around doors and narrow passages when playing as Spy in Team Fortress 2,among many, many other things.
Oh, and don't forget dice humping. Dice humping occurs when a player constantly gets the worst possible outcome in a game due to luck. Let's say you're trying to play Battleship, but your opponent just happens to hit all your ships without fail-That's when a player gets dice humped. All luck-heavy games offer the chance for a player to get dice humped, but good games try to make it more and more unlikely, or eliminate chance in vital areas of the gameplay.
Balance would be very, very important for a concept like that. You don't want one buid to be stronger than the others, yet still make different builds distinct. This is complicated by the unpredictable nature of strategy. A long-range weapon with a fast build may not end up with much damage output, but be practically unstoppable due to being too far away to be hit and able to stay that way, for instance. Its tricky to balance a game with a wide range of abilities when they are well-known, such as with team fortress. Balancing arbitrary skill sets gainst each other would be very, very tricky. Though if you can pull it off, the result would be pretty cool.
I think that having classes being better than another class would be a good difference for gameplay. It would make teamwork a requirement. You couldn't take a slow moving heavy class against a sniper, you'd need a fast cqb class.
One obnoxious thing about chance (that makes some games more interesting) is hidden or privileged information. Say you're playing as the Spy in Team Fortress 2, you decide to walk through narrow door-And a Pyro just happens to enter through as soon as you go through.
With this kind of stuff, you face player-created randomness. So you have to be cautious around doors and narrow passages when playing as Spy in Team Fortress 2,among many, many other things.
thats not so much randomness as a rich interactions. There is a difference between the unpredicatability of a setting or player and randomness.
Oh, and don't forget dice humping. Dice humping occurs when a player constantly gets the worst possible outcome in a game due to luck. Let's say you're trying to play Battleship, but your opponent just happens to hit all your ships without fail-That's when a player gets dice humped. All luck-heavy games offer the chance for a player to get dice humped, but good games try to make it more and more unlikely, or eliminate chance in vital areas of the gameplay.
Yes, that is the real downside of adding chance. That is the reason you need the highly iterative chance, to drive the chance of that really low. In a computer game, it can be possible to specifically detect and counter that,
I think that having classes being better than another class would be a good difference for gameplay. It would make teamwork a requirement. You couldn't take a slow moving heavy class against a sniper, you'd need a fast cqb class.
better against other classes, sure. Better at certain things, certainly. But one can't be definitevely better than the others.
If people are interested, I could make this into a series, and do periodic posts about game design. I have a lot of topics I could delve into.
I'd love to read more on this. I'm planning on being a programmer in the future, and your first post actually made me think about the types of games that would not be unfair to the player.
You proved a very good point in the OP, Mystify. But I do actually have something to say about some points you made. If you are using an American Penny, the chances of it landing on heads are slightly lowered, as tails on the American Penny weighs more. Not everyone knows this, so the people who do know this would pick tails, thus giving them a higher chance of winning the game.
Roulette
Quote from Wikipedia »
Just felt like giving that to you as a reference for the Roulette odds :biggrin.gif:
I didn't read the whole post, but I could tell you thought it out and did math. You have proof to backup your reasons, and not very many posts anywhere do that anymore. I am glad to see someone who actually puts thought into their post, and doesn't just throw out a simple opinion and expect everyone else to have the same opinion as them.
Today, I will be analysing randomness in games. The degree of randomness in a game occurs on a sliding scale. On one end, you have no randomness. Everything is perfectly predictable. You take an action, you can predict the result. The interest in such games comes from the actions of another entity, be it another player or the game world itself.
Chess is a great example of a chanceless game. Everything is visible. Every action predicatable. You move a pawn, you know exactly where it will end up. This allows you to formulate and execute elaborate strategies, and the only determiner of who wins is the players. This is a very solid game design.
On the oppisite extreme, you have games that are pure chance. This is gambling. A slot machine. You put in your coin, pull the lever, and the result is pure chance. A roullete wheel, or Bingo, you have no influence over the outcome, you can only wait for the result. Whether it is a series of random events, as is the case with bingo, or a singular event, as with a roullette wheel, the result is the same, one just teases it out longer. This is a poor game, as there is no real involvement from the player, but it can play off some of the same triggers that a game does- the desire to win, the feeling of risk, etc.
A variation of that is the guided game of chance. Most card games fall into this category. You have a strong element of chance- what hand you are dealt, what cards you draw, etc. However, that chance is not completely out of your control. You can still win with a poor hand if you play your cards right. Skill can work to counter chance, so a skilled person with a poor hand can beat an unskilled person with a good hand. You can again enact strategies. Only now, strategies are not deterministic, but based on chance. You develop strategies to increase the odds of your own victory. This generally does not occur on an explicit level; you typically don't think that playing a certain card has a 5% greater chance of winning over another card. But you still make choices as to the best card to play.
