My impression of the play of some top Planet Wars bots from the 2010 contest, plus oddshrimp itself at the end.
#1 bocsimacko Bocsimacko’s play impresses as foresightful—as it should, it is far stronger than #2. Its ships have a way of seeming to land by luck where they cause the opponent trouble, a clear sign of skill. (Chess players say “the good player is always lucky.”) Unlike other bots, it understands suppressing attacks: It knows how to reduce your options with small attacks here while it makes a killing attack there. It also tends to dispatch small fleets to their doom for no apparent reason and doesn’t quite have a clear understanding of what captures are feasible. I think of it as strategically brilliant and tactically sloppy. As a bot facing much weaker opposition, bocsimacko’s big concern is not to play the very best moves but to avoid risk, a job in itself in a game with simultaneous moves, and to take advantage of mistakes. In other words, bocsimacko is vulnerable and stronger bots are possible. I intend to beat it.
#2 iouri Called “iori” in my result tables, for “historical reasons”. Iouri I think of as scrappy. There are games where it gets nervous and sits on a stalemate, winning or losing by a small margin, but its favorite game is a messy slugfest. It makes a lot of mistakes but it also has a lot of smarts—it didn’t earn #2 for nothing. One of only two bots that I have seen show the tactical skill to beat oddshrimp3.3 from a level start in complications where a lot of planets get traded back and forth. (#31 Manwe56 can also do it rarely. Bocsimacko can’t.) Against some opponents iouri’s play seems more conservative, so maybe it really has an all-around style.
#3 Slin Slin does not want to wipe you out. Slin wants to get solidly ahead and lock you in a wrestling hold that you can’t break, and then it is content to wait for the game to run out. It avoids the risk of going wrong in a finishing attack, a serious problem for other bots against oddshrimp4.1. Big thanks to its author Fredrik Persson who shared a copy with me. I’m delighted partly because I now have three test opponents stronger than oddshrimp4.1 and partly because its playing style is unique.
#8 GreenTea After reading the author’s description of the rigid phases that the bot steps through to choose its moves, I can’t help but think of GreenTea as bureaucratic. It’s effective but it’s cautious and sticks to its own rules. Like any bureaucracy, it often holds fleets back from the front lines, causing it serious trouble against oddshrimp4.1 and later.
#12 dmj111 Highly aggressive, a style that poses problems for oddshrimp.
#17 wagstaff Wagstaff only wants calm. It is acutely aware of whether it is ahead or behind. When ahead it plays conservatively, occasionally to the point that it doesn’t consolidate its advantage when it should. When behind it grows desperate, trying anything to return to the calm of being ahead. “I must punch you down so I can relax!” Its desperate struggles do not help it against oddshrimp2.3 or later. Curiously, my tries to improve oddshrimp by making it aware of when it is ahead or behind have never come to much.
#18 medrimonia A sly opponent. It loves to hang back and snipe neutrals you’ve just taken, and plays tricky moves that conceal its intentions from weaker bots. One of its favorite ploys is to send extra forces to take a neutral and redirect them from there to new targets. Oddshrimp up to version 3.4 does not foresee that ships in flight can relaunch after landing and often gets caught out.
#19 smloh “That planet’s mine. That one too. Mine, mine, mine.” Smloh’s strategy is to build a strong base and then overwhelm you with numbers. It works poorly against oddshrimp, which usually attacks before smloh’s base is built.
oddshrimp2.x bots are crazy reckless. They are familiar with the idea of defending against incoming ships, and that is the total extent of their self-preservation. “The enemy might attack next turn? I don’t understand you.” It works surprisingly well. Apparently attack often is the best defense.
oddshrimp3.x bots are tough and aggressive. Even oddshrimp3.4, the most balanced of the series, still tends to prefer aggression over greed. Leaving aside the odd blunder, they win efficiently when ahead and fight stubbornly when behind. They prefer attacking the enemy over taking neutrals, and are likely to get into trouble when the camps are far apart and taking neutrals is the only way (including oddshrimp3.3 and 3.4, which have major adjustments to reduce this weakness). They attack when I think they shouldn’t, and succeed more often than I expect, commonly by switching targets when the first attack draws enemy ships out of position (“tacking”). They don’t have much more sense of strategy than that. Opponents with greedy styles lose to oddshrimp3.x bots, opponents with matching aggressive styles do well.
oddshrimp4.x bots own the middle game but are weaker in the opening. Oddshrimp4.1 sometimes follows this winning strategy: 1. Play a poor opening and get behind. 2. Wait for the enemy to finish you off. 3. Turn around the resulting battle from behind and brush the dust from your hands as if it were nothing. “I planned it that way all along.” Oddshrimp2.x bots already won that way occasionally, but oddshrimp4.1 does it a good 10% of the time against a range of opponents—even bocsimacko, rarely. See #3 Slin above!
Oddshrimp4.2 and 4.3 redefine “somewhat behind” as “winning”. They no longer wait for the enemy to try to finish them off, but go make trouble themselves. Sometimes they fight back and forth until they can gain an advantage, sometimes they blithely pick up an enemy planet, even one that looks invulnerable, and the enemy is surprised to find itself helpless. “Oh, was this yours? Well, I guess it’s mine now.” These bots would be untouchable if they didn't get behind in the first place.
last updated 4 June 2014 with new oddshrimp4.x notes
last big update 21 March 2014 with Slin and oddshrimp4.1
original version January 2014