existence-trace · Methodology: PAD, G, XP

PAD情绪的形状The shape of mood

愉悦 P、激活 A、掌控 D 三个轴，每个轴是 0–5 的连续值。Pleasure P, Arousal A and Dominance D, each a continuous value from 0 to 5.

读数可以带精确时间（HH:MM）或只标时段（早、中、下午、其他）。只标时段的读数画在固定位置，图注会说明。A reading has an exact time (HH:MM) or only a slot (morning, noon, afternoon, other). Slot-only readings are drawn at fixed positions and the caption says so.
曲线是高斯核局部线性拟合：一日图带宽 2.5 小时，一周图带宽 0.75 天，结果裁剪到 0–5。没有散点，没有虚线。Lines are Gaussian-kernel local linear fits: 2.5-hour bandwidth for a day, 0.75 days for a week, clamped to 0-5. No dots, no dashed lines.
27 个情绪词只是粗投影：每个轴四舍五入后 0–2 为低、3 为中、4–5 为高。词永远不替代连续值。The 27 mood words are only a coarse projection: after rounding, 0-2 is low, 3 is mid and 4-5 is high on each axis. A word never replaces the continuous values.

原方法论目前没有写出逐轴评分锚点、27 词与 125 词映射、校准程序与准确性验证。本页不补写这些内容。下表只是本仓库预设使用的粗投影表，不是经过验证的映射。The source methodology does not yet define per-axis scoring anchors, a 27- or 125-word mapping, a calibration procedure or an accuracy validation. This page does not fill them in. The table below is only the coarse projection this repository's preset uses, not a validated mapping.

P	A	D 低low	D 中mid	D 高high
P 低low	A 低low	无助helpless	沮丧dejected	轻蔑contemptuous
P 低low	A 中mid	担忧worried	烦躁irritated	不满dissatisfied
P 低low	A 高high	恐慌panicked	紧张tense	愤怒angry
P 中mid	A 低low	茫然blank	平静calm	淡定composed
P 中mid	A 中mid	困惑confused	专注focused	笃定assured
P 中mid	A 高high	震惊shocked	惊讶surprised	警觉alert
P 高high	A 低low	依恋attached	满足content	惬意at ease
P 高high	A 中mid	感动moved	开心happy	自豪proud
P 高high	A 高high	惊喜delighted	兴奋excited	豪迈exhilarated

G稳定与探索Stability and exploration g-1.4.0

G 是根据日记证据编码的单人纵向复合指数，不是临床量表，也不是对人的客观真值。G is a single-person longitudinal composite coded from journal evidence; it is not a clinical scale and not an objective truth about a person.

B = 0.28 B1 + 0.18 B2 + 0.36 B3 + 0.18 B4；E = 0.22 E1 + 0.14 E2 + 0.36 E3 + 0.28 E4；G = (B + E) / 2B = 0.28 B1 + 0.18 B2 + 0.36 B3 + 0.18 B4; E = 0.22 E1 + 0.14 E2 + 0.36 E3 + 0.28 E4; G = (B + E) / 2
刻度 0–10。全局锚点：Scale 0-10. Global anchors: 0 明确而强烈的不足clear and strong shortfall · 2 明显不足clearly short · 5 支持与阻力并存support and resistance side by side · 8 较充分mostly sufficient · 10 明确而充分clear and sufficient

八个细项与逐项锚点Eight items and their anchors

细项Item	权重Weight	0–2	5	8–10
B 稳定B stability
B1 睡眠与恢复Sleep and recovery	0.28	明确失眠、严重压缩或几乎未恢复clear insomnia, badly cut short or barely recovered	有休息，也有不足或质量不明some rest, with shortfall or unclear quality	恢复充分且本人感到被修复fully recovered and feels restored
B2 身体精力Physical energy	0.18	明确疲惫、疼痛或难以承担日常clearly exhausted, in pain or unable to carry the day	有能量也有身体成本some energy, with a bodily cost	精力充足、身体余力明显plenty of energy, clear physical reserve
B3 安全与支持Safety and support	0.36	明确不安全、孤立或支持断裂clearly unsafe, isolated or cut off from support	支持与威胁/孤立并存support alongside threat or isolation	关系与环境带来清楚承接relationships and surroundings clearly hold you
B4 注意力与负担Attention and load	0.18	明显被多线、义务或反复思考占满clearly swamped by parallel threads, duties or rumination	能推进，但切换与负担明显can move forward, but switching and load are clear	注意力自主、负担可控且有余地attention is self-directed, load is manageable with room to spare
E 探索E exploration
E1 内在兴趣Intrinsic interest	0.22	几乎只有被迫执行或明确厌倦almost only forced execution or clear boredom	兴趣与义务混合interest mixed with obligation	持续被问题、声音或创作吸引steadily drawn in by a problem, a sound or making something
E2 自主与自由Autonomy and freedom	0.14	被外界要求或证明冲动主导driven by outside demands or the urge to prove yourself	有选择，也有明显外部牵引some choice, with a clear outside pull	能按自身愿望选择、拒绝或无功利用途地尝试can choose, decline or try things for no purpose, by your own wish
E3 期待与意义Anticipation and meaning	0.36	明天缺少可感知方向或明确绝望no felt direction for tomorrow, or clear despair	有下一步，但意义感不稳定a next step exists, but the sense of meaning is unsteady	对明天有本人认可的期待与方向an expectation and direction for tomorrow you endorse yourself
E4 边界与可持续性Boundaries and sustainability	0.28	没有停止点、代价明显不可承受no stopping point, the cost is clearly unbearable	开始设边界但执行不稳starting to set boundaries, but keeping them is unsteady	能停、能退出、强度可长期承受can stop, can step out, the intensity is sustainable

