5 ms·

The worst part is that even if the algorithm perfectly computed probabilities and was completely open, humans would take away incorrect information. I am remin

The worst part is that even if the algorithm perfectly computed probabilities and was completely open, humans would take away incorrect information. I am reminded of this talk by Sid Meier: https://youtu.be/bY7aRJE-oOY?t=19m27s https://youtu.be/bY7aRJE-oOY?t=19m27s People don't understand probabilities. They don't accept that if someone has a 75% chance of being guilty, they may very well not be. It happened again with FiveThirtyEight when everybody accused their "70% chance of Mrs Clinton winning" of being wrong, when it doesn't preclude Mr Trump winning.

FullMtlAlcoholc 10y ago
> They don't accept that if someone has a 75% chance of being guilty, they may very well not be. Nothing drove this point home to me more than X-COM: EU. Missing 3 successive 90% shots at the worst possible time would cause me to fling my keyboard across the room, but eventually, I learned to accept 90% isn't 100% :).
- Gravityloss 10y ago
  Well, considering the bayesian, I'd consider it likely there to be a bug, either in the probability display or the outcome calculation...
  - csydas 10y ago
    Nah, it's just how these games are. The first two fire emblem games on GameCube/wii were merciless in the same way on the normal modes. 80% hit chance sounds good until you miss and the entire battle goes south because of it. Since these games had permadeath, a miss could mean restarting if you didn't just hit reset mod battle. Sometimes those odds are just those odds. A lot of gaming which uses pseudorandom had given a lot of people the gamblers fallacy as their understanding of probability. In some games it perhaps makes sense; in league of legends I believe the Critical hit chance is pseudo random to feel less swingy, which is probably healthier for a competitive game. Diablo 3 weights it's unique drop rates to ensure that players get a steady supply of relevant uniques since there is no trade economy in d3 anymore. This also makes sense in isolation. Other games though that don't have as strong of a reason to fudge the roll give a really bad impression of random to players.
    - foota 10y ago
      You might consider using a term other than pseudo random for that, since pseudo random number generators generally try to be uniformly random, as opposed to the sort of thing you're talking about.
      - madenine 10y ago
        The term comes from the games community - its what they call it. For example, in DotA: http://wiki.teamliquid.net/dota2/Pseudo_Random_Distribution http://wiki.teamliquid.net/dota2/Pseudo_Random_Distribution
        foota 10y ago
        Interesting, thanks for making that point. I think it might be best to refer to one as a prng and the other as a prd (generator vs distribution) although implementing a prd creates a prng, some prng attempt to target a real random distribution.
        csydas 10y ago
        Late reply but thanks for the heads up - I wasn't aware that prng was a term that I could be using as opposed to prd. The above comment is accurate in how I first came across the term as it's pretty commonly used in a lot of games in a lay person's fashion.
    - slavik81 10y ago
      Most Fire Emblem games have misleading hit percentages. When it says 80%, you really have a 92.2% chance to hit. http://web.archive.org/web/20141004060133/http://old.serenesforest.net/general/truehit.html http://web.archive.org/web/20141004060133/http://old.serenes...
- golergka 10y ago
  And that's why it's a bad idea for game designers to expose honest probabilities as naked numbers to the players.
  - FullMtlAlcoholc 10y ago
    Why? Does something like 90% get rounded up to 100%?
    - golergka 10y ago
      Because games are created for humans, and humans are bad at understanding probabilities. So when 90% shot fails three times in a row, overwhelming majority of players will think that the game is cheating.
- Dylan16807 10y ago
  While you learned a good lesson, hit percentages in games are very often lies. I'm seeing conflicting data for X-COM: EU specifically. Some say it's fair (except for cheating in your favor when you're losing on easier modes), but this guy collected a quite-significant number of samples and got results that looked like the square of the displayed hit chance: https://steamcommunity.com/app/200510/discussions/0/613937943073634826/ https://steamcommunity.com/app/200510/discussions/0/61393794...
Mathnerd314 10y ago
The 70% probably was wrong, in the sense of being uncalibrated. 538 did their best to fit the data to past elections, but there haven't really been enough elections to fit a good model.
- SixSigma 10y ago
  Plus, of course, Nate was no longer an impartial observer, he has moved over to punditry. https://wikileaks.org/podesta-emails/emailid/28878 https://wikileaks.org/podesta-emails/emailid/28878 From:brentbbi@webtv.net To: john.podesta@gmail.com Date: 2015-07-09 09:44 Subject: CNN and bed-wetting > And when I see Dan, Nate Silver and various other insider pundits saying what they are saying, I have been to enough rodeos to know that thoughts are being planted from the Obama-Clinton consultant class (not you, and you know who I mean). This is not helpful to Hillary, quite the contrary, Bernie has zero chance of being nominated and Hillary is going to need every liberal supporting her to vote, which is why I am being very friendly to Bernie and will have some credibility when the time comes to encourage his people to get on the train with Hillary. Examples of his work https://i.sli.mg/fYycbE.jpg https://i.sli.mg/fYycbE.jpg
  - Houshalter 10y ago
    There's no question that 538 is anti-Trump. That shouldn't affect their statistical model, which gave much higher odds to Trump than most other models. I don't see the relevance of either of those links.
    - SixSigma 10y ago
      The point of the other links is that Nate is compromising his professional image through punditry. If I asked a consultant for advice on the future of my sector of the economy and then found a series of blog posts bashing my industry which were all unfounded, I would call his judgment into question. At best, he is trying to influence the very thing he is supposed to me impartially analysing.

