Grobots - How to write the next champion - Warren's Thoughts

Looking at the July 15th tournament a few patterns in the columns are striking:

I see two keys to building a good side:

A good side must be reasonably good at both objectives, and a champion is usually superb at one.

Growing in the absence of other sides

While growing in the absence of other sides sounds easy (and it is), many serious sides fail to grow quickly when no enemies are present. Growth in the absence of enemies is important in real matches, because there are usually areas of the map that are relatively unoccupied. Sides that take advantage of such areas grow faster, allowing them to overwhelm other enemies later.

Doubletime - measure of success

The standard double-time test is rather flawed by testing sides in the absence of enemies, but it still useful measure of performance in the absence of enemies. All of the champions since (and including) Teledont have doubletimes of between 1100 and 3100 (measured using modern versions of the champions). If you ignore the highest and lowest doubletimes (Missile Bacteria and Commune), the range is about 1700-2800. Sides with doubletimes faster than this cannot afford enough defense to keep from becoming food for better-armed competitors. Sides with doubletimes slower than this get swarmed and killed later in the game. I recommend doubletimes 1500-2500.


Autotrophs can reach the upper end of this range, but only by sacrificing enough defense to fail when enemies show up. The time-honored technique for growing in the absence of enemies is to use lightly armed and armored theotrophs that spread out to gather from all available unused territory. It is critical to have one's territory expand to fill all available space; no one has yet built a colony with a doubletime under 2000. I recommend autotrophy only when a side has saturated an area with gatherers.

Growing in the presence of other sides

The techniques for surviving in the presence of other sides are more diverse, and harder to evaluate.

Techniques include (in no particular order)

Run away

Exemplars: innumerable skittish gatherers

If your current location is too crowded, find somewhere else that's less crowded. A very useful tool, but do not rely on it exclusively, since there are not always empty places to flee to.


Exemplars: Productive, Sucker

Blends into running away, but is more pre-emptive.


Exemplars: Bacteria, Eventually (early), Commune (at long range)

You don't always have to distance yourself from enemies to grow well, especially if the enemy has a different preferred food source. This technique works best when near opponents who are willing to cooperate. Co-existance might work against moderately armed opponents

Scare others away

Exemplar: Not Quite Wise

If your enemies prefer the run-away strategy, you can steal their territory by shooting at them to scare them away. A small weapon (5-15% of capital cost) is sufficient to do this.

Kill others in area

Exemplars: Commune and Revenge Killer

If you are better at combat than your opponents, you can kill your enemies and take their food. This technique sometimes works wonders, but has a major drawback: half the time, you are the weaker side. This technique is also negative-sum - it hurts the loser more than it helps the winner.

Distract enemies

Exemplar: Not Quite Harmless

If you offer a target that dodges or shields well, you can force an enemy to co-exist even if it would prefer to kill you.

Combining techniques

The agressive and passive techniques do not mix well. All the robots in an area should decide, together, on compatible strategies.

Other suggestions

Long-range guns are a supplement

Long-range guns are tempting because they work even if one does not think about regulating range. However, they stopped working well when sides began to chase shots. I recommend against using a long-range gun (>20 range) except for an offensive supplement.

Keep it medium

Gatheres seem to do best relatively small (500-1000). Fighters tend to have more expensive processors, sensors, and radios, so fighters tend to be bigger (1000-1500). If your types are smaller than about 300 or larger than about 2000, expect unfavorable rule changes in response.

Analysis of Champions

I discuss recent champions that stayed champion after the rules were fixed. See Devon's analysis page for a complete list.


Intoxicated was the best fighter at the time. Its approach to gaining territory was simple: steal it. It partly made up for its high combat fraction by continuously moving to new territory, ensuring an ample supply of manna. Its strategy floundered once its combat superiority ended.


Active's strategy is pretty antique. It did well because its active dodging made it difficult to defeat at combat, not due to any strategic prowess.


Commune has excellent defense, enabling it to kill anything that gets close or shoots at it. Its colonialism deprives it of territory, and hence growth. This lets it be defeated by enemies with similar fighters but a better economy.

Missile Bacteria

Missile Bacteria combines unmatched gatherers with defensive missiles.

Its performance has degraded against Harmless and RK because both are resistant to missiles. I expect a version of Bacteria with better defense (e.g. Revenges) to do quite well.

Not Quite Harmless

Not Quite Wise's doubletime of about 1500 is comparable to RK's gatherer's and only a little worse than Escherichia's. When enemies show up, Wise combines two coping techniques that aren't intuitively combinable: scare others away and run away. Wise's bizzare behavior when it sees an enemy (shoot at it, then run away) has some nice effects:

Wise's weakness was sides that take killing seriously. Harmless takes advantage of the fact that most sides behave terribly in the presence of shielded robots. (robot-shield-fraction not being implemented yet encouraged this.) Against some opponents, such as Gunner 2 or Untouchable 1, a Harmless encouraged the side to commit suicide. Even against opponents that handled Harmlesses relatively well (e.g. Commune), the Harmlesses provided enough of a distraction to keep the better-armed enemies from stealing Wise's territory.

Revenge Killer

Revenge killer was the first side to combine an opportunistic gatherer's no-enemies performance with a coordinated, dedicated defense. (Trying to explain RK's success led to the creation of this article.)

Grobots by Devon Schudy ( and Warren Schudy (