Methodology 数据说明

Data & definitions — what each figure means and when it refreshes · 英文正文 / 中文摘要

This almanac is assembled from the weekly echo statistics and two published data modules; each release is mirrored in the site's data directory and pinned to the tag listed on the changelog. Nothing in the tables is hand-entered.

本站数据由每周回响统计与两个已发布的数据模块组成;每个版本都镜像在站内数据目录,并固定到 changelog 中列出的 tag。表格中的数字均不做手工录入。

Data modules 数据模块

  • The weekly echo table — win rate and pick rate for every echo and bracket — is archived daily as raw snapshots in the repository archive/ (not deployed) and rebuilt into the aggregated trends.json the site actually reads. Each snapshot carries its own period label; that label becomes the data week.

    周胜率 / 选取率数据每日归档为仓库 archive/ 下的原始快照(不部署),再聚合成站点读取的 trends.json;每个快照自带周期标签,即站内的「期」。

  • Cards and echoes are published as two independent release modules in hpma-bits/hpma-spellbook: cards (cards-v*) for cost, type, rarity, every level and the bilingual text, and echoes (echoes-v*) for the echo list, rarity variants and affixes. The site pins each release in the module's version.json, verifies the downloaded files against the release checksums, and refreshes both modules weekly.

    卡牌与回响以 hpma-spellbook 数据仓的两个独立模块发布:cards-v*(费用、类型、品质、全部等级与中英文文案)与 echoes-v*(回响列表、品质变体与词条)。站点在各自 version.json 中固定版本,按发布 checksums 校验,并每周同步。

  • The Magic Quiz bank is published by hpma-quizbank as quiz-v* releases. The dataset keeps the correct answer in the first option; the site shuffles options on display and always follows the is_correct flag, so a reshuffled question cannot drift from its answer.

    魔法测验题库由 hpma-quizbank 以 quiz-v* 发布。数据集中正确答案固定于第一个选项;站内展示时重新洗牌,并始终以 is_correct 为准,避免洗牌后答案错位。

Card art and echo icons are hot-linked from the immutable release tag on raw.githubusercontent.com; the same-origin /cardimg/* and /echoimg/* redirects are the fallback for networks that cannot reach GitHub. Artwork stays © NetEase and the rights holders.

Fields 字段口径

Every echo row is one echo × one bracket × one period. Three values do all the work:

Win rate winrate
The share of matches won with the echo in a given period, exactly as reported in the dataset. The site prints one decimal place and does not re-weight by bracket size or sample count.
Pick rate attendancerate
The share of matches the echo was used in. The dataset field and the README call it attendance rate; the charts and tables label the same number pick rate. It is a usage share, not a raw number of selections.
Delta (Δ) win(t) − win(t−1)
The week-over-week move for the same echo in the same bracket, in percentage points. The front page rounds it to one decimal and treats a missing previous row as 0; the standings table prints an arrow and the absolute move, or — when there is nothing to compare.

每行数据 = 一个回响 × 一个 bracket × 一个周期。胜率(winrate)是该周期内使用该回响的胜场占比;选取率在数据集字段中叫 attendancerate(README 称 attendance rate),指该回响在比赛中的出场占比,不是出场次数。Δ 是同回响、同档位相邻两期的百分点差:首页保留一位小数,上一期缺失按 0 处理;排行榜显示箭头与绝对值,无可比数据时显示「—」。站点不做插值、补行或加权。

Brackets 四档

The dataset splits the ladder into four brackets at 7,500 rating points. The site keeps all four separate; every chart, plate and standings table has a bracket switch, and no view averages across brackets.

The four rating brackets used across the site
BracketQueueRatingDataset label
1v1 Low1v1Below 7,5001v1:7500分以下
1v1 High1v17,500 and above1v1:7500分以上
2v2 Low2v2Below 7,5002v2:7500分以下
2v2 High2v27,500 and above2v2:7500分以上

数据集按 7,500 分把匹配分成四档:1v1 低分(低于 7500)、1v1 高分(7500 及以上)、2v2 低分、2v2 高分。站内四档始终分开呈现,不做跨档平均。

Update cadence 更新节奏

Scheduled workflows keep the data current; the deploy is triggered by the resulting commit. Times below are UTC.

Data refresh schedule
DataRefreshMechanism
Echo statisticsDaily archive · 02:30 UTCarchive workflow; raw snapshot → trends.json
Cards & echoesMondays · 03:00 UTCSpellbook Data Update workflow
QuizMondays · 03:30 UTCQuiz Data Update workflow
Site buildOn every push to mainNetlify runs the prerender build

回响统计每日 02:30 UTC 归档,周期随数据仓换周更新;卡牌与回响每周一 03:00 UTC 同步;题库每周一 03:30 UTC 同步;数据提交后 Netlify 自动执行预渲染构建并部署。定时任务可能被 GitHub 延迟,延迟只会推后数据,不改变口径。

Disclaimer 免责声明

An unofficial, non-commercial fan project. Not affiliated with or endorsed by NetEase or Warner Bros. Game data, card art, echo icons and quiz text remain © their rights holders and are shown for personal research and reference only.

Figures describe the data sample, not the whole player base. A one-week move can be noise; treat small differences between echoes cautiously rather than as proof of a stronger or weaker card.

本站为非官方、非商业的同人项目,与网易及华纳兄弟无隶属或背书关系。游戏数据、卡面、回响图标与题目文案版权归各自权利人所有,仅供个人研究参考。数字反映的是数据样本而非全体玩家;单周波动可能只是噪声,小幅差距不宜当作强弱结论。

Spotted a wrong number or an out-of-date tag? Open an issue on GitHub ↗