FullMtlAlcoholc 10y ago

> They don't accept that if someone has a 75% chance of being guilty, they may very well not be. Nothing drove this point home to me more than X-COM: EU. Missing 3 successive 90% shots at the worst possible time would cause me to fling my keyboard across the room, but eventually, I learned to accept 90% isn't 100% :).

Gravityloss 10y ago
Well, considering the bayesian, I'd consider it likely there to be a bug, either in the probability display or the outcome calculation...
- csydas 10y ago
  Nah, it's just how these games are. The first two fire emblem games on GameCube/wii were merciless in the same way on the normal modes. 80% hit chance sounds good until you miss and the entire battle goes south because of it. Since these games had permadeath, a miss could mean restarting if you didn't just hit reset mod battle. Sometimes those odds are just those odds. A lot of gaming which uses pseudorandom had given a lot of people the gamblers fallacy as their understanding of probability. In some games it perhaps makes sense; in league of legends I believe the Critical hit chance is pseudo random to feel less swingy, which is probably healthier for a competitive game. Diablo 3 weights it's unique drop rates to ensure that players get a steady supply of relevant uniques since there is no trade economy in d3 anymore. This also makes sense in isolation. Other games though that don't have as strong of a reason to fudge the roll give a really bad impression of random to players.
  - foota 10y ago
    You might consider using a term other than pseudo random for that, since pseudo random number generators generally try to be uniformly random, as opposed to the sort of thing you're talking about.
    - madenine 10y ago
      The term comes from the games community - its what they call it. For example, in DotA: http://wiki.teamliquid.net/dota2/Pseudo_Random_Distribution http://wiki.teamliquid.net/dota2/Pseudo_Random_Distribution
      - foota 10y ago
        Interesting, thanks for making that point. I think it might be best to refer to one as a prng and the other as a prd (generator vs distribution) although implementing a prd creates a prng, some prng attempt to target a real random distribution.
        csydas 10y ago
        Late reply but thanks for the heads up - I wasn't aware that prng was a term that I could be using as opposed to prd. The above comment is accurate in how I first came across the term as it's pretty commonly used in a lot of games in a lay person's fashion.
  - slavik81 10y ago
    Most Fire Emblem games have misleading hit percentages. When it says 80%, you really have a 92.2% chance to hit. http://web.archive.org/web/20141004060133/http://old.serenesforest.net/general/truehit.html http://web.archive.org/web/20141004060133/http://old.serenes...
golergka 10y ago
And that's why it's a bad idea for game designers to expose honest probabilities as naked numbers to the players.
- FullMtlAlcoholc 10y ago
  Why? Does something like 90% get rounded up to 100%?
  - golergka 10y ago
    Because games are created for humans, and humans are bad at understanding probabilities. So when 90% shot fails three times in a row, overwhelming majority of players will think that the game is cheating.
Dylan16807 10y ago
While you learned a good lesson, hit percentages in games are very often lies. I'm seeing conflicting data for X-COM: EU specifically. Some say it's fair (except for cheating in your favor when you're losing on easier modes), but this guy collected a quite-significant number of samples and got results that looked like the square of the displayed hit chance: https://steamcommunity.com/app/200510/discussions/0/613937943073634826/ https://steamcommunity.com/app/200510/discussions/0/61393794...

Mathnerd314 10y ago

The 70% probably was wrong, in the sense of being uncalibrated. 538 did their best to fit the data to past elections, but there haven't really been enough elections to fit a good model.

SixSigma 10y ago
Plus, of course, Nate was no longer an impartial observer, he has moved over to punditry. https://wikileaks.org/podesta-emails/emailid/28878 https://wikileaks.org/podesta-emails/emailid/28878 From:brentbbi@webtv.net To: john.podesta@gmail.com Date: 2015-07-09 09:44 Subject: CNN and bed-wetting > And when I see Dan, Nate Silver and various other insider pundits saying what they are saying, I have been to enough rodeos to know that thoughts are being planted from the Obama-Clinton consultant class (not you, and you know who I mean). This is not helpful to Hillary, quite the contrary, Bernie has zero chance of being nominated and Hillary is going to need every liberal supporting her to vote, which is why I am being very friendly to Bernie and will have some credibility when the time comes to encourage his people to get on the train with Hillary. Examples of his work https://i.sli.mg/fYycbE.jpg https://i.sli.mg/fYycbE.jpg
- Houshalter 10y ago
  There's no question that 538 is anti-Trump. That shouldn't affect their statistical model, which gave much higher odds to Trump than most other models. I don't see the relevance of either of those links.
  - SixSigma 10y ago
    The point of the other links is that Nate is compromising his professional image through punditry. If I asked a consultant for advice on the future of my sector of the economy and then found a series of blog posts bashing my industry which were all unfounded, I would call his judgment into question. At best, he is trying to influence the very thing he is supposed to me impartially analysing.