缺失与补值Missing values and gap filling

一组四项都有才算该组，两组都有才算 G；否则显示为「缺」，不当作 0。计算中不取整，只在显示时保留一位小数。A group is computed only when all four items exist, and G only when both groups exist; otherwise it shows n/a, never 0. Nothing is rounded during computation; display uses one decimal.
补值按固定顺序逐项进行：historical_bridge（前后真实值线性插值）→ continuity_carry_forward（沿用最近的前值）→ scale_prior（5.0）。补值一律低置信度，带上方法和邻近日期，整天不进入训练和准确率统计。Gaps are filled per item in a fixed order: historical_bridge (linear interpolation between real values) → continuity_carry_forward (nearest earlier value) → scale_prior (5.0). Fills are always low confidence, carry their method and neighbour dates, and keep the whole day out of training and accuracy.
后来的真实值改变了补值时，追加一个新修订，旧修订保留。When a later real value changes a fill, a new revision is appended and old revisions stay.

先冻结，后校准Freeze first, calibrate later

每天的估计先冻结成快照。只有在快照之后表达的反馈才能关联到它；只有用户明确给出的数字才是标签，文字只作证据。Each day's estimate is frozen into a snapshot first. Only feedback expressed after the snapshot links to it; only numbers the user explicitly gave are labels, text is evidence only.
反馈可以针对细项、B、E 或 G。设计向量：细项是单位向量，B/E 是该组权重，G 是两组权重各乘 0.5。Feedback may target an item, B, E or G. Design vectors: a unit vector for an item, the group weights for B or E, and both groups' weights x 0.5 for G.
Ridge 回归为每个细项拟合一个加性偏移：λ = 5；样本权重 2^−Δ天/60（半衰期 60 天）；同一天的标注合计权重为 1；偏移裁剪到 ±2，校准后分数裁剪到 0–10。Ridge regression fits one additive offset per item: λ = 5; sample weight 2^−Δdays/60 (60-day half-life); the labels of one day share a total weight of 1; offsets are clipped to ±2 and corrected scores to 0-10.
前瞻评估：前 10 个标注日之后的每一天，只用更早的日子拟合再预测。MAE 先在一天内平均，再跨天平均。快照晚于当天次日冻结的回顾性回填可以参与训练，但不计入前瞻准确率。Prospective evaluation: each labelled day after the first 10 is predicted with offsets fitted only on earlier days. MAE is averaged within a day first, then across days. Retrospective backfills (snapshots frozen later than the next day) may train but never count towards prospective accuracy.
门槛：至少 10 个训练日 + 10 个前瞻测试日，且 MAE 改善 ≥ 0.2 才启用；最近 10 个测试日恶化 ≥ 0.3 就停用。Gate: at least 10 training days + 10 prospective test days and an MAE gain of ≥ 0.2 to switch on; a loss of ≥ 0.3 over the last 10 test days switches it off.
不同评分标准版本的标注不混用。版本变化必须升级版本号，不静默重算历史。时区默认 America/Chicago，可在 trace.toml 中修改。Labels from different rubric versions are never mixed. Any change bumps the version; history is never silently recomputed. The default timezone is America/Chicago and can be changed in trace.toml.

趋势图Trend chart

etrace g chart DATE --days N 画 B、E、G 三条平滑线：高斯核局部线性，带宽 4.5 天；纵轴在数据范围外扩 0.2，按 0.2 对齐，跨度至少 1.2，并保持在 0–10 内；少于 3 个点的线不画。etrace g chart DATE --days N draws three smooth lines for B, E and G: Gaussian-kernel local linear, 4.5-day bandwidth; the y axis pads the data range by 0.2, aligns to 0.2, spans at least 1.2 and stays within 0-10; a line with fewer than 3 points is not drawn.

维度（default 预设）Dimension (default preset)	上限Cap	L1	L2	L3	L4
节律与稳定Rhythm and stability	18	4.5	9	13.5	18
学习与吸收Learning and absorbing	16	4	8	12	16
边界与休息Boundaries and rest	16	4	8	12	16
行动与尝试Action and trying	14	3.5	7	10.5	14
反思与发现Reflection and discovery	14	3.5	7	10.5	14
联结与探索Connection and exploration	12	3	6	9	12

三个视角：PAD、G、XPThree lenses: PAD, G and XP