Sea Power : Naval Combat in the Missile Age

汉化不全

---- 一、游戏体验 / 1. Gameplay impressions 【中】 游戏好玩得很呐~~~ 比如说我最喜欢的反潜，那种和敌方核潜艇斗智斗勇的时候，时不时冒出来一些鲸鱼瞬间硬控所有人的感觉。 而且配上全球海域地图，给人一种真的在直布罗陀和北约对线的感觉。 而且强大的社区产出的模组结合优秀的游戏底子，直接 1+1 大于 2 的感觉。 【EN】 This game is an absolute blast. Take ASW, my favourite part: the cat-and-mouse duel against an enemy nuclear submarine — and then, out of nowhere, a pod of whales hard-CCs the entire bridge crew for a moment. Put that on a global sea-area map and it genuinely feels like you are squaring off against NATO at Gibraltar. And when a strong modding community meets a solid base game, it really is 1 + 1 > 2. ---- 二、关于我用 AI / 2. A note on how I used AI 【中】 以下是希望制作组添加的东西。 这是把我自己在游戏里面遇见的问题借助 AI 整理表述出来。若有并不喜欢 AI 的人，请不必看下面的内容。 我使用 AI 编辑单位，主要是自己不懂代码；我用 AI 编辑任务，是因为如果是自己出题自己解，有一种索然无味的感觉。 【EN】 What follows is my wish list for the developers. These are problems I ran into myself in the game, written up with the help of an AI. If you are someone who dislikes AI, feel free to skip the rest. I used AI to edit units mainly because I cannot code; I used AI to edit missions because setting your own puzzle and then solving it yourself feels rather pointless. ---- 三、简短版本 / 3. Short version 【中】 在让 AI 做完整战役（15 个任务 + 60 页简报）的过程中，我最希望制作组补三件事： ① 发布前自动校验（编辑器内置 + 命令行）：区域/单位是否压在陆地、必杀陆上目标是否真能被打死、战役节点是否孤立不可达、航点是否写了危险的 |Loop； ② 单位与弹药的伤害模型文档：现在 DamagePoints 与各模块 Integrity 的摧毁条件完全没写，玩家会花几个小时才明白“某个雷达为什么 80 枚导弹打不死”； ③ 给 AI / 外部工具用的机器可读接口：字段 schema + 地形合法性查询，让 AI 生成任务时不会写出“旱地行舟”。 有了这三样，社区任务的质量会立刻上一个台阶。 【EN】 Having used AI to build a complete campaign (15 missions + 60 briefing pages), there are three things I most want the developers to add: Automatic pre-release validation (built into the editor plus a command-line tool): whether zones/units sit on land, whether mandatory-destruction land targets can actually be killed, whether campaign nodes are orphaned and unreachable, and whether a waypoint uses the dangerous |Loop. Documentation for the unit and ammunition damage model: the destruction conditions linking DamagePoints and each module's Integrity are completely undocumented, and players can spend hours before understanding why 80 missiles will not kill one radar. Machine-readable interfaces for AI and external tools: a field schema plus terrain-legality queries, so that AI-generated missions do not end up with dry-land sailing. With those three, community mission quality would jump up immediately. ---- 四、详细版本 / 4. Detailed version -- T0 · 直接影响“能不能通关 / 会不会卡死” 【中】 发布前校验器（编辑器面板 + CLI，可挂 CI） 证据：官方任务格式文档里 0 处涉及伤害与地形合法性；我不得不自己写 4 个校验器。 危险配置的防御性拦截 证据：Waypoints=单航点|Loop 会让飞机到达后每帧重建行为树 → 主线程 100% CPU、整个 UI 硬锁死；应该在加载时就拒绝，而不是进了游戏才炸。 必杀目标可摧毁性检查 证据：us_cobra_dane_ewr 等 EWR 标了 Role=Target、500 分，但可扣总量 80 < 血量 180 → 玩家 80 枚战斧打不死；而官方 41 个必杀目标全部满足 ∑Integrity ≥ DamagePoints，说明本该有这条不变量。 战役节点可达性告警 证据：我生成的一个剧情页（free_event_7）没有前驱、不在链路上，玩家永远看不到。 【EN】1) Pre-release validator (editor + CLI, CI-friendly): the official mission-format document covers no damage or terrain legality, so I wrote four validators myself. 2) Defensive rejection of dangerous configs: a single-waypoint |Loop rebuilds the behaviour tree every frame, pinning the main thread at 100% CPU and hard-locking the UI — reject it at load time. 3) Killability check for mandatory targets: EWRs like us_cobra_dane_ewr are Role=Target and worth 500 points, yet their damageable Integrity (80) is below 180 DamagePoints, so 80 Tomahawks cannot kill one; all 41 official mandatory targets satisfy sum(Integrity) >= DamagePoints. 4) Reachability warning for nodes: a story page I generated (free_event_7) had no predecessor and no path led to it, so players never saw it. -- T1 · 编辑器可用性（让 AI 能编任务） 【中】 5. 机器可读 schema + 版本化字段清单：现在只能靠逆向 INI 文件；给了 schema，AI 与社区工具才能安全生成任务。 6. 