From 81b100c0d3a11189392377f69aad8377f186ae1a Mon Sep 17 00:00:00 2001
From: Scout <scout@vector-hoops.local>
Date: Tue, 13 Oct 2026 00:00:23 -0500
Subject: [PATCH 1/2] Atlas search: rank suggestions by recency, not file order
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The suggestion sort only ordered identical names and left everything
else in index order — and the index is chronological, so a surname
query filled all 8 slots with the oldest namesakes: 'Curry' showed
Dell and Michael Curry seasons and never Stephen. Now: match tier
(name/word-start before substring), then most recent season first,
then name. Verified in a browser: 'Curry' now leads with Stephen
Curry 2025-26; 'Dell' and 'Jordan' behave consistently; no page
errors.
---
 assets/atlas-map.js        | 14 +++++++++++++-
 index.html                 |  2 +-
 public/assets/atlas-map.js | 14 +++++++++++++-
 public/index.html          |  2 +-
 4 files changed, 28 insertions(+), 4 deletions(-)

diff --git a/assets/atlas-map.js b/assets/atlas-map.js
index 0defbf64..5c37e285 100644
--- a/assets/atlas-map.js
+++ b/assets/atlas-map.js
@@ -330,8 +330,20 @@
         if (n.indexOf(q) === 0 || n.indexOf(' ' + q) > -1) starts.push(i);
         else if (n.indexOf(q) > -1) contains.push(i);
       }
+      // Rank: match tier first (name/word-start before substring), then
+      // MOST RECENT season first. The old comparator only ordered identical
+      // names and left everything else in file order — and the index is
+      // chronological, so a surname query filled the 8 slots with the
+      // OLDEST namesakes ("Curry" showed Dell and Michael, never Stephen).
+      var tier = {};
+      starts.forEach(function (i) { tier[i] = 0; });
+      contains.forEach(function (i) { tier[i] = 1; });
       var all = starts.concat(contains);
-      all.sort(function (a, b) { return P[a].n === P[b].n ? (P[b].s > P[a].s ? 1 : -1) : 0; });
+      all.sort(function (a, b) {
+        if (tier[a] !== tier[b]) return tier[a] - tier[b];
+        if (P[a].s !== P[b].s) return P[b].s > P[a].s ? 1 : -1;
+        return P[a].n < P[b].n ? -1 : P[a].n > P[b].n ? 1 : 0;
+      });
       results = all.slice(0, 8); active = results.length ? 0 : -1; render();
     });
     input.addEventListener('keydown', function (e) {
diff --git a/index.html b/index.html
index d439e771..0d3e7a82 100644
--- a/index.html
+++ b/index.html
@@ -359,7 +359,7 @@ <h2 class="label fig-label" style="margin:0 0 4px">Fresh from the lab <b>· new
 
 <script src="/assets/site-nav.js?v=1d646996"></script>
 <script src="/assets/mtnn.js?v=81d491f8"></script>
-<script src="/assets/atlas-map.js?v=7aad5d98"></script>
+<script src="/assets/atlas-map.js?v=d186fea5"></script>
 <script src="/assets/career-trails.js?v=a0e9e836"></script>
 <script src="assets/site-facts.js?v=cbb639f4" defer></script>
 <script>
diff --git a/public/assets/atlas-map.js b/public/assets/atlas-map.js
index 0defbf64..5c37e285 100644
--- a/public/assets/atlas-map.js
+++ b/public/assets/atlas-map.js
@@ -330,8 +330,20 @@
         if (n.indexOf(q) === 0 || n.indexOf(' ' + q) > -1) starts.push(i);
         else if (n.indexOf(q) > -1) contains.push(i);
       }
+      // Rank: match tier first (name/word-start before substring), then
+      // MOST RECENT season first. The old comparator only ordered identical
+      // names and left everything else in file order — and the index is
+      // chronological, so a surname query filled the 8 slots with the
+      // OLDEST namesakes ("Curry" showed Dell and Michael, never Stephen).
+      var tier = {};
+      starts.forEach(function (i) { tier[i] = 0; });
+      contains.forEach(function (i) { tier[i] = 1; });
       var all = starts.concat(contains);
-      all.sort(function (a, b) { return P[a].n === P[b].n ? (P[b].s > P[a].s ? 1 : -1) : 0; });
+      all.sort(function (a, b) {
+        if (tier[a] !== tier[b]) return tier[a] - tier[b];
+        if (P[a].s !== P[b].s) return P[b].s > P[a].s ? 1 : -1;
+        return P[a].n < P[b].n ? -1 : P[a].n > P[b].n ? 1 : 0;
+      });
       results = all.slice(0, 8); active = results.length ? 0 : -1; render();
     });
     input.addEventListener('keydown', function (e) {
diff --git a/public/index.html b/public/index.html
index d439e771..0d3e7a82 100644
--- a/public/index.html
+++ b/public/index.html
@@ -359,7 +359,7 @@ <h2 class="label fig-label" style="margin:0 0 4px">Fresh from the lab <b>· new
 
 <script src="/assets/site-nav.js?v=1d646996"></script>
 <script src="/assets/mtnn.js?v=81d491f8"></script>
-<script src="/assets/atlas-map.js?v=7aad5d98"></script>
+<script src="/assets/atlas-map.js?v=d186fea5"></script>
 <script src="/assets/career-trails.js?v=a0e9e836"></script>
 <script src="assets/site-facts.js?v=cbb639f4" defer></script>
 <script>

