Changelog v3 120s - Heliomance/Uncharted-Horizons GitHub Wiki

_{Changelog · Design document — Changelog}

Breakthrough IDs link to the breakthrough that holds the ID today. A few early entries predate a renumbering or a merge, so an ID there may since have passed to another breakthrough.

• Ring Fortress merged into Reinforced Concrete (CO11, Modern, STANDARD; Tamsin, 15 October, the row survey's merge 7). The Military Base and the Bunker move to the row that brings the concrete, with Liège and Verdun in its notes. v3.88 had created Ring Fortress so that the Military Base's Arsenal was certain to come first; Reinforced Concrete does not require Star Fort, the Arsenal's row, so the Military Base now requires a Castle instead (§7's requirement overrides; the Castle's row is CORE), and the defensive line skips only its STANDARD rung. Star Fort keeps the Arsenal and the Kremlin.
• 537 breakthroughs: CORE 139, STANDARD 244, UNCOMMON 61, RARE 39 and UNIQUE 54.
• Decided with it (Tamsin): the four CORE rows that are CORE only for being on an FTL route (Quantum Entanglement Mastery, the Laser, Composite Materials, Carbon Fiber) stay CORE; an audit of the Competing Solutions is parked (design/notes/deferred_tasks.md).
• Measured (three worlds, all 15 victory conditions, three benchmark players, 2,700 games with rivals made under v3.125, none unfinished): victory medians 304 (expert; middle half 298–313), 307 (novice) and 381 (average), against base Civ V's 305, 305 and 380. Civilisations enter every era by count within 9 turns of their development reaching it. From the Modern era on, a good economy makes 1.16 discoveries a turn counting rivals' knowledge, an average one 0.93.

• Three merges (from the row survey, design/notes/row_survey_20261015.md; Tamsin, 15 October):
  • General Relativity into Special Relativity, as Relativity (RC01, Modern, CORE). General Relativity granted nothing of its own. The merged row is the prerequisite for every FTL theory, its seven dependents follow it, and it proposes Warp, Wormhole and Tachyonic together (§4). The proposals had been staggered only so that each theory arrived with the breakthrough that proposed it in real life (Tamsin), and the merged row holds both theories.
  • Quantum Electrodynamics into Quantum Field Theory (QP05, Atomic, CORE), which takes QED's +10% Insight to Electronics and its prerequisites; the Ion Drive and Metamaterials follow it. In step, Aetheric Electrodynamics into Unified Field Theory (Elemental) (QM-E04), which now always carries its grants.
  • Uranium Discovery into Radioactivity (NP01, atomic worlds, Modern, CORE): Radioactivity reveals Uranium and needs Deep Shaft Mining, as its elemental mirror, the Boundary Survey, already did both.
• Two cuts. Sealed Environments: its prerequisites (Reinforced Concrete, Oxygen Theory, Glassmaking) move onto the Submarine and Life Support Systems, and its +25% Production towards Submarines onto the Submarine. Atmospheric Science: its +1 Gold from Harbors moves to Wireless Transmission, as storm warnings sent to ships by wireless; Extremophile Engineering and Planetary Engineering Theory no longer need it.
• Five demotions to UNCOMMON: Desalination, Hydroponics, Flying Boat, Nuclear Thermal Rocket and Helium Extraction. Helium was long one nation's near-monopoly, so most civilisations' airships now end in the Airship Disaster, as they did in history (Tamsin: fine as the most common outcome; §1.2). About half of civilisations hold Helium Extraction by the race's finish, two in five of them by diffusion.
• Fixes found in passing. The production pass of v3.10x removed building bonuses from atomic rows but missed their elemental mirrors: Aetheric Electrodynamics' +2 Production from Factories goes with its merge, and Resonant-State Computing loses the +1 Science from Research Labs and Public Schools that the Transistor lost. The same pass left a stray semicolon in the Laser's effects. Aetheric Vacuum gets Casimir Effect Amplification's report, on the medium's energy rather than the FTL method, so the Wormhole rows of both kinds of world report alike. The Submarine's note names its base-game type's unique correctly (Invisible to non-adjacent units; parked since 13 October).
• 538 breakthroughs (from 544): CORE 139, STANDARD 245, UNCOMMON 61, RARE 39 and UNIQUE 54 (§8, whose tier philosophy's counts now match its table).
• Competing Solutions are measured. The simulator now charges each alternative its ×1.5 detour margin while its primary exists in the world, and a civilisation that discovers the primary first never develops the alternative (§1.2; before, every measurement counted alternatives found after their primaries). A civilisation locks out about 7–8 alternatives a game, 4.5–6 of them from the Modern era on. Several races are lopsided: the Industrial Canal Network, the Atmospheric Engine, Advanced Fuel Cycles and Direct-Current Stations are developed in at most 11% of games, the Thorium Reactor in none of the expert's, while Stirling Engine Mastery, Analog Computing Mastery and Phage Therapy usually come before their primaries (design/simulator/power/results/v3124_confirmation.txt; parked for a decision).
• The Space era's price ×1.05 (1,469 per discovery, from 1,399; §2). The merges and lock-outs shortened the science victory to 301 / 305 / 378 (expert / novice / average) at the old price, the expert 4 turns short of base Civ V's; ×1.05 re-centres it. The prototype now takes about 5 turns on a good economy and 6 on an average one (§4), the Drive still about 10 and 12.
• Measured (three worlds, all 15 victory conditions, three benchmark players, 2,700 games with rivals made under the same settings, none unfinished): victory medians 305 (expert; middle half 298–314), 307 (novice) and 379 (average), against base Civ V's 305, 305 and 380. Civilisations enter every era by count within 10 turns of their development reaching it. From the turn development enters the Modern era to the race's finish, counting rivals' knowledge, a good economy makes 1.17 discoveries a turn (v3.123: 1.26–1.27) and an average one 0.93 (1.01); a game holds 283–289 discoveries by the finish (294–302). Turns per discovery while development is in each era (expert / novice / average): Modern 0.9 / 0.9 / 1.1, Atomic 0.9 / 0.8 / 1.0, Space 0.8 / 0.9 / 1.2 (v3.123: 0.9 / 0.8 / 1.0, 0.8 / 0.7 / 0.9, 0.7 / 0.8 / 1.1).
• Still to do. One a turn from the Modern era on would need about 17–19 fewer late discoveries a game for the good economies; the row survey's remaining options are there. Ring Fortress (merge into Reinforced Concrete, or demote) awaits Tamsin. The Competing Solutions' races want tuning. §7's building cost scaling is still to be re-solved for v3.122's and v3.123's eras.

