Grobots - Sides - Analysis

Growth Rates and Efficiency

There are various conditions under which growth rates (and other things) can be measured. I ran these sides alone, starting from a seed in the center of the world, for 4500 frames (two and a half minutes, or a quarter of a standard round). This measures their early growth rate in the absence of enemies. The July 15 2003 tournament has doubletimes for more modern sides.

• Many of these measurements are old. Rules have changed, so be careful about comparisons.
• For theotrophs, and for sides that do anything randomly, results are quite variable. So pay no attention to the less significant digits.
• The lack of enemies makes this test unrealistic, since there's no need for defense and no opportunity for predation or parasitism.
• Since the side can gather manna from the entire world, theotrophs do much better than they would in a real round.
• Efficiency matters. Missile-lunatic 3 wastes a lot of energy running in circles, and has a terrible doubletime; Missile-lunatic 4 doesn't and doesn't.
• Many theotrophs do much better started in the center than in a corner. I've been starting them in the center, but maybe this should be changed.

Eventually's doubletime

Different versions of Eventually have very different doubletimes. The following were all taken with the same version of Grobots, so they do not depend on rule changes.

Versions 5 and 6 took first or second place in tournaments, and survived about half the time. Version 7, despite its heavier armor, survived only a quarter of the time, and fell to third place. It even had a high early death rate. Version 8 is back in first place. The moral: growth matters. Don't be too quick to cut those expensive solar cells to save energy.

Past Champions

Below are, in order, the sides that have taken first place in a tournament. Many of these sides were around long before they won; the version described is the one that became champion.

Hardcoded Predator

In the early days, when sides were written in C++ and built into the engine, the best of them was Predator. It was a simple gatherer with blasters and heavy armor. It won consistently, despite a serious friendly-fire problem, because there was nothing else that could both eat and fight.

After the hardcoded sides stopped working, there was a recreated Predator with (inaccurate) grenades. I think it was the best side for a while, for the same reason.

Homesick

It's dangerous to go too far from home - you might get shot at. Gatherers have a problem with this. So Warren wrote a side whose gatherers prefer food that's close to their birthplaces. They also make stationary grenade launchers to fortify territory. This combination of caution and expansion made Homesick the first champion worth the title.

Old fighting Rat

Once upon a time, it was possible to fire a weapon even when you had shields up. Rat was the first side to take advantage of this. (Actually, there was a hardcoded side called Leech that used blasters and syphons while shielded, but I don't think it was ever finished enough to be effective.) It ignored enemies and blindly gathered food, raising a shield when shot at and only shooting back when its food-gathering brought it close to a target. With the advantage of shields, this was good enough to give it scores of over 70%.

Rat made us think about the roles we wanted for shields, and after some confusion, we changed them so weapons no longer worked. So Warren split Rat into two sides: the modern Rat, which uses shields but has no weapon, and Rat-without-shields, which abandoned the shield but kept weapon and championship.

Rat without shields

Shields weren't the only reason for Rat's success, and this version held on to the championship (narrowly). It had a high repair rate, which in combination with heavy armor (instead of shields) kept it alive despite its nearsightedness. It had enough armament to win most battles, and its mobility (in pursuit of food) gave it a good chance of eliminating most enemies, or at least surviving.

Megadont

Megadont exploited several imbalances and a common weakness of its enemies. It consisted of one huge cell with range-40 grenades. It had a constructor, but that was a waste of energy, as in practice it never reproduced. Its size gave it an advantage in combat. Large grenades were too powerful, and long range too cheap, so it exploited these. More important, it exploited its enemies' indifference to attack: most sides at the time made no attempt to defend themselves unless they could see an enemy robot. Megadont won by killing everything else from long range.

Fixing the grenade size and range imbalances weakened Megadont. The arrival of a side which did not ignore bombardment defeated it.

Life and Death

Life and Death was designed specifically to beat Megadont. It was the first side to chase shots, and therefore the first that could resist Megadont's long-range bombardment. Like Gnats, it uses a syphon and short-range blaster to stun and kill its enemies. It also has an unarmed type which is supposed to survive when it loses battles. It beat Megadont (and everything else) surprisingly easily.

Later versions zigzag while charging, use shields, and eat while fighting.

Teledont

Teledont was champion for a long time. It's like Megadont, but with a blaster with a damage rate higher than that of Megadont's grenade launcher. It defeats Life and Death by storing enough energy to delay stunning by syphons. It reproduces, and avoids friendly fire by sending its cells to different parts of the world.

Teledont won as long as nothing could kill it, but it grows so slowly that it dies if it ever meets serious resistance. In the presence of opponents that chase shots, it picks more fights than it can handle and dies quickly.

Version 5 makes the first use of multiple sensor returns, to select food and targets better. Version 6f briefly recovered the championship in the days of Intoxicated. We're not sure why.

Eventually...

Eventually was the first side to change its strategy over time: it devotes energy to growth (and defense, in later versions) at first, and later switches to offense. This gives it faster growth than most opponents, so its offense can overwhelm them. The offense is only a lot of simple missiles, but that's more offense than most sides have, and more than any side at the time could resist. Even sides that could shoot down almost all the missiles eventually miss a few. Eventually won outright more than half the time, until Productive 3 came along and did the same thing better. Versions 4 and later ignore Productive's decoys. Version 7 was a disaster, with a doubletime over 4000, and fell to third place; version 8 fixed this. Later versions coordinate their missiles in an effort to find targets faster and overwhelm their defenses.

