Repository navigation
Dune City v1.0.869
Frozen source a6e26c5; producing run 43597776305. All desktop platforms complete; mobile channels are independent.
DuneCity 1.0.869
Campaign AI opponent
- The Campaign AI opponent you can choose for a game now follows the original Dune II campaign rules in new games. Its factories pick what to build the original way. Each unit it builds goes on the attack as it is produced, rather than gathering into timed waves. The teams written into each mission run their original team scripts, read from your own DUNE.PAK. Destroyed buildings are rebuilt from up to five remembered slots, and the Death Hand aims at the first enemy building instead of the strongest base. Damage to its base no longer sends its whole army out.
- This is the original behaviour adapted to DuneCity's engine, not a frame-by-frame copy of the original game. QuantBot is a separate choice and is not changed by this. Games saved before 1.0.869 keep the previous Campaign AI behaviour when loaded.
- It was tested for correct operation, not for a particular win rate: 1,665 automated games finished without an engine error, and saved games were loaded and continued.
QuantBot AI
- The campaign rule that denies a Repair Yard by difficulty now applies only to Easy opponents. In 1.0.867 a Medium helper that started a mission without a Repair Yard never added one, so its damaged tanks were not repaired. Helpers of every difficulty, and Medium and harder opponents, now build one. The usual conditions still apply: tech level, credits, a place to build, and in city games the technology the map allows.
- Campaign QuantBots no longer upgrade their Construction Yard straight away. They wait until they have income (spice, or taxes in a city game) and the power they are required to have, and can pay for the upgrade while keeping a reserve of at least 2,000 credits.
- Measured in Single mission games on the original rules with an idle human, missions 14, 20 and 22 for three houses: a Medium helper against Hard opponents won 16 of 36 games, up from 10. A Hard helper against Medium opponents was unchanged within the margin of the test (32 of 36 against 31).
- Medium QuantBots in Custom games also no longer inherit the campaign Repair Yard ban. On the original rules a Medium QuantBot used to be limited to the Repair Yards it started with, or a single one if it started with none. In 16 matched Custom games against a Hard QuantBot it built 1 to 3 Repair Yards in every game with the change and none without; it won 3 of the 16 either way, so no change in strength is established.
- The campaign setup screen described the QuantBot opponent levels wrongly, and now describes what the game does. It used to say Easy and Medium opponents attack one house at a time and Hard two at a time. In fact every enemy house runs its own attack timer on every level, so houses can attack at the same time. What differs between levels is the size of the attack waves and the strength of the opponent's economy. Correcting these descriptions changes no combat rule.
Known issues
- Late missions against Hard opponents are still mostly lost with a QuantBot helper and an idle human. In 576 Full campaign test games on this version's game code, Medium, Hard and Brutal helpers won 20 to 24 of 24 games on each of missions 5, 8 and 14, and an Easy helper won 23, 18 and 18. On missions 20 and 21 together an Easy helper won 3 of 48, a Medium helper 11, a Hard helper 16 and a Brutal helper 21. On mission 22 they won 0, 3, 14 and 13 of 24, so a Brutal helper did no better than a Hard one. Harkonnen mission 21 was lost in all 32 games. This release does not change that.
- The wave-size changes we tested for Hard opponents did not fix it, so that value stays at 3,500. After 2,880 exploratory games, the three most promising settings were each played for 576 games on new seeds against the current value. None gave a consistent improvement on the late missions. On missions 20 and 21 a Brutal helper won 21 of 48 at the current value and 15, 19 and 20 with the alternatives; a Hard helper won 16 at the current value and 12, 20 and 20. On mission 22 a Brutal helper fell from 13 of 24 to 9, and a Hard helper from 14 to 9, with all three. The two settings that allowed larger waves on the early single-base missions also cut an Easy helper's wins on mission 14 from 18 of 24 to 3. Harkonnen mission 21 was won once in those 96 games. Removing the rule that makes the helper wait until its army is strong enough compared with the enemy's did not win that mission either: 36 of 36 games lost. The cause of the late-mission losses is not yet known. The settings used for these tests remain in the game, switched off.
Compatibility
- Multiplayer uses protocol 125: all players need 1.0.869, not 1.0.867 (protocol 123).
- The save format is now 9886. Two campaign saves made with 1.0.867 on the original rules were loaded and continued in this version.
- Android has a separate version and publication process.
- There is no 1.0.868 release. It was this same game; its Mac package could not be completed, so 1.0.869 replaces it.