Score every metric equally in the overall total (#218)

The overall climate-shift score was a weighted roll-up (feels-like and humidity
2x, wind and gusts 0.5x, and so on). Make it an equal average instead: every
metric — and, through each metric's own mean over the 10/25/50/75/90 percentiles,
every percentile — counts the same.

- scoring._overall: plain mean of the present metrics' divergences, mapped to
  0-100, replacing the weighted sum. Missing metrics still drop out cleanly.
- Drop the now-unused per-metric weights from the METRICS table and the scored
  entries (nothing rendered them; only the roll-up read them).
- views.SCORE_VER s4 -> s5 so cached weighted scores are recomputed.
- score.js: the hero note now reads "Every metric counts equally" instead of
  "Temperature, feels-like and humidity are weighted most".

Verified against Seattle: overall mad equals the equal mean of the eight metric
mads (7.9), and the entries no longer carry a weight.

Claude-Session: https://claude.ai/code/session_013dRZmX9D3JEntfMKWMTWZ8

data/scoring.py

@@ -7,8 +7,8 @@ between where it lands and ``q`` — in percentile points, so it is unit-free an
 comparable across metrics — is the divergence. A metric's score is the mean
 absolute divergence over the categories, mapped to 0-100; its sign (bias) says
 which way it drifted. Metrics are computed per meteorological season and the
-seasonal differentials are averaged into an annual view, then weighted (temps /
-humidity / feels-like heaviest) into one overall score.
+seasonal differentials are averaged into an annual view, then averaged equally
+across all metrics into one overall score.
 
 Baseline is the ENTIRE record, including the recent years — the same all-years
 climatology every other view grades against. The recent window's ~13% overlap
@@ -26,21 +26,20 @@ SEASONS = {"djf": (12, 1, 2), "mam": (3, 4, 5),
            "jja": (6, 7, 8), "son": (9, 10, 11)}
 SLICES = ("annual", *SEASONS)                 # payload slice keys
 
-# Per-metric descriptor — display label, overall-score weight, and the
-# (positive-drift, negative-drift) direction words — in ONE table so a scored
-# metric is defined in a single place. Temps / humidity / feels-like dominate the
-# weighting; wet bulb medium; wind + precip lightest. wetbulb is derived at the
-# read boundary (climate.py); the rest are raw daily columns. Insertion order is
-# the canonical compute order (the frontend renders in its own display order).
+# Per-metric descriptor — display label and the (positive-drift, negative-drift)
+# direction words — in ONE table so a scored metric is defined in a single place.
+# The overall score weights every metric equally (see _overall). wetbulb is derived
+# at the read boundary (climate.py); the rest are raw daily columns. Insertion order
+# is the canonical compute order (the frontend renders in its own display order).
 METRICS = {
-    "tmax":    {"label": "High temp",  "weight": 1.5,  "dir": ("warmer", "cooler")},
-    "tmin":    {"label": "Low temp",   "weight": 1.5,  "dir": ("warmer", "cooler")},
-    "feels":   {"label": "Feels like", "weight": 2.0,  "dir": ("warmer", "cooler")},
-    "humid":   {"label": "Humidity",   "weight": 2.0,  "dir": ("muggier", "drier")},
-    "wetbulb": {"label": "Wet bulb",   "weight": 1.25, "dir": ("warmer", "cooler")},
-    "wind":    {"label": "Wind",       "weight": 0.5,  "dir": ("windier", "calmer")},
-    "gust":    {"label": "Gusts",      "weight": 0.5,  "dir": ("gustier", "calmer")},
-    "precip":  {"label": "Precip",     "weight": 0.75, "dir": ("wetter", "drier")},
+    "tmax":    {"label": "High temp",  "dir": ("warmer", "cooler")},
+    "tmin":    {"label": "Low temp",   "dir": ("warmer", "cooler")},
+    "feels":   {"label": "Feels like", "dir": ("warmer", "cooler")},
+    "humid":   {"label": "Humidity",   "dir": ("muggier", "drier")},
+    "wetbulb": {"label": "Wet bulb",   "dir": ("warmer", "cooler")},
+    "wind":    {"label": "Wind",       "dir": ("windier", "calmer")},
+    "gust":    {"label": "Gusts",      "dir": ("gustier", "calmer")},
+    "precip":  {"label": "Precip",     "dir": ("wetter", "drier")},
 }
 SCORE_METRICS = tuple(METRICS)
 