Productive (Bombs and Factories)

The first version of Productive was helpless - no wonder; it was the author's first side! (Seeing it not working made Devon write Eventually.) The second version was the first side to move deliberately: it goes to the nearest corner, where wandering enemies are less likely to find it. It tends to get killed by blundering into an enemy, but this is more than made up for by the safety of the corner. Some versions also communicate to arrange some of the cells in a line (to reduce traffic jams and vulnerability to missile accidents) and to disperse the missiles. The third version was the first side to have an effective defense against missiles: it uses tiny, cheap missiles as decoys and antimissiles. The combination of hiding in a corner, offense, and missile defense works pretty well, even after missiles were weakened.

World Toad

Written to exploit the size advantage, World Toad is a giant gatherer with a lot of attention to combat. It doesn't fight unnecessarily, and has good defense, so it used to survive more than half the time. Because it chases shots, it resists long-range attacks (it was designed with Teledont in mind) and tends to kill sides that use them. This has a substantial effect on game balance.

Most of the earlier champions often won by elimination, but World Toad rarely did. Its strength was in survival.

Intoxicated

Few sides have good defense, so Intoxicated can often straightforwardly kill them all. It has passive dodging, which lets it take on large groups of enemies. It's also the first side to travel in a group. This leads to a good deal of accidental group combat.

Even after several rule changes to weaken large cells, its size helps. A large cell concentrates firepower without any coordination effort. It also reduces friendly fire. Between its size and its agressiveness, Intoxicated's success is very annoying from a game-balance standpoint. This sort of side isn't supposed to do well.

Bacteria

This was supposed to be a clone of Microb, plus some armed types. But when I entered an incomplete version in a tournament, it won by a large margin. (This wasn't repeatable, as that tournament had an unusually large world, but it often scores well.) So I left it as is.

It has only two types: an autotroph, and a gatherer that runs away from shots. Running is a good idea for any unarmed type, but it works especially well with a trial-and-error strategy. Instead of dying, misplaced cells just run to better pastures. This works much better than shields.

Active Large

Active (of any flavor) is spooky to watch. It appears to dodge shots before they're even fired, and to somehow cause enemies to shoot at places it isn't. It wins by elimination a lot, but it's hard to resent something so clever.

Active 2 doesn't work very well, because it spends half its construction on missiles, which aren't very useful in most situations. Since Active Large costs twice as much as its missiles, it accidentally spends less on missiles. This doesn't completely explain its success, though; Active 3 builds even fewer missiles, but doesn't do quite as well. Apparently size is still an advantage.

Commune

This is a communication-intensive side. It completely separates defense from economy in order to concentrate as much power as possible in its fighters. The fighters respond to threats as a group, so it is very hard to kill Commune by normal means. Its gatherers respond to other cells' requests for energy with friendly syphons. They tend to eat all the food in the area, so enemy gatherers are not attracted to the colony.

Because the fighters have short range, Commune is vulnerable to missiles. The separation of gestation from gathering (but not from solar power) also means it grows rather slowly.

Missile Bacteria

Business Cycle uses missiles for defense, with great success. So I added some simple defensive missiles (the type took five minutes to write) to Bacteria. They chase shots, so shooting at Missile Bacteria is usually fatal.

Not Quite Harmless

This uses Not Quite Wise as its economy, supporting a shielded leech type. The idea of a leech is years old, but this is the first one that works. No existing side can handle it; they either kill themselves by friendly fire or run out of energy. It even resists missiles, thanks to its shields. It should be helpless against syphon users such as Gnats or Life and Death, though.

We let syphons work through shields to encourage leeches, even though we knew it wasn't a good idea. Stopping this weakened leeches, but they remain useful as decoys.

Revenge Killer

Warren thought short-ranged blasters were overpowered, but no side was specialized to defeat them, so he wrote one. Revenge Killer keeps Revenges at arm's length by running away and pushing them with a force field, and kills them with a large blaster. It's rather agressive, so it can capture lots of territory for its skittish gatherers.

This also works well against missiles, and indeed against pretty much everything. And Warren no longer thinks short-ranged blasters are overpowered.

Untouchable

Like Business Cycle, Untouchable collects food with force fields, without risking its gatherers. It earns its name by also using the force fields on enemies - it pushes them away, which is easier than killing them, and makes Untouchable almost impossible to attack. The first version had long-range grenades, but it was hard to coordinate these with the force fields, so they were removed in favor of medium-range passive-dodging fighters. Feeding all this takes a lot of syphons, which make the side rather pretty.

When force fields were weakened, Untouchable became easier to touch, but because of its good fighters it remains one of the best sides.

Commune Plus

Commune has not changed much since the first version, because it matches Devon's vision of what it should be. But such a vision is not a reliable guide to quality, and Warren was not interested in following it. He added an autotrophic type: Solar Outlet (from Untouchable). This feeds Revenges more reliably than Sustenance, and improves the growth rate when the colony is large. Together with a few other upgrades, which will probably make their way back into Commune, this fixes many of Commune's weaknesses.

Wise Men

Business Cycle's Day Trader combines gathering and fighting pretty well. Wise Men splits it into gatherers, fighters, and decoys. Version 2 won by building lots of those fighters and using them as a group. (The decoys seem to be a waste of energy; the side does better if it builds more fighters instead.)

Segregated Eaters

Since the beginning, gatherers have wasted time when more than one cell tries to eat the same food. Sides which communicate about who's eating what can grow much faster. Segregated Eaters assigns each tile to a cell, intead of tracking individual foods like Frog Celestial. Also unlike Frog Celestial, it has a separate fighter type. These fighters are agressive and numerous enough that Segregated Eaters often wins by elimination.

Side Lengths

omg, homesick was simple! - Warren, April 2004

Champions