Static defenses, experiences?

[DrClaes]

I notice that some bots are posting "fixed" or static defenses around their home hills, for example at the points directly adjacent to the hill or the diagonal points around the hill. I also noticed that my timed out clumps of 4-5 ants in the maze leading to a (highly visible) hill stumped a tactically very good, but perhaps too cautious opponent (A)...

What experiences have you had with static defenses? I'm going to try some variation of this, but need to rework some of my ant movement code to allow "tunneling" through a wall of stationary ants... my guess is that if you're able to fortify your home base you will fend off quite a few attacks by weaker bots who "get lucky" and find your hill, but eventually succumb--maybe stalling for a ltitle longer--to a stronger enemy who is more successful at building his numbers.

Looking at the problem from the attacker's point of view, part of the challenge is in knowing when to accept losses of your own ants, even for a smaller number of enemy casualties. I'm nowhere near this kind of deep code for my tactics yet--indeed I'm relying on very simple heuristics of backing off when I'm one-on-one, then running for the enemy when there's more than one ant on my side... and running like a maniac for the enemy hill when it's in sight ...

[lorill]

My bot uses diagonal static defenses (not always, depends of the perceived mood of the game). As you guess, it helps with weaker bots getting lucky. I don't lose too much in gathering time, since I don't allow my ants to idle too soon in the game.

It doesn't help against rushers or heavy casualty bots (hi, strcat!). It doesn't help either against good tactical bots, but I still believe it's worthy, even in the top 50.

I might change it to have more of a patrolling squad, which would be less easy to detect, and see enemies farther away, but I have to do some tests to decide if it's any good first.

[Parasprites]

In my old bot, I used a static defence of 3x3/5x5 squares around the mound, but unfortunately, it performed very poorly. In most cases, the bot was unable to collect enough food, and eventually sent the defenders away to gather more, thus defeating the purpose. And of course, it doesn't help much against tactical bots.

In my current bot, I've created a dynamic guarding behavior which I'm happier with. This has the advantage of allowing guards to gather food as well.

[RejectCode]

I was testing out this recently as well and noticed this exact problem; ants that are defending aren't helping unless an enemy comes near. So every turn an ant didn't move towards some objective or guard from an enemy ant it is contributing negative value. I witnessed the same behavior as Parasprites 3x3/5x5, limited usefulness and a quick demise once under the guard count since mine stupidly scramble for food, leaving the hill unprotected. On multi-hill maps this negative value can be seen quite quickly and has always ended poorly for me even against the HunterBot/GreedyBot.

Dynamic guarding sounds interesting, glad to hear it's better.

[dimkadimon]

In my opinion the usefulness of a static defense is completely dependent on the level of bots you are playing. If an opponent bot does not know how to fight and comes to your hive in a long string of ants then your static defense (of 4) can kill all of his ants easily. On the other hand, a bot who knows how to fight should be able to destroy your static defense with even a few ants. Here is an example of how my bot uses 3 ants to take down a static defense of AVAVT (turns 350-383):

http://aichallenge.org/visualizer.php?game=109026

If you are playing against bots who know how to fight then a static defense is a bad thing, because you can make much better use of those "static" ants and ask them to fight dynamically.

[DrClaes]

I agree with what's been said here, and have spent some time looking more closely at my own and others' games. The static defenses right around the nest are generally only good against naive bots which come at you single file, and won't get anyone into the top 100 here... there will be exceptions where clever bots will be too cautious to attack static formations of ants, but I haven't worked out what formations might prove most useful in that respect.

I've got some thoughts for setting up a "patrol" system around my hills, with a number of points which must be "spotted", and where their attractiveness grows the longer it's been since they were last seen... a set of 6 of those points spaced evenly around the nest at some appropriate radius (2 viewradii? 1.5?) would provide some spotting in all directions. Most bots coming your way won't know your hill location until they stumble on it, regardless of algorithms for symmetry detection (maybe I'm wrong), and being able to keep them out of view will keep you alive and spawning ants that much longer.

/CW

[mac]

The static defenses right around the nest are generally only good against naive bots which come at you single file, and won't get anyone into the top 100 here...

Agree with this (and also with the rest of your post), yet your sentence could be formulated as:

"Static defences work 98.6% of the times"

[tmc]

But if you aim for 100th place now, you might find that you were aiming what will be 400th place in the final tournament.

Static defenses are a waste of ants.

[BenJackson]

A waste? "Static defense" is the only way to have visibility of your hill and it's your only hope of having ants in the area to defeat lone scouts.

[mac]

A waste? "Static defense" is the only way to have visibility of your hill and it's your only hope of having ants in the area to defeat lone scouts.

I think you missed the static bit of the "static defense" expression!

Nobody says that defending the hill keeping some fighter around it is a waste of ants, but keeping your ants frozen in a position is probably not the most effective way to use them...