• Parity with base Civ V's science victory, the design's hardest constraint (Tamsin, 15 October): a science victory must take as long as base Civ V's for the same player on the same economy, within five turns either way, because domination, cultural and diplomatic victories are unchanged and balanced against base Civ V's game length. The reference is base Civ V's own tree run to a launched spaceship on the simulator's income (Unciv's Gods & Kings techs, researching only what the ship needs, the Apollo Program and the six parts built from a typical city's Production): turn 305 on a good economy and 380 on an average one. The era model's 320 is base Civ V's whole-tree research time, not its victory. §2 now states the constraint.
• The late prices back to their v3.108 values (§2): 951 per discovery in the Modern era, 1,369 in the Atomic and 1,399 in the Space era, against v3.122's ×1.25, ×1.25 and ×2. With the SS FTL Drive built from Production rather than pencilled in at 4 turns, v3.122 measures 327 / 332 / 412 (expert / novice / average), not the 323 / 328 / 407 its changelog gave: 22–32 turns longer than base Civ V's. v3.122's era timing stays. The prototype now takes about 4 turns on a good economy and 5 on an average one (§4), and the Drive about 10 and 12.
• A smoother opening (§2; Tamsin asked for the discovery rate smoothed, counting knowledge from rivals). The Ancient era's many small breakthroughs had come at up to 1.25 a turn (the novice's). Its price is now 1.5× its even-pace price (22 per discovery, from 14), with its span to match (858 Insight, from 572), and the Primitive span is 1.3× longer at its old price (194, from 149), so that development still enters the Ancient era within 6–9 turns of civilisations' entry by count, rather than 10–14. Every era up to the Early Modern now runs at one turn per discovery or slower for all three benchmark players. The cost: civilisations get through their Ancient breakthroughs later, entering the Ancient and Classical eras by count about 4–7 turns later than before (the Medieval 3–4, the Early Modern 1–2; screened on three victory conditions). Development's eras now last about 33, 36, 28, 36, 49, 46, 36 and 56 turns on a good economy, and idea windows follow (§1.2).
• Measured (three worlds, all 15 victory conditions, three benchmark players, 2,700 games with rivals at the same prices, none unfinished): victory medians 304 (expert; middle half 298–313), 307 (novice) and 380 (average), against base Civ V's 305, 305 and 380. Civilisations enter every era by count within 9 turns of their development reaching it. Turns per discovery while development is in each era, counting rivals' knowledge (expert / novice / average): Primitive 1.1 / 1.1 / 1.4, Ancient 1.1 / 1.0 / 1.2, Classical 1.2 / 1.3 / 1.6, Medieval 1.1 / 1.1 / 1.3, Early Modern 1.2 / 1.2 / 1.5, Modern 0.9 / 0.8 / 1.0, Atomic 0.8 / 0.7 / 0.9, Space 0.7 / 0.8 / 1.1.
• The late pace stays as it is, for now. A game holds 280–308 discoveries by the race's finish whatever the prices or diffusion (measured 15 October), so with the game's length fixed, prices can only move the fast stretches around. The Modern to Space breakthroughs take 85% of the Insight spent on discoveries, so they cannot get dearer without lengthening the science victory, and slowing knowledge diffusion only turns free breakthroughs into the player's own. Fewer late discoveries need fewer late breakthroughs: merges, cuts and demotions to UNCOMMON are under review.
• Still to do. §7's building cost scaling, solved for v3.108's era timing, is to be re-solved for v3.122's late eras and v3.123's early ones. The FTL methods' balance (v3.122). The simulator's income follows the Insight accumulated, not the breakthroughs held, so it cannot price the later Ancient-era buildings: to be confirmed in play.