单位支持绝对经纬度摆放：现状是单位用相对地图中心的 RelativePositionInNM、区域用绝对 GeoPoint，两套坐标系混用正是“旱地行舟”的根因。 7. 实时水陆/深度提示 + 一键吸附：拖部署框时直接显示“陆地面占比 %”，并提供一键“移到最近水域”。 8. 地形叠加显示（岸线 / 等深线）：现在必须外挂 NOAA 高程数据自己算，精度还受栅格限制（小岛礁会整个漏掉）。 9. 任务编辑器的撤销 / 重做 + 差异对比：手工改 INI 极易互相覆盖，我只能自己写备份脚本。 【EN】5) A machine-readable schema and versioned field list: today you can only reverse-engineer the INI. 6) Absolute lat/lon placement for units: units use RelativePositionInNM while zones use GeoPoint, and mixing the two is exactly what puts them on land. 7) Live land/water and depth readout while dragging a deployment box, plus move-to-nearest-water. 8) Terrain overlays (coastline, depth contours): today you must bolt on NOAA data yourself, and the raster still misses small islands. 9) Undo/redo and diff in the editor: hand-edited INI overwrites other people's work easily, so I wrote my own backup scripts. -- T2 · 玩法层（让“打不死”不再是玄学） 【中】 10. 陆上目标命中区可视化：高亮剩余命中区、标灰已摧毁区。否则玩家永远不知道自己差哪一块——这正是“某个指挥部砸了 30 多枚炸弹还不死”的真实原因（命中区没打全）。 11. 伤害模型说明，或者干脆补上溅射：弹药只有 Power / Penetration / CEP，没有任何 AoE 伤害字段（ImpactSize 只影响特效）。所以“精确制导炸弹炸了雷达、20 米外的房子毫发无伤”是模型使然；要么在文档里明说，要么真的加上溅射。 12. 修数据：us_cobra_dane_ewr / nato_ewr_station / nato_large_ewr_station —— 三个都是 Role=Target、值 500 分、却结构性不可摧毁；而它们在 72 个官方任务里出现 0 次，所以一直没人发现。 13. 武器数据卡：直接显示 Power / Penetration / CEP / 推荐目标类型。玩家不该为了搞懂“该用舰炮还是导弹”去逆向弹药文件。 14. 辅助幸存者搜救：游戏里搜索幸存者时，幸存者在夜间缺少明显标志。是否可以高亮标出幸存者的位置？例如按下 Z 使用载具视野、打开夜视模式后就能发现幸存者。 【EN】10) Hit-zone visualisation for land targets: highlight surviving zones and grey out destroyed ones, otherwise players can never tell what is left — the real reason a headquarters survives 30-plus bombs. 11) Document the damage model or add splash: ammunition has only Power / Penetration / CEP and no area-of-effect field (ImpactSize is purely visual), so a guided bomb killing a radar while a building 20 m away is untouched is by design. 12) Fix the data for us_cobra_dane_ewr / nato_ewr_station / nato_large_ewr_station: 500-point Role=Target units that are structurally indestructible, and they appear in none of the 72 official missions. 13) Weapon data cards showing Power / Penetration / CEP and the recommended target type. 14) CSAR assist: survivors have no clear marker at night; could they be highlighted, e.g. pressing Z for vehicle view with night vision? -- T3 · 战役 / 简报系统 【中】 15. FreeEvent 模板 + 实时预览 + 校验：官方已经有一套很好的排版报纸模板，但没有任何文档；内置预览可以避免“固定行高把图裁掉 60%”“文字重复”“节点孤立”。 16. 多语言字体自动回退：中文正文套 Times New Roman 会直接丢字（例如“能源与货▯”），引擎应当自动回退到系统中文字体。 17. 本地化预览（中文 / 日文 / 韩文）：现在只能靠肉眼看游戏里是否缺字。 【EN】15) FreeEvent templates with live preview and validation: the built-in newspaper layout is excellent but undocumented, and a preview would prevent a fixed row height cropping 60% of the image, duplicated text and orphaned nodes. 16) Automatic multi-language font fallback: Chinese in Times New Roman drops glyphs outright (energy and cargo ▯). 17) Localisation preview (CN/JP/KR) instead of checking by eye in game. -- T4 · 给 AI 生成任务的“最小可用集” 【中】 如果只能给一样，我建议是一个 headless 校验 / 查询 CLI： seapower-mission validate <mission.ini> # 水陆 / 引用 / 目标可摧毁性 / 航点风险 seapower-terrain query --rect <lat,lon,w,h> # 返回陆地面占比、最近水域距离 seapower-assets list --schema # 单位 / 弹药字段清单（机器可读） 有了它，“AI 写任务”才不是把风险转嫁给玩家——你原本担心的“旱地行舟”这类问题，本质上就是缺少合法性校验，而不是 AI 不够聪明。 【EN】If only one thing could be added, I would ask for a headless validation / query CLI: seapower-mission validate <mission.ini> # land/water, references, killability, waypoint risks seapower-terrain query --rect <lat,lon,w,h> # returns land fraction and distance to nearest water seapower-assets list --schema # machine-readable unit / ammunition field list With it, AI-written missions stop being a risk pushed onto players: the whole class of problems you worry about — units and zones ending up on dry land — is really a missing validity check, not a lack of AI cleverness.