Wednesday 5 March 2014

Frequentist v Bayesian Statisticians



If you are not an XKCD reader then you are missing out on some of the wittiest and geekiest humour on the net.

Statistics is a useful tool, but terms and conditions apply, and those are that most people don't understand statistics and what they mean. This includes gamers too, as most don't seem to understand that while probabilities of a number coming up that it doesn't mean that the conclusion one should draw is to always follow the numbers.  See the Frequentist side of the cartoon for clarification.  As gamers we would in my opinion be better served if we put a Bayesian hat when calculating strategies when based on the probablitiy of any number being rolled on a D6.

Just saying, especially at Ogre players.
  

7 comments:

  1. There are lies, damn lies and every said by Nigel Farage?

    ReplyDelete
    Replies
    1. It's even more complicated than that, but yes politicians lie, mostly by omission.

      Delete
  2. I suspect that anyone who's done any sort of gaming for a while knows that, while you follow the odds, you don't rely on them.

    Consider OGRE, where a 1:1 attack gives you 2/6 chance of a D, 2:1 3/6 and 3:1 4/6. And to take an obvious example, say you're attacking secondary batteries with three missile tanks. You can work out the odds: all on one gives you 0: 33%, 1: 67%. One on each of three gives you 0: 30%, 1: 44%, 2: 22%, 3: 4%. So you've got a better chance of killing at least one battery by splitting your fire, and so that's what you do. But when all those dice come up 1s and 2s, you're not surprised.

    ReplyDelete
    Replies
    1. I disagree with you at the point where you say, "and so that's what you do."

      My answer would be it all depends on what stage you are it in the game. Turn one of combat when the Ogre still has its main battery, then no. A toothless Ogre, then yes.

      There is IMNSHO a strategy for using probabilities effectively, and that is driven by the consequences of failure. Ignoring costs means one is relying on luck, I prefer strategy and tactics; and these change during the game from the dynamic of the Ogre as a big hammer of concentrated force that wants to defeat in detail the overwhelming numbers of the defence versus the needs of the defence to concentrate and defeat the weapon arrays of the Ogre.

      I hope that is a clear explanation?

      Delete
    2. Yes, I was implicitly assuming that the immediate objective was "kill as many SBs as possible" to keep the complexity down. Otherwise I'd be showing off my full how-to-play-Ogre algorithm, which is very elegant but will not fit in this comments box. :-)

      Delete
  3. Put this way, the Bayesian simply can't lose the bet. ;-)

    Ogre demonstrates the idea of risk very well. You might have a chance of destroying more SB by splitting fire, but you run the risk of having those guns shoot back at you if you fail. Going for the more certain attack may be the better choice.


    ReplyDelete
    Replies
    1. Good to see you making a comment here, seems like a long time since we last exchanged thoughts. I have got a draft of something about BattleTech I'm going to post soon, but I've been a bit consumed by the need to get my Ogre forces up to snuff to play demo games, as you can probably tell by all the recent posts. ;-)

      Delete