From f3d782d722484f4b485f9938f7245afade8d14d3 Mon Sep 17 00:00:00 2001
From: Scout <scout@vector-hoops.local>
Date: Tue, 13 Oct 2026 08:08:24 +0000
Subject: [PATCH 2/2] Fix ruff PTH/E731/format in facts/sitemap scripts
 (pathlib)

---
 scripts/build_site_facts.py | 23 ++++++++++++-----------
 scripts/build_sitemap.py    | 15 +++++++--------
 scripts/check_site_facts.py |  9 +++++----
 3 files changed, 24 insertions(+), 23 deletions(-)

diff --git a/scripts/build_site_facts.py b/scripts/build_site_facts.py
index ae31d756..38c17429 100644
--- a/scripts/build_site_facts.py
+++ b/scripts/build_site_facts.py
@@ -9,15 +9,19 @@
 
 Stdlib only. Every fact names its source file so a reader can recompute.
 """
+
 import json
-import os
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+
 
-ROOT = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
-A = lambda p: os.path.join(ROOT, "assets", p)
+def A(p):
+    return ROOT / "assets" / p
 
 
 def load(name):
-    with open(A(name), encoding="utf-8") as f:
+    with A(name).open(encoding="utf-8") as f:
         return json.load(f)
 
 
@@ -53,8 +57,8 @@ def compute_facts():
         "recall_at_10": meta["test_recall_at_10"],
         "purity_at_20": round(meta["purity_at_20"], 4),
         "archetype_top1": round(meta["archetype_top1_acc"], 4),
-        "onnx_bytes": os.path.getsize(A("mtnn.onnx")),
-        "embeddings_bytes": os.path.getsize(A("mtnn_embeddings.f32")),
+        "onnx_bytes": A("mtnn.onnx").stat().st_size,
+        "embeddings_bytes": A("mtnn_embeddings.f32").stat().st_size,
         "eligibility": vectors["eligibility"]["method"],
         "drift_pairs": len(drift["pairs"]),
         "drift_biggest": {
@@ -63,10 +67,7 @@ def compute_facts():
             "rotation_deg": pairs[0]["rotationDeg"],
             "shared_players": pairs[0]["sharedPlayers"],
         },
-        "drift_top5": [
-            {"from": p["from"], "to": p["to"], "rotation_deg": p["rotationDeg"]}
-            for p in pairs[:5]
-        ],
+        "drift_top5": [{"from": p["from"], "to": p["to"], "rotation_deg": p["rotationDeg"]} for p in pairs[:5]],
         "trails_players": len(trails),
         "insights_cards": len(insights["insights"]),
         "ghost_games": len(ghost["g"]),
@@ -82,7 +83,7 @@ def compute_facts():
 
 def main():
     out = compute_facts()
-    with open(A("site_facts.json"), "w", encoding="utf-8") as f:
+    with A("site_facts.json").open("w", encoding="utf-8") as f:
         json.dump(out, f, indent=1, sort_keys=False)
         f.write("\n")
     print("wrote assets/site_facts.json —", len(out["facts"]), "facts")
diff --git a/scripts/build_sitemap.py b/scripts/build_sitemap.py
index 8374cc1c..9dedf124 100644
--- a/scripts/build_sitemap.py
+++ b/scripts/build_sitemap.py
@@ -2,29 +2,28 @@
 """Build sitemap.xml from the site's 9 canonical routes. (Insights cards
 were removed from the sitemap in the 2026-10-12 focus cuts.)
 Run: python3 scripts/build_sitemap.py. Stdlib only."""
+
 import json
-import os
 from datetime import date
+from pathlib import Path
 
-ROOT = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
+ROOT = Path(__file__).resolve().parent.parent
 BASE = "https://hoops.dumbmodel.com"
-PAGES = ["/", "/play", "/live", "/players", "/insights", "/taxonomy",
-         "/methods", "/props-lab", "/playoff-runs"]
+PAGES = ["/", "/play", "/live", "/players", "/insights", "/taxonomy", "/methods", "/props-lab", "/playoff-runs"]
 
 
 def main():
-    with open(os.path.join(ROOT, "assets", "insights.json"), encoding="utf-8") as f:
+    with (ROOT / "assets" / "insights.json").open(encoding="utf-8") as f:
         slugs = [c["slug"] for c in json.load(f)["insights"]]
     # Insight stubs stay live for social unfurls but leave the sitemap
     # (Wave 5): the hub page is the indexable surface.
     urls = list(PAGES)
     today = date.today().isoformat()
-    lines = ['<?xml version="1.0" encoding="UTF-8"?>',
-             '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">']
+    lines = ['<?xml version="1.0" encoding="UTF-8"?>', '<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">']
     for u in urls:
         lines.append(f"  <url><loc>{BASE}{u}</loc><lastmod>{today}</lastmod></url>")
     lines.append("</urlset>")
-    with open(os.path.join(ROOT, "sitemap.xml"), "w", encoding="utf-8") as f:
+    with (ROOT / "sitemap.xml").open("w", encoding="utf-8") as f:
         f.write("\n".join(lines) + "\n")
     print(f"wrote sitemap.xml — {len(urls)} URLs ({len(PAGES)} pages)")
 
diff --git a/scripts/check_site_facts.py b/scripts/check_site_facts.py
index 3243dec5..674049b4 100644
--- a/scripts/check_site_facts.py
+++ b/scripts/check_site_facts.py
@@ -2,20 +2,21 @@
 """Gate: assets/site_facts.json must equal a fresh compute from the served
 