@@ -174,7 +173,7 @@ def _direction(metric: str, bias: float) -> str:
 
 def _null_entry(metric: str, reason: str) -> dict:
     return {"key": metric, "label": METRICS[metric]["label"], "score": None,
-            "weight": METRICS[metric]["weight"], "reason": reason}
+            "reason": reason}
 
 
 def _build_entry(metric: str, mad: float, bias: float, per_q: list,
@@ -189,7 +188,7 @@ def _build_entry(metric: str, mad: float, bias: float, per_q: list,
         "score": score, "mad": round(mad, 1), "bias": round(bias, 1),
         "tier": tier, "class": css, "direction": direction,
         "grade": tier if score < 15 else f"{tier}, {direction}",
-        "weight": METRICS[metric]["weight"], "per_q": per_q,
+        "per_q": per_q,
         "n_recent": n_recent, "n_base": n_base,
     }
     if freq is not None:
@@ -206,8 +205,9 @@ def _metric_entry(metric: str, base: np.ndarray, rec: np.ndarray) -> dict:
 
 
 def _overall(entries: dict) -> dict | None:
-    """Weighted roll-up of the present metrics in one slice. Weights renormalize
-    over whatever is present (a metric missing for the source drops out cleanly).
+    """Equal-weight roll-up of the present metrics in one slice: every metric (and,
+    through each metric's own mean, every percentile) counts the same. Metrics
+    missing for the source simply drop out of the average.
 
     The total reads as a direction-agnostic NET CHANGE: colored by magnitude on the
     intensity ramp and labeled by tier alone, not 'warmer/cooler' — the per-metric
@@ -215,8 +215,7 @@ def _overall(entries: dict) -> dict | None:
     present = [e for e in entries.values() if e.get("score") is not None]
     if not present:
         return None
-    wsum = sum(e["weight"] for e in present)
-    mad = sum(e["weight"] * e["mad"] for e in present) / wsum
+    mad = sum(e["mad"] for e in present) / len(present)
     score = score_of(mad)
     tier, css = tier_of(score, 1.0)   # magnitude only — net change, not a direction
     return {"score": score, "mad": round(mad, 1), "tier": tier, "class": css,

tests/data/test_scoring.py

@@ -160,14 +160,26 @@ def test_wetbulb_stress_frequency_rises_with_heat():
     assert freq["f6"] > freq["f45"] and freq["d"] > 0
 
 
-# --- missing metric / weight renormalization ----------------------------------
+# --- overall aggregation ------------------------------------------------------
+
+def test_overall_is_equal_average_of_metric_mads():
+    # Every present metric counts the same — no per-metric weighting.
+    entries = {
+        "a": {"score": 20, "mad": 3.0},
+        "b": {"score": 80, "mad": 12.0},
+        "c": {"score": None},              # missing -> dropped from the average
+    }
+    ov = scoring._overall(entries)
+    assert ov["mad"] == 7.5                                # (3.0 + 12.0) / 2
+    assert ov["score"] == scoring.score_of(7.5) == 50
+
 
 def test_missing_gust_column_nulls_out_but_overall_survives():
     out = scoring.build_scores(make_frame(drop=("gust",)))
     ann = out["slices"]["annual"]
     g = ann["metrics"]["gust"]
     assert g["score"] is None and "not available" in g["reason"]
-    assert ann["overall"]["score"] is not None          # renormalized over present metrics
+    assert ann["overall"]["score"] is not None          # averaged over present metrics
 
 
 def test_all_null_metric_reports_not_enough_data():

web/views.py

@@ -108,7 +108,7 @@ def forecast_key(today, days: int) -> str:
 # is the plain `history_token` (expires when the archive tail advances). The key
 # just carries the scoring-math version, so a math change invalidates only score
 # rows without disturbing any other kind.
-SCORE_VER = "s4"
+SCORE_VER = "s5"   # s5: overall is an equal average of metrics (was weighted)
 
 
 def score_key() -> str: