1 · Attestation network — the 58-formula bank一 · 征引网络：58 条公式库
Each formula in the closed vocabulary, with the documents in which it was detected. Counts are attestations in this index — the number of inscriptions where the normalised phrase matched.封闭词表中的每条公式，及其被检出的文献。计数为本索引中的征引数，即规范化短语匹配到的铭文数量。
⚠ Corpus bias is inline, not a footnote. The I.PHI and EDCS corpora are a survival- and digitization-biased sample, not a census. A count measures attestation in this index, never historical frequency; a formula absent here is "not in this sample", never "did not exist". Matching is deterministic key/variant matching over a closed vocabulary — high precision, but it finds only what the bank already names (see §2 for what it doesn't).⚠ 语料偏差为正文，非脚注。 I.PHI 与 EDCS 语料是经存留与数字化偏选的样本，非全数。计数衡量的是本索引中的征引，绝非历史频率；此处缺席的公式意为"不在此样本中"，绝非"不曾存在"。匹配为封闭词表上的确定性键／变体匹配：精度高，但只能找到库中已命名者（其余见第二节）。
2 · Candidate formulae — proposed, awaiting review二 · 候选公式：已提出，待审核
3 · Period drift — formula attestation across centuries三 · 公式漂移：公式征引随世纪之变
Each formula's attestation across the dated decree corpus, by century. Cells show distinct decrees attesting the formula in that century (darker = more); hover for span and lacunose counts. A purely deterministic aggregation over the dossier's dated shards — no model.每条公式在已纪年法令语料中按世纪的征引。格内为该世纪中征引此公式的不同法令数（越深越多）；悬停见词例与残缺数。纯属对文献集已纪年残篇的确定性统计，无模型。
⚠ This is the decree corpus (6,617 shards, 6,500 Greek / 117 Latin) — a survival- and digitization-biased sample dominated by Athenian/Greek civic decrees; counts are attestation in this sample, not historical frequency. The Latin slice is too small for robust Latin drift. 2,247 decrees (34%) are undated — held in a separate hatched column, never bucketed into a century. 37% of spans are lacunose (counted, surfaced on hover). No period is assigned beyond the shard's own
date.century_bucket.⚠ 此为法令语料（6,617 残篇，希腊 6,500／拉丁 117），系经存留与数字化偏选、以雅典／希腊城邦法令为主之样本；计数为本样本中之征引，非历史频率。拉丁部分过小，漂移不可靠。2,247 篇（34%）无纪年，置于单独斜纹列，绝不归入世纪。37% 词例残缺（已计入，悬停可见）。除残篇自带之 date.century_bucket 外，不擅断年代。