Sigma · Ekumene v3
Ekumene · 寰宇 —— 逐期动画Ekumene — Period-by-period viewer
这是一次跑完的历史模拟。世界从公元前四千年的一千万人起步，粮食、人口、耕地、技术、政体、战争、瘟疫按各自的规则逐期推演，一直跑到公元 1500 年。下面的动画每十期取一帧，可以切换八个图层，拖动时间轴看它们各自怎么变。 One completed run of a historical simulation. The world starts in 4000 BC with about ten million people; food, population, cropland, technology, polities, war, and epidemics each advance by their own rules, period after period, through to AD 1500. The animation below samples every tenth period; switch between eight layers and drag the timeline to watch each one change.
这一版做了什么What changed in v3
前两版的世界人口跑到几十亿，是真实历史的几十倍。这一版把偏高的成因逐条查清并改掉了，改的都是底层的物理与生物规则，没有一处是照着历史数字去调参数——项目的纪律是模型参数不许用历史结果标定，否则模型"对上历史"就不再说明任何事情。 In the two earlier versions world population ran into the billions — dozens of times the real figure. This version traced the causes one by one and fixed them in the underlying physical and biological rules. None of the fixes was made by tuning a parameter to match a historical number: the project's discipline forbids calibrating parameters on historical outcomes, because a model tuned that way tells you nothing when it then "matches history."
最要紧的一条是单产。此前石器时代的农民和铁器时代的农民拿到同一个每公顷产量。现在按技术档分开：石器档的产量取实测的烧垦农业，按整个轮歇周期摊算（一块地种一年休二十年，不能按种植当年算），中位是每公顷每年 0.15 吨；定耕铁农具档取中世纪英格兰庄园账、罗马埃及、美索不达米亚神庙田庄的实测，中位 0.65 吨。比值 0.33。 The most important change is yield. Previously a Stone Age farmer and an Iron Age farmer were given the same output per hectare. Now they differ by technology tier: the Stone Age figure comes from measured swidden agriculture, amortized over the whole fallow cycle (a plot cropped one year and rested twenty cannot be counted at its cropping-year rate) — a median of 0.15 tonnes per hectare per year; the settled, metal-tooled tier comes from medieval English manorial accounts, Roman Egypt, and Ur III temple estates — a median of 0.65. The ratio is 0.33.
只砍单产不行：世界当场全饿死了。原因是模拟里一直没接采集渔猎的口粮通道，等于假定石器时代的人只吃农田产出，这不符合考古事实。接上之后人口稳在温饱线附近，形成正常的马尔萨斯均衡。此外还解开了一个循环：技术档原本要"有政体且中心格已炼铁"才升级，而政体涌现要余粮、余粮要单产、单产要技术档——四者成环，世界永远停在石器时代。现在技术档直接按每格自己的冶金水平判，因为农业技术的扩散本来就不需要国家。 Cutting yields alone was not enough — the world starved outright. The reason: the simulation had never connected its foraging channel, in effect assuming Stone Age people ate only what their fields produced, which is not what the archaeology shows. With foraging connected, population settles near subsistence in a normal Malthusian equilibrium. A circular dependency was also broken: the technology tier used to require a polity whose capital had reached iron, while a polity required surplus, surplus required yield, and yield required the tier — a closed loop that froze the world in the Stone Age. The tier is now read from each cell's own metallurgy, since the spread of farming technique never depended on states.
这一版的人口曲线The population curve
| 年份Year | 模型Model | HYDE | 比值Ratio |
|---|---|---|---|
| 3000 BC | 5.6M | 43M | 0.13× |
| 2000 BC | 5.8M | 71M | 0.08× |
| 1000 BC | 9.1M | 107M | 0.09× |
| AD 1 | 80.4M | 225M | 0.36× |
| AD 1000 | 270M | 314M | 0.86× |
| AD 1500 | 427M | 488M | 0.88× |
HYDE 是荷兰环境评估署的全球历史人口与土地利用数据库，这里只作事后对照，没有参与任何取值。公元前的三千年模型明显偏低，成因还没查完，如实留着。 HYDE is the Netherlands Environmental Assessment Agency's history database of the global environment; it is used here only for after-the-fact comparison and played no part in setting any parameter. The model runs clearly low through the three millennia BC; the cause has not yet been traced, and the gap is left on the record.