However, even with all of the strategies in play, you don't win every hand, even with a perfect stategy. That element of chance is still there, and still matters significantly. However, there is an important property of card games; they don't rely on a single attempt. You don't play a single hand, you play a series of hands. You don't rely on the outocme of a single draw, you work off of a series of draws. If you have any understanding of probability, you know that this heavily normalizes the average outcome. Flipping 2 coins, the chance of various number of heads is very wild. 25% of 0, 50% chance of 1, and 25% chance of 2. That is a very wide spread, because there are only 2 samples. If you flip it ten times, then it balances out. There is .097% chance of 0 heads, but a ~25% chance of getting 5 heads. There is a 65% chance of getting 4,5,or 6 heads. The likelyhood of getting an normal result is much, much higher than the chance of getting abnormal results. The more iterations, the smaller the standad deviation. This means that the alterations you do to the probability have a much more profound impact. If you have a 75% chance of gettign heads on any flip, then you end up with a nearly 30% chance of rolling an 8, an you will get 8+ heads over 50% of the time.with 100 flips, you have practicalyl no oflanding outside the 60-80 heads range.
So, multiple iterative chances has the impact of heavily normalizing the results. Chance still matters moment to moment, which helps keep the flow of the game exciting, but the final outcome is still mostly dependant on how well you play, and shift the probabilities in your favor. In our coind flip example, 100 flips will favor heads 50% of the time. if you can shift the probabiltty by just 5%, the chance of winning shoots up to 80-90%. So, with iterative chances, swaying the probablity, even slighlty, can be a huge advantage. Hence, your skill at playing becomes very important, and the chance involved ceases to have a strong influence on the final result. Yet, moment-to-moment, chance does matter. The unpredicatability is still there, the tension of whether the card is a 7 or a 9, and the course of the game is interesting.
This is even more true when the quality of a hand is bounded. Take hearts, for example. Basic idea of the game is you play cards from your hand, and whoever plays the highest card of the right suit takes all of the played cards. Certain cards are wroth poitns, which you want to avoid. Hence, a good hand allows you to not take many points if played well. However, there is an additional rule. If you take all of the points in a round, you shoot the moon, and get no points, whilst everyone else takes the max number of points. This is a tricky thing to pull off, but is a huge advantage in winning the overall game. This also means that a hand that will take many cards does not have to be an awful hand; it can instead be a good hand. This helps bound the quality of your hand, and hence limits the probablity from outright dooming you.
In addition, you have accumulating value. Be it the cards within a hand or the chips you earn between hands, you get to accumulate the value of your successes, building towards a goal. This acts as a overall measurement of the probabilities. Without it, you are eternally on the influence of the last chance, and the benefit of iterative chances is lost, if not inverted. If every hand is double or nothing, it does not matter how good you do overall. A failure on the current hand becomes a complete failure, and hence you are back to the singular chance situation. In fact, multiple iterations make this worse, as it is no longer the chance that you have accumulated more, but the chance that you have never failed. Even if you have a 99% chance of winning, you will have a 64% chance to have failed by 100 iterations. a 90% chance of success becomes a .00026% chance of sucess after 100 iterations. Any chance of failure is quickly magnified. Of course, just becuase you are not accumulating value doesn't mean you risk losing everything on a single failure. As shown, a single point of failure with multiple iterations is a bad idea. If instead you are iterating until success, those probabilities flip, and even small chance of success will quickly build to a near-certainty. Which makes the entire excercise a waste if time and resources. This is why instant-death in a game is a bad idea. Even if its rare or easily avoidable, eventually it will get them, and becomes the single point of failure over repeated iterations.
The magnitudes of the probabilities involved influence how many iterations are needed to smooth it out. If you are talking about 50% chances of sucess, even a hundred iterations clmaps down on the variance by quite a bit. If you are talking about .5%, you need many more iterations. It is important to match these up properly.
Applying these principles to other games, you can identify some interesting features. Take D&D, for instance. The central mechanic for combat is you roll a dice, and it determines whether you hit or miss, and if you hit, you roll dice to determine how much damage. Lots of chance involved. However, the players receive a very high degree of control over this chance. The characters get modifiers they can apply to the rolls, boosting them higher, or making the opponent need a higher result. This comes into play at multiple levels. At one level, it reqards the character creation. your ability to create a good character influences your bonuses, and hence your relative advantage. Another level is tactical. The maneuvering and strategies enacted on the battlefield can have an even more profound impact than character creation. Manuevering can give you specific bonuses to hit, or for certain builds to damage, further shifting the probabilities. Good tactics can also supress an enemies ability to attack, making it so they have fewer chacnes to hit you. Cover can increase your defenses, concealment can add an additional miss chance, and as a result, you have a very high degree of control over the probabilities involved. Couple this with the fact that you are rolling a lot of chances, and you meet these criteria. You have a large number of iterative chances to normalize the results, and the ability to shift the probabilities to your advantage.
