Talk:Minimax/Archive 2

Archive 1

Archive 2

Another Wiki page that is effectively worthless

The block that is called "Lua Example" has no description. Is it how to build the tree or traverse the tree. No code comments and a reader is just supposed to work it out for themselves. Thus, why bother inserting the block at all.

And as to the other text. It is written by someone who understands the method/algorithm but has no ability in passing on this knowledge to other readers.

Another effectively worthless piece of technical description. — Preceding unsigned comment added by 81.187.174.42 (talk) 16:02, 9 August 2011 (UTC)

error in pseudocode

I strongly believe that the negamax pseudocode has an error. When exiting at a terminal node, this possibly happens at a depth <> 0. Then, the returning heuristic value should have a minus, if depth is an odd number. The correct algorithm at that point, in my opinion, is that: if node is a terminal node or depth <= 0:

       return the heuristic value of node

if depth = 0:

       return the heuristic value of node

if node is a terminal node

      if (depth is a even number):
               return the heuristic value of node
       else            
                return (-1) * the heuristic value of node  — Preceding unsigned comment added by 155.207.112.128 (talk) 09:42, 23 August 2011 (UTC) 

pseudocode suggestion

for faster parse ::

 max(a,b) == min(-a,-b)

based on code from negamax

 function minimax(node, depth)
   if node is a terminal node or depth = CutoffDepth
       return the heuristic value of node
   let α := -∞                                   (* set to min, so that max is found *)
   foreach child of node                         (* evaluate *)
       α := max(α, -minimax(child, depth+1))
   return α

it isn't going to be oldschool if else minimax then, is it. but isn't it better ? given min(a,b)=max(-a,-b) fact

---

ca1 16:19, 3 April 2008 (UTC)

It probably won't result in faster code, considering the cost of having two times as many function calls. But it is a nice idea and simplifies the pseudocode. However, I think what you mean is max(a,b) = -min(-a,-b). I changed the text accordingly. --Zvika (talk) 05:28, 4 April 2008 (UTC)

This is not minimax, this is negamax, and while the algorithims return the same result, and performs equivalent computation, this is not the actual minimax algorithm. Additionally, it must be noted that for Negamax, the heuristic depends on the current player to move (it must be negated when current player is Min), while in Minimax it does not. Also, Negamax must be severely altered when turns are not guaranteed to alternate. As such, I believe this pseudocode should be labeled as negamax, or reverted to actual minimax code. 131.151.90.233 (talk) 19:18, 20 March 2009 (UTC)

I also agree and have made the change back to minimax code. GamePlayerAI (talk) 15:19, 3 October 2013 (UTC)

Graph theory

This term is also used in graph theory, e.g. [1]. — Preceding unsigned comment added by Subcelestial (talk • contribs) 22:09, 22 February 2015 (UTC)

Clarity in example

The first example uses the word "choices" in a way that looks a lot like it should be "options". The word "choice" in the context of selections is ambiguous. When you say the player "has three choices" it could mean "has three options" or "makes three selections". It's too early in the morning for me to read the whole thing, but if someone familiar with it could clarify it that would be grand. 67.168.176.62 (talk) 15:30, 27 February 2016 (UTC)

When minimax fails

This article would be considerably improved by a section on games/situations where minimax fails to give a good strategy, e.g. the prisoner's dilemma. As first aid I have added a See Also wikilink to Tit for Tat, the alternative strategy that is preferred in the prisoner's dilemma & many other games and interactions. --D Anthony Patriarche (talk) 05:16, 28 September 2018 (UTC)

maximin vs minimax

I think there's already too much going on in this article (minimax equilibrium concept; for statistics; for social welfare...). In any case, I think I want to un-redirect maximin back to its own article. In a few cases (such as zero-sum games) maximizing the minimum of one thing (maximin) is equivalent to minimizing the maximum of another thing (minimax). But it's certainly not always the case.

There is enough room in this article for both, with reorganization, but would it make more sense to have them both in the same article, or separate? Cretog8 (talk) 03:44, 4 June 2008 (UTC)

why isn't it always the case? minimax of a loss function f is always the same as maximin of the utility function -f. Zvika (talk) 04:56, 4 June 2008 (UTC)

I'm thinking in game theory terms. In a zero-sum game, minimax is usually described as minimizing the maximum payoff to the other guy. It could instead be minimizing your own maximum loss, or maximizing your own minimum gain. But I usually see the first description. In a non-zero-sum game, maximin (maximizing your own minimum gain) still has some appeal, and that's the same as minimizing your own maximum loss, but it's not the same as minimizing the other guy's maximum gain. whew!

So, technically, you're right that you can redefine the problem so any maximin is a minimax instead, but I think the exposition would be unpleasant. Probably that's also true for Rawls' stuff. Cretog8 (talk) 05:09, 4 June 2008 (UTC)

There is way too much going on here for someone looking for something specific. — Preceding unsigned comment added by 103.40.80.138 (talk) 09:17, 9 March 2020 (UTC)

cases where minimax does *not* give the optimal strategy

• games where at least one of the players doesn't seem to play rationally (such as when a non-zero-sum game is converted to a zero-sum game by adding "nature": we can often predict exactly what nature would do. Nature hardly ever acts like a "rational" Homo economicus.)
• infinite games

Are there any other cases where minimax is non-optimal? --DavidCary 07:15, 20 November 2005 (UTC)

Prisoner's dilemma for one --D Anthony Patriarche, BSc (talk) 21:58, 26 April 2020 (UTC)

Who invented?

If this is a theory someone had to invent it. Anyone know who? -Ravedave ^{(help name my baby)} 17:45, 25 August 2006 (UTC)

von Neumann and Morgenstern, for starters. Needs a history section; who developed it when; why -- e.g. minimax played an important role in WWII strategy. The whole article seems way out of balance & non-encyclopedic to me; all tech talk with a lot of repetition, no history, applications other than pure game theory, no examples where minimax fails & other context, critical reception…. --D Anthony Patriarche, BSc (talk) 22:18, 26 April 2020 (UTC)