07-29-2017, 07:24 PM | #41 | |
Join Date: Oct 2005
|
Re: I was hoping for this
Quote:
The AI will likely need to at least use some way to generalize better, and may need some clever tricks from Gordon. (Hopefully the crew at Auroch won't resort to tweaking the odds to keep the game interesting. That'd be disappointing.) |
|
07-30-2017, 10:58 AM | #42 |
Join Date: Aug 2004
Location: Pacheco, California
|
Re: I was hoping for this
Does anybody else think that an AI that just considers the moves that will generate the greatest potential VP gain over the current turn and the enemy response turn will lose badly?
__________________
-HJC |
07-30-2017, 11:08 AM | #43 |
Join Date: May 2012
|
Re: I was hoping for this
AIs that look ahead only one turn do usually lose to more perceptive opponents. The trick is to look at all of the current possible moves, AND all of the opponents possible responses to each of those, AND all of your possible responses to each of those possible responses, etc, as far ahead as you have memory and processing power, and pick the immediate move that leads to the best result X turns ahead, even if it's not the best immediate result. This is called "min-maxing" because you minimize your opponents score and maximize your own on each iteration of the loop.
|
07-30-2017, 11:12 AM | #44 | |
Join Date: May 2007
|
Re: I was hoping for this
Quote:
Edit: And HeatDeath explained it a lot better. |
|
07-30-2017, 12:43 PM | #45 |
Join Date: Aug 2004
Location: Pacheco, California
|
Re: I was hoping for this
When I get a copy I'll quickly find out if it's burdened with a lot of common sense rules like say "Don't run your GEVs into swamp".
Hint: If the computer knows for a fact that I'll never do this then I'll win Breakthrough in the northeast corner.
__________________
-HJC Last edited by hcobb; 07-31-2017 at 12:45 PM. Reason: An example |
07-31-2017, 09:52 AM | #46 |
Moderator
Join Date: Jun 2017
Location: Bristol, UK
|
Re: I was hoping for this
It's fantastic to see the enthusiasm for the adaptation, and to hear your feedback and technical questions. What we're going to do is gather up these questions, for example those regarding how the AI will work, and then address them all in one big post after interviewing the designers and programmers. We want to ensure that your questions are answered in as much detail as possible, and the level of technical knowledge from these guys will hopefully do so :)
We hope this is OK for everyone. |
07-31-2017, 10:49 PM | #47 | |
Ogre Line Editor
Join Date: Sep 2006
Location: Plainfield, IL
|
Re: I was hoping for this
Quote:
Here's some more to chew on... New Stuff - I'd like to think the Core game will be as complete as possible (i.e., be a reflection of the ODE ruleset), but what kind of things are expected for DLC? I can understand things like paint schemes, factions, etc being DLC, but I'm personally hoping that rules are not DLC; it would be a bummer if you had to purchase rules to make a "complete" game. Overlays - Being able to modify maps with terrain overlays is a big deal that was introduced with ODE. Will overlays be in the computer version?
__________________
GranitePenguin Ogre Line Editor |
|
08-01-2017, 05:19 AM | #48 |
Join Date: Aug 2004
Location: Carrboro, NC
|
Re: I was hoping for this
As long as you're gathering questions...
What can they tell us about how they bridge the gap between strategy and tactics? How does the system plan its strategy for a given (or previously unseen!) scenario? How does it know to hold this town, abandon this outpost, or make a swift strike up a particular road to catch the enemy out of position? The simplest approach is to ignore strategy, and min/max (A/B tree, whatever) for as many future turns as resources allow. Not too satisfying in the long run, but very flexible. Built-in heuristics could be made to work really well for predefined scenarios, but may not be very adaptable to user made scenarios. There could be a layer of static analysis at the start of the scenario, which produces an abstracted view from which it devises a strategy. This could be run again during the game to adapt and refine the system's objectives, as long as it doesn't switch back and forth between goals (analysis paralysis!) so often as to be ineffective. Anyhow, the "how to do" (tactical) problems are solvable in a variety of ways, but how are they approaching the "what to do" (strategical) issue? I'm looking forward to the game, and thanks for posting here. Last edited by tomc; 08-18-2017 at 04:14 PM. |
08-01-2017, 11:26 AM | #49 |
Join Date: Oct 2005
|
Re: I was hoping for this
Riffing on GP's question: To what extent will things be customizable?
One of the great strengths of Ogre as a board game is that the game is simple and so is easy to tweak. Can we edit maps? To what extent can we tweak scenarios? To what extent can we pick and choose rules? To what extent can we program the AI ourselves? Can we create new maps from whole cloth? Can we create new scenarios ? With new victory conditions? With special rules? Can we create new units? With special rules? --- And one really basic question - how will combined fire be handled in the UI? |
08-02-2017, 04:52 AM | #50 | |
Moderator
Join Date: Jun 2017
Location: Bristol, UK
|
Re: I was hoping for this
Quote:
Completely agree with your points above that one of Ogre's many strengths is the ease with which you can modify the rules and settings to suit the preferences of the player. Indeed, that's most likely one of the main reasons it's enjoyed such a long shelf life, and has as many fans worldwide as it does. It's definitely something we're trying to retain in the adaptation too. We'll address these questions in more detail when we publish the post from the designer and programmers :) Combined fire is working really well in testing so far. The basic process is; you select your target first, then any friendly units in range are displayed with a highlight, and you can select as many of these to attack as you wish. Units such as Ogres that have more than one weapon have a flyout menu, and you can then select the individual weapons you want to commit to this attack. Thanks! |
|
|
|