4 ms·

There's an interview with Sid Meier floating around somewhere where he discusses how, in Civilization, people will assume the computer is cheating at combat, un

by GlennS 6y ago

There's an interview with Sid Meier floating around somewhere where he discusses how, in Civilization, people will assume the computer is cheating at combat, unless you actually bias it quite strongly in the human's behaviour. Civilization shows you the odds for each fight based on the relative strength of the units. People expect to win a 90% odds fight every time. (I'm also currently partway through reading Thinking In Bets, which has some interesting things to say about this sort of thing.)

AussieWog93 6y ago
Fire Emblem is one of the games that shows fake probabilities to get around this problem - if the actual probability of an attack connecting is 95%, it will display a value of 80-85% to the user. The odd part is that I've played hundreds of hours of those games and never noticed anything wrong.
- cma 6y ago
  Maybe people expect it to be one some kind of logarithmic scale.
hutzlibu 6y ago
I am reminded about XCOM, which some players complained about being unfair, too. In fact, the game is cheating - but in favor of you. (first enemy critical hits get negated and after someone of your troop dies, the rest gets better chances of surviving and so on) But actually, I don't like that behavior either ..
avianlyric 6y ago
If I remember correctly Sid Meier’s point was a little more nuanced than that. It wasn’t that people expect to win 90% of the time, it’s that humans have a very poor intuition for odds and statistics. Pretty much everyone makes the common mistake of not understanding how a 33% chance of winning impacts independent events. People assume that 33% chance of winning, means a pretty such sure fire win after 3 attempts (after all 33% * 3 is approx 100%). But every battle in Civilization is an independent event, so to get a sure fire win (90% confidence) you need to fight something 6/7 times (I don’t know the exact number, been some time since I studied this). Having the number of fights you win against an opponent be 2 or 3 times greater than the intuitive number of fights makes people think the computer to cheating. When really they just don’t understand how stats work. To fix this the Civilization games actually skew their stats so they better align with human intuition. So if the game says you have a 33% chance of winning, then that means you’ll probably win after 3 fights. Despite the fact that really you have a 40%-50% chance of winning that fight.
- devonkim 6y ago
  Sid Meier also worked with the Firaxis folks on the recent XCOM reboot and this is well known and documented now how the stats work out in the game and how the game will cheat for the player eventually on lower difficulties. Nothing more frustrating than lining up shots perfectly and missing a 98% point blank shot from 2 feet away. I’ve always wondered what that 2% could be - gun jams, the character slips on a banana peel, who knows?
  - username90 6y ago
    > . I’ve always wondered what that 2% could be - gun jams, the character slips on a banana peel, who knows? Or just, the other person moving around in a hectic life or death situation? Trying to shoot someone in short range isn't 100% hit. The characters calmly standing there waiting to get shot is just a game representation.
smichel17 6y ago
Counter-hypothesis: these games' higher level gameplay loops demand a higher level of consistency; players' mismatched expectations come partially from a place of "if that chance is actually 90%, then what you're asking me to do is impossible". Let me be more concrete. Take XCOM as the clearest example. On one hand, your troops level up with experience. A veteran soldier is more effective than a new recruit. So you need your them to survive, otherwise you'll be at a significant disadvantage on later missions. On the other, the developers want there to be some difficulty, to create a sense of tension (that you might not succeed) and make strategy matter. These two goals are diametrically opposed, given the genre. It is difficult-to-impossible to create a scenario where there is both a real risk of losing and where a player can recover from losing a nominally 90% roll. Maybe once per mission. To keep the gameplay balance intact, devs would need to raise the nominal chances to match the actual chances, not the other way around. The problem is, doing this would make it obvious to the player how much stronger they are than their opponents, in a game where you are supposedly vastly outgunned. So this type of fudging the numbers is more of a psychological trick by the developers, who want to have their cake and eat it, too. It's in the same class as shooters which prevent the AI from shooting you in the back or sniping you from offscreen, or horror games where the monster moves more slowly when you're not looking at them. An illusion to create a certain type of experience. Note: I don't dispute that people are bad at judging probabilities. But I also ask that you consider that players may use this reasoning (probably subconsciously): "The developer wouldn't set me an impossible goal. If I lose this battle, it will be impossible to reach the goal they have set. Therefore, my chance of winning must be 100%. The game says it is 90%. Therefore, when the game says 90%, it must actually mean 100%. Therefore, when I lose a roll that the game says is 90%, the game is cheating." They don't really mean the game is cheating, they mean the developer is setting them an unfair task, giving them the illusion that there is a chance at victory through smart play, when in fact there is none. I imagine you could test this hypothesis, or rather, how large of an effect it has, by taking the first 11 letters of the alphabet, randomizing their order, and then assigning them to probabilities 0-100% in steps of 10. Then have people play the game, showing only the letters. Then at the end, ask them to guess what each letter's probability is. This should show how much of the delta is the developer's fault vs player's. I think you'd want to do it with players whose perceptions have not already been poisoned by the lying of previous games, and use a less uniform distribution, so players can't guess the % just by figuring out the relative order.