版图为什么长成这样Why the map looks like this
看政体图层会注意到两件事：国家覆盖的格子只占陆地的一成，其余大片是无人统治的灰色；国家的个头也不大，最大的一个占 216 格（约 38 万平方公里）。这两件事的成因查过，一件是缺陷、已经修了，一件是模型在如实反映自己的规则。 Two things stand out in the polity layer: states cover only about a tenth of the land, the rest staying grey and stateless; and states are not large, the biggest holding 216 cells (some 380,000 km²). Both were traced. One was a defect and has been fixed; the other is the model faithfully following its own rules.
缺陷是军队只在建国那一期招募一次，之后不再扩编，而驻军需求随领土增长——一个政体长大之后，全部兵力都被驻防吃掉，野战军只剩几百人，再也打不动。现在改为常备军：兵力目标等于治下人口乘以聚居度除以军民战力比，缺额每期补募，用的是建国门本来那条口径，没有引入新参数。改完之后最大政体从 70 格涨到 216 格，一半是单格小国的局面降到不足三成。 The defect: an army was recruited once, at founding, and never expanded, while garrison demand grew with territory — so a state that expanded had its whole force absorbed into garrison duty, leaving a few hundred men in the field and no capacity to campaign. It now maintains a standing army: the target strength is the subject population times the clustering share divided by the soldier-to-civilian strength ratio — the same formula the founding threshold already used, with no new parameter. The largest polity went from 70 cells to 216, and the share of single-cell statelets fell from about half to under three in ten.
另一件不是缺陷。那些无人统治的格子，每格的可耕地上限中位只有 2 平方公里，而有国家的格子是 241 平方公里——它们是沙漠、高山、极地和雨林，本来就没多少地可种。模型里开战要求征收所得能覆盖驻防开销，于是国家不去占它们。前现代的实际情形与此相符：国家控制的是河谷、平原与灌溉农业区，山地、干旱带和森林里的人长期不在任何国家治下。 The second is not a defect. Those stateless cells hold a median of just 2 km² of cultivable land each, against 241 km² in cells under a state — they are desert, high mountain, polar, and rainforest, with little to farm in the first place. The model requires that what can be extracted from a cell cover the cost of garrisoning it, so states do not take them. This matches the pre-modern record: states held river valleys, plains, and irrigated country, while people in the mountains, the drylands, and the forests lived outside any state for long stretches.
八个图层各是什么The eight layers
政体归属：每个国家一种颜色，灰色是无人统治的格。人口：每格人数取对数着色。耕地：每格熟地面积。口粮比值：每人每天分到的口粮相对于温饱线，1 是刚够吃，红色是饥荒。城市：城内人口。文字与铸币：各格所达的档位。官僚：在编官吏按治下格摊算。 Polities: one hue per state, grey for cells under no rule. Population: people per cell, shaded logarithmically. Cropland: cultivated area per cell. Food ratio: daily ration per person against the subsistence line, where 1 is just enough and red is famine. Cities: urban population. Writing and coinage: the tier each cell has reached. Bureaucracy: officeholders spread across the cells they govern.
怎么读它How to read it
这是模型，不是历史。它能回答的是"给定这套规则，会长出什么样的世界"，不能回答"历史为什么是这样"。看的时候值得注意的是形态而不是具体数字：人口先长期低平再加速上升；政体从无到有、从零星到密布；技术从几个格子扩散到大半个世界。哪里与真实历史对不上，那个偏差本身就是有用的——它指向模型缺了哪条机制。 This is a model, not history. It can answer what kind of world grows out of a given set of rules; it cannot answer why history went the way it did. What is worth watching is shape rather than exact numbers: population stays low for a long time and then accelerates; polities appear from nothing, first scattered and later dense; technology spreads from a handful of cells across much of the world. Where it fails to match the real record, that gap is itself useful — it points to a mechanism the model is missing.