• Back to base Civ V's game lengths (Tamsin, 15 October): a science victory should take as long as in base Civ V at every skill level, about 320 turns for a good player and 400 for a casual one, because cultural and diplomatic victories are balanced against those lengths. The targets used since 13 October (about 295 for a good economy, 360–375 for an average one) came from v3.91's description of the era model ("the fork's eras run about a tenth shorter"), which v3.108's redraw had superseded; v3.112's 20% cut in Space-era prices and its 5× prototype were made to reach them.
• The last three era boundaries at base Civ V's timing (§2). Development now leaves the Modern era at 66,231 Insight, enters the Space era at 113,629 and leaves it at 209,676: on a good economy around turns 228, 264 and 320, base Civ V's own. Before, development entered the Space era around turn 285, 24–29 turns after civilisations entered it by count, so 81–88% of the Space-era breakthroughs a player held at the finish came before its development reached the Space era, and the ahead-of-time premium took 36–47% of the Insight spent studying them. Now civilisations enter every era by count within about ten turns of their development reaching it. Idea windows follow the new era lengths (§1.2).
• Dearer late research (§2). Prices per discovery ×1.25 in the Modern and Atomic eras (951 → 1,189; 1,369 → 1,712) and ×2 in the Space era (1,399 → 2,798, replacing v3.112's ×0.8). A Space-era discovery had cost half a turn of a good economy's Insight, since §2 divided the era's span by all its expected content, most of it rows no single game reaches; it now costs about 1.25 turns. A civilisation's own discoveries come at about 0.8 a turn or fewer in every late era, and about one counting knowledge from rivals, where they had run at up to 1.4 (2 in the expert's Space era). The prototype, at 5× the Space era's price, takes about 8 turns on a good economy and 10 on an average one (§4).
• Measured (three worlds, three benchmark players, 2,700 games with rivals at the same prices, none unfinished): victory medians 323 (expert; middle half 313–333), 328 (novice) and 407 (average). Rivals had been replayed from records made at v3.120's prices, which understated any change in price; they are now made under each setting's own. The simulated expert was retuned for the new prices: it never targets a row whose ahead-of-time premium exceeds 2, and commissions an Endeavour only for the last thing its race waits on (900 games: median 326 → 323).
• Still to do. The methods have drifted apart: Warp and Tachyonic win 62–78% of races against their world's field, Wormhole, Tunnelling and Plenum 22–44%, against a band of 40–60%; part of it predates v3.122. §7's building cost scaling, solved for v3.108's era timing, is to be re-solved. The simulator's income follows the Insight accumulated rather than the breakthroughs held, so dearer research does not delay its buildings and real games may run somewhat longer: to be confirmed in play.
• Corrected: v3.121's figures for counting the study already done after the premium were 15% and 22.5%; with every Endeavour taking at least a turn, they are 14% and 22%.