However, there are also flaws in that system. Sometimes, you don't get a large number of iterative chances. You get one chance. You need to jump over a pit, for instance. You have a 75% chance of making it. In an iterative situation, that is a very significant advantage. However, in a single event, that is a 25% chance of straight-up failure. There is no iterative advanatage to be gained over multiple tries, such as when you are attacking an enemy and wearing down its health. You come down to a single roll. Its a gamble. The overall game design has broken down.
Now lets examine minecraft. The largest source of randomness is finding ores while mining/spelunking. Each block of stone has a chance of being an ore. Mining or spelunking, you expose a large number of blocks, hoping to find ore. Since this is random, there is a lot of chance, and you never know when you will find ore. This makes it more exciting when you do find ore than if you get it reguarly. However, since it is a highly-iterative chance, the rate of mineral gathering is fairly predictable, and does not suffer wide variance on average. You do have control over the chances. Efficient mining techniques can expose more stone per stone mined than inefficent techniques, and expose stone that is more likely to have ores, and so shifts the probabilites to your favor. This also follows the accumulation rule, as every ore found accumulates to keep track of your overall progress. And even though individually, you have a low chance of finding ores, the volume of tests makes up for it. Anyone who has mined can tell you that you see plenty of ores, without being drowned in it, so this aspect is reasoanbly balanced. So, minecraft's mining follows these principles, and at this level, is a solid mechanic.
Now lets look at its enchantments. you gather a large number of resources, and enchant an item. This is a singular event, and you either get something good, or something bad. You can't accumulate the results of successes. You can influence the probabilities, but at an ever-increasing cost, and since it is a singular check, that influence is weakened. These enchantments break from the good design of the rest of the game, standing out as a unmatching mechanic, and violating the game design principles.
In summary, chance can be a useful element in game design. You want to leave the player in control, since that interaction is what makes a game as opposed to a movie, and is the point. Chance would seem to defy that principle, but when used correctly, it can work with it. A highly-repteated chance with accumulating returns and the ability to influence the probablities can help ensure the player's control is meaningful and interesting while still allowing chance to have its influence. However, it is not a perfect design. Even with highly iterative chances to even it out, there still remains a chance that a player, not matter how skilled or how well they played, just gets screwed due to chance. This is not a desireable outcome. However, it is the trade-off for adding in the interest of randomness into the general game flow. And if that singular failure is not the end overall, the game itself can become a highly iterative chance with outcomes influenced by the player, and drive the impact of random unfair failure even lower. If the chance of unfair failure is low, then it does not take many iterations to smooth it over. But chance can be a succefully incorporated element of a game if done right.
Game Design Theory 2 - Linearity
I can continue but I don't think I need to.
I wonder how in the **** you read books.
"tl;dr"
-Rips Book-
You're meant to ignore those people. (they're trying to troll...)
Fien. I won't feed.
Only feed the fat trolls! Hehhe!
Balance would be very, very important for a concept like that. You don't want one buid to be stronger than the others, yet still make different builds distinct. This is complicated by the unpredictable nature of strategy. A long-range weapon with a fast build may not end up with much damage output, but be practically unstoppable due to being too far away to be hit and able to stay that way, for instance. Its tricky to balance a game with a wide range of abilities when they are well-known, such as with team fortress. Balancing arbitrary skill sets gainst each other would be very, very tricky. Though if you can pull it off, the result would be pretty cool.
With this kind of stuff, you face player-created randomness. So you have to be cautious around doors and narrow passages when playing as Spy in Team Fortress 2,among many, many other things.
Oh, and don't forget dice humping. Dice humping occurs when a player constantly gets the worst possible outcome in a game due to luck. Let's say you're trying to play Battleship, but your opponent just happens to hit all your ships without fail-That's when a player gets dice humped. All luck-heavy games offer the chance for a player to get dice humped, but good games try to make it more and more unlikely, or eliminate chance in vital areas of the gameplay.
I think that having classes being better than another class would be a good difference for gameplay. It would make teamwork a requirement. You couldn't take a slow moving heavy class against a sniper, you'd need a fast cqb class.
thats not so much randomness as a rich interactions. There is a difference between the unpredicatability of a setting or player and randomness.
Yes, that is the real downside of adding chance. That is the reason you need the highly iterative chance, to drive the chance of that really low. In a computer game, it can be possible to specifically detect and counter that,
better against other classes, sure. Better at certain things, certainly. But one can't be definitevely better than the others.
I'd love to read more on this. I'm planning on being a programmer in the future, and your first post actually made me think about the types of games that would not be unfair to the player.
Roulette
Just felt like giving that to you as a reference for the Roulette odds :biggrin.gif:
I didn't read the whole post, but I could tell you thought it out and did math. You have proof to backup your reasons, and not very many posts anywhere do that anymore. I am glad to see someone who actually puts thought into their post, and doesn't just throw out a simple opinion and expect everyone else to have the same opinion as them.