by GlennS 6y ago

AussieWog93 6y ago

Fire Emblem is one of the games that shows fake probabilities to get around this problem - if the actual probability of an attack connecting is 95%, it will display a value of 80-85% to the user. The odd part is that I've played hundreds of hours of those games and never noticed anything wrong.

cma 6y ago
Maybe people expect it to be one some kind of logarithmic scale.

hutzlibu 6y ago

I am reminded about XCOM, which some players complained about being unfair, too. In fact, the game is cheating - but in favor of you. (first enemy critical hits get negated and after someone of your troop dies, the rest gets better chances of surviving and so on) But actually, I don't like that behavior either ..

avianlyric 6y ago

If I remember correctly Sid Meier’s point was a little more nuanced than that. It wasn’t that people expect to win 90% of the time, it’s that humans have a very poor intuition for odds and statistics. Pretty much everyone makes the common mistake of not understanding how a 33% chance of winning impacts independent events. People assume that 33% chance of winning, means a pretty such sure fire win after 3 attempts (after all 33% * 3 is approx 100%). But every battle in Civilization is an independent event, so to get a sure fire win (90% confidence) you need to fight something 6/7 times (I don’t know the exact number, been some time since I studied this). Having the number of fights you win against an opponent be 2 or 3 times greater than the intuitive number of fights makes people think the computer to cheating. When really they just don’t understand how stats work. To fix this the Civilization games actually skew their stats so they better align with human intuition. So if the game says you have a 33% chance of winning, then that means you’ll probably win after 3 fights. Despite the fact that really you have a 40%-50% chance of winning that fight.

devonkim 6y ago
Sid Meier also worked with the Firaxis folks on the recent XCOM reboot and this is well known and documented now how the stats work out in the game and how the game will cheat for the player eventually on lower difficulties. Nothing more frustrating than lining up shots perfectly and missing a 98% point blank shot from 2 feet away. I’ve always wondered what that 2% could be - gun jams, the character slips on a banana peel, who knows?
- username90 6y ago
  > . I’ve always wondered what that 2% could be - gun jams, the character slips on a banana peel, who knows? Or just, the other person moving around in a hectic life or death situation? Trying to shoot someone in short range isn't 100% hit. The characters calmly standing there waiting to get shot is just a game representation.

smichel17 6y ago

Counter-hypothesis: these games' higher level gameplay loops demand a higher level of consistency; players' mismatched expectations come partially from a place of "if that chance is actually 90%, then what you're asking me to do is impossible". Let me be more concrete. Take XCOM as the clearest example. On one hand, your troops level up with experience. A veteran soldier is more effective than a new recruit. So you need your them to survive, otherwise you'll be at a significant disadvantage on later missions. On the other, the developers want there to be some difficulty, to create a sense of tension (that you might not succeed) and make strategy matter. These two goals are diametrically opposed, given the genre. It is difficult-to-impossible to create a scenario where there is both a real risk of losing and where a player can recover from losing a nominally 90% roll. Maybe once per mission. To keep the gameplay balance intact, devs would need to raise the nominal chances to match the actual chances, not the other way around. The problem is, doing this would make it obvious to the player how much stronger they are than their opponents, in a game where you are supposedly vastly outgunned. So this type of fudging the numbers is more of a psychological trick by the developers, who want to have their cake and eat it, too. It's in the same class as shooters which prevent the AI from shooting you in the back or sniping you from offscreen, or horror games where the monster moves more slowly when you're not looking at them. An illusion to create a certain type of experience. Note: I don't dispute that people are bad at judging probabilities. But I also ask that you consider that players may use this reasoning (probably subconsciously): "The developer wouldn't set me an impossible goal. If I lose this battle, it will be impossible to reach the goal they have set. Therefore, my chance of winning must be 100%. The game says it is 90%. Therefore, when the game says 90%, it must actually mean 100%. Therefore, when I lose a roll that the game says is 90%, the game is cheating." They don't really mean the game is cheating, they mean the developer is setting them an unfair task, giving them the illusion that there is a chance at victory through smart play, when in fact there is none. I imagine you could test this hypothesis, or rather, how large of an effect it has, by taking the first 11 letters of the alphabet, randomizing their order, and then assigning them to probabilities 0-100% in steps of 10. Then have people play the game, showing only the letters. Then at the end, ask them to guess what each letter's probability is. This should show how much of the delta is the developer's fault vs player's. I think you'd want to do it with players whose perceptions have not already been poisoned by the lying of previous games, and use a less uniform distribution, so players can't guess the % just by figuring out the relative order.