Jump to content

  • Log In with Google      Sign In   
  • Create Account


Member Since 27 Aug 2003
Offline Last Active Today, 07:16 AM

Posts I've Made

In Topic: When is it okay for a player to get "stuck"?

Yesterday, 05:24 PM

Failure of a puzzle game really should either be effectively instant, such as falling into the acid in portal, or should allow the user to seamlessly return to the beginning if they feel they may have messed something up.


Add a 'teleport pad' or something beyond the 1 way door, which carries the user back to the initial part of the level, and clearly tells them that the task has been reset. 

In Topic: OGRE Bicycles

22 December 2014 - 02:04 PM

Remind me never to visit if I want to buy a bike.

 Around here there are hundreds of bikes under $250 USD  .



I'm sure you can get lots of cheap bikes in Europe as well. However they'll be much like the cheap ones you can get in North America: Insanely heavy frames for the stiffness and durability, cheap gears, cheaper shifters, cheap wheels, horrendous breaks, and frequently using non-standard parts and mounting brackets that are hard to fix/replace/upgrade.


Two of my friends bike a lot, and one laughed at the other for spending $3k on a bike nearly 10 years ago when the other was riding something that cost around $300. He has now gone through nearly $2000 in 'cheap' bikes, while the guy who bought the expensive one in the first place has put a few hundred into it, and has enjoyed a far superior bike for the last decade.

In Topic: Development in 4k

18 December 2014 - 08:25 AM

I've tried working with single large monitors, but I generally go back to a multi-monitor setup rather quickly. For one I just find it far easier on the eyes in the long term when I'm moving back and forth between different focus points, so I keep my monitors at slightly different distances.


I tend to go for a single 'large' monitor in my setups as the main central screen, and then flank it with a smaller cheaper monitor, ideally a matched pair on either side, and with the option to rotate them between portrait and landscape modes. 

In Topic: AI in 4X

16 December 2014 - 01:52 PM

Just trying to throw out a bunch of questions to hopefully drive some more conversation and input from others as well:


So will Tier 1 continue to demand a given planet, regardless of resource cost to get it? Can it decide that "Hey, attacking this planet from the large empire next to us is kind of a bad idea, lets go after some of their smaller enemies instead? The enemy of my enemy is my friend, and maybe they won't attack US instead if we're helping them?"

>Edit: How does the AI see things if they change, such as that small single planet opponent just got annexed by BIG_BAD_SCARY_EMPIRE_3?


How do the various levels of AI gain information to base their choices on? How do they calculate potential profit? Risk vs Reward?


How are the actions/threats of other "Players" accounted for, if at all? Does the AI have the ability to make educated guesses of what other Players or AIs are working towards, and decide if they pose a real threat? (ie, do they focus just on growing in a relative vacuum and expand by picking off the easiest targets nearest them, or do they evaluate everyone around them to prioritize where they attack/expand? A simple 'easiest growth path first', or would "Large Empire A" ever attack "Small Empire B" instead of the easier target "Small Empire C" just so as to deny "Large Empire D" from easily advancing into "B"'s planets?)

In Topic: AI in 4X

16 December 2014 - 08:14 AM


I like your idea! If I understand it correctly there will be set of "advisors" on many levels, that can deliver their requests to the higher level for consideration, right?
No... Not exactly. The order is given by high levels and executed by low levels. So it's the other way round (the high levels giving advices/guidelines/gloas for low levels).


The lower level could give the feedback of the orders received, but that's just an extra (probably not needed really).




Every turn your goal values would be updated to help decide the next move
No, no, no. That's the thing I want to break from. Thinking in terms of "where to move units", instead I want thinking in terms of "what the AI wants to archieve" (territory focused, not units focused). Where exactly move units is the last (lowest level) step and not that important.


I don't want it min-max algorithm driven, where the AI reevaluates the situation each turn. I want it data driven where the AI builds on already existing past orders.


"Next move" doesn't have to be as low level as an actual unit move, but can also be an update to your grand strategy target.


However, I think you're going to hit a rather awkward wall if you try to break away from weights and values map evaluations. What exactly is your AI going to use to make choices? Arbitrary dice roll? You need some means to map out data and come to conclusions. Humans do the same thing, but are far more arbitrary and imprecise in our evaluation of those numbers. 



So you have your highest level AI. Describe how it makes a decision. Break it down into pusdo-code level and explain [i]how[//i] it is deciding to attack planet A or B.