Strategic factors

Come in here to talk about your sky-net style world-destroying super bots!

Moderator: Defcon moderators

User avatar
smc
level1
level1
Posts: 21
Joined: Sun Jun 13, 2010 2:18 am

Strategic factors

Postby smc » Tue Jun 22, 2010 11:51 pm

Disclaimer - I'm not an expert DEFCON player, but based on a few games played, and few more observed these are my notes on the important factors that determine a winner in 1v1 play. At the end I've tried to think how they might be applied to writing a bot. Any comments/criticisms welcome.


Primary factors

1. Distributed vs concentrated structure placement

Concentrating structures (particularly silos) to provide overlapping fields of fire produces a huge gain in their resistance to attack. In most cases this heavily outweighs the negative impact of the reduced coverage of the player's cities and the reduced ability to project power (e.g. silo AA, fighter cover, anti-sub bomber cover).

2. Being beaten at sea

If a fleet vs fleet battle produces a clear winner then that winner has direct sea access to his opponents territory. This allows the winner to scout the loser's structures, protect surfaced subs while they launch nukes, and attack cities and structures with sea-based bombers. These advantages often swing a game.

3. Opponent firing silos without cover for enemy subs or fleet AND Having concealed subs/fleet in range when enemy silos launch

This is a conditional factor. One player must leave themselves open to it occurring, the other player must plan for it occurring. A player who fires silos only to have them immediately attacked by subs or bombers (to which he has no immediate defense) will often lose the game.


Secondary factors

1. Aggressiveness

Hard exactly to quantify but it appears that other things being equal that the more aggressive player, who takes the fight to the opponent does better. This may be confounded with player skill - better players may be more aggressive.

2. Harvesting (i.e. picking off undefended enemy cities)

In close games one player can often gain an advantage of the order of 20-30 points by methodically using a small force (often 2-3 carriers, 2-3 subs, or more rarely airbase bombers) to repeated nuke enemy cities that are outside the protection of enemy silos. This usually happens in the middle game while the main fleet combat is happening. It is quite geography dependent and often asymmetric between continents.

3. Sub usage

A significant strategic choice is whether to use subs for naval combat, often revealing their position by using active sonar, or using them as nuclear strike platforms. In naval combat they may be effective at destroying opponent fleet units and/or disrupting opponent fleet formations but are often lost to enemy subs, enemy carriers, or naval nuking. The consequences of this choice, either in winning the naval war or in allowing strikes at launching silos, are often significant.


Tertiary factors

1. Multi-focal structure placement

As noted in the "Primary factors" above, there are disadvantages to concentrating all structures together. A careful balance of structures in different locations, in some circumstances, may convey an advantage. For example in Africa vs Asia the African player might place all silos in extreme West Africa, but all airbases in South Africa with most fleet elements in the Indian Ocean. This allows the airbases to re-arm the fleet with planes (though making the airbases highly vulnerable if the enemy fleet breaks through and manages to scout them).


Confounding factors

1. Close/Intermediate/Far placement

Games have a very different character based on the proximity of the two territories. The most obvious difference is the use of airbase bombers, in close games they can be used to directly attack opponent terrain, while simultaneously being available for naval combat and coastal defence from subs. In far games they must be often be transferred to distant fleets to repeated and efficiently attack enemy cities or structures.

2. Single/Double/Late naval approach

In some combinations of territory (e.g. North America vs South America, or Russia vs either America) there are two mostly independent naval battles. Other match-ups (e.g. EU vs Russia) have only one major naval battle. The distribution of forces between these battles (if more than one) is often significant. Depending how each player splits forces and the skill at naval battles they can achieve naval dominance in one, both or neither of the naval theatres, with the advantages noted above. In these circumstances there may be some profit in game-theoretic approaches to deciding the force balance of the two main naval fleets.

In most match-ups there is a single obvious direct naval route to the opponent, plus one or more "late arriving routes". Sending a force along a late arriving route may allow naval dominance later in the game in one region, at the cost of making it more likely that the opponent will achieve it early in the game in the 'primary' naval region.

In some situations (e.g. Russia vs Asia if Russia places fleet in Artic) there are no direct naval conflicts possible until late in the game.


So then...

A successful bot would need to at least take into account the primary and secondary factors above. Against the built-in CPU a concentrated structure placement would give it a significant advantage, but other competitive bots or humans with a little experience would be expected to do the same.

Naval skill is vital. Naval battles require a mix of all unit five mobile unit types. While general patterns for naval battles exist (e.g. BattleShips in front) the range of options and large set of exceptions would imply that rule-based methods would be less effective than more organic methods. This tactical area of the bots behaviour might require a huge amount of attention.

The ability to move subs into strike range and wait for an opportunity, and the ability to guard its own coastline, while fighting one or more naval engagements require a bot to execute multiple priorities simultaneously. Some of these priorities will compete for resources, necessitating some form of resource (re-)allocation. Some of these priorities are event driven - the bot will need to re-plan and re-allocate resources depending on events.

Strategy will be strongly affected by the distance between continents and routes between them. A brute-force approach might work given that there are only 30 possible match-ups.

The secondary factors identified are more conditional on the details of a particular game that the primary ones. Deciding where these tactics are appropriate depends on an assessment of the opponent's placement of units and, arguably, style of play. Exploiting these secondary factors might require modelling of the opponent and/or inductive reasoning.

All in all...

Writing a competitive DEFCON bot appears to be a interesting project... :)
User avatar
poptart911
level1
level1
Posts: 33
Joined: Thu Jan 13, 2011 6:04 pm
Location: Iowa, US
Contact:

Postby poptart911 » Mon Jan 24, 2011 9:48 pm

To add to the above, it seems like there are about 5 discrete goals for the bot.

> Forming alliances
> Placing Structures
> Conducting air/naval combat
> Finding/Selecting Targets
> Balancing nuclear assault with defense

Oddly enough, there are five discrete stages in the game. (DEFCON 5...1)

Hope this helps someone.

Return to “AI Bots”

Who is online

Users browsing this forum: No registered users and 4 guests