• A Grand Endeavour may again target an idea already under study (Tamsin, 15 October 2026), to have it now rather than when its focus gets round to it: bomb shelters, the turn word comes that a rival has the bomb. v3.120 had allowed only Revealed ideas that had not occurred, because reordering research at 2× cost the benchmark players 4–9 turns of victory (measured 14 October, before the safety net). Re-measured on v3.120 (familiar world, 900 games a setting), allowing it costs the novice and the average player 5–6 turns and the expert nothing measurable. That stays true when nothing presses, so the simulated players keep the restriction (ENDUNOCC=1): the needs that make such an Endeavour worth its price come from the game situation, which the simulator does not model. §1.5's rule is now Under study or not; §1.4's REVEALED state, §4's Bad luck never blocks a science victory and Guarantee Rule 12 say the same. Open: whether the Insight already spent studying the idea counts towards the Endeavour.
• Measured (the expert, familiar world, 300 games; endeavour_oracle.py): at moments sampled every ten turns, an Endeavour would have had an idea under study sooner than funding its focus for 12% of such ideas: 2% of those alone in their focus, 8% with one other, 26% with two or three others, 52% with four or more. Counting the Insight already spent would raise the 12% to 14% (in the funded focus, 13.5% to 22%).
• §1.2's study rule corrected. A breakthrough's progress, which with 0.1 sets its share of its focus, is the Insight it has received measured in its price, not the share of its hidden meter filled. This is what the simulator has always done (the workshop report: "progress as a fraction of the average cost").
• No row changed.

• §1.2 now describes the discovery model in use. Every measurement since 1 October 2026 has run on it, while §1.2 still described the first design's Insight thresholds, hidden discovery rolls and excess Insight feeding the roll. It now sets out what the simulator does:
  • the active focus, and a spread that goes only to foci with something under study, weighted by the work waiting there and by §1.3's environment, with spill-over;
  • development and price: a breakthrough costs its era's price per discovery × 0.83, × 4 for each era it lies ahead of development;
  • ideas: a breakthrough is studied only once its idea occurs, at once for CORE, otherwise by chance through a window sized to the tier's odds (STANDARD 85% within half an era; UNCOMMON 40%, RARE 10%, UNIQUE 6.5% within a whole era), the window shrinking behind history;
  • study: hidden meters filled with momentum, every tier alike once its idea has occurred;
  • the tier table restated as each tier's odds and moment; accelerators as head starts (decided 14 October); knowledge diffusion as modelled, with the meaning of the old roll bonuses marked as proposed.
• The safety net (adopted). Rediscovery on demand: a lost STANDARD idea comes back when something waits on it, 10% a turn for each waiting thing. Grand Endeavours only for revealed ideas that have not occurred: at 2×, reordering work already under study cost the benchmark players 4–9 turns. §1.5 and Guarantee Rule 12 say so.
• W16 dropped. §4's rule that a Space-era row's window stays open while its focus is funded gives way to rediscovery: with rediscovery, W16 only bought speed (about 8 turns for the expert), and its written form did nothing measurable for the simulated players.
• Measured (three worlds, three benchmark players, 2,700 games; these are now the simulator's defaults): no game unfinished, against about 1% under the previous rules; victory medians 297–298 (expert), 301–302 (novice) and 374–375 (average), within the balance targets of about 295 for a good economy and 360–375 for an average one. The weaker players gain about ten turns because Endeavours are no longer spent reordering work.
• Knock-on wording. §1.3's necessity bonus is Insight alone. §1.4 and §1.5 lose their rolls and thresholds: proximity revelation is the glimpse (10% a turn for STANDARD ideas, 3% for UNCOMMON, provisional), serendipity is the spread, and an Endeavour costs 2× the breakthrough's price. Likewise §4's prototype paragraph, the Early Steam, Clockwork and Antikythera worlds, the elemental gateways' difficulty (prices, not thresholds), and §6's diffusion summary, Almost Discovered notice and dual paths. Mycological Knowledge's accelerators are head starts. §1.4's reveal rules themselves await the reveal decision.
• ~544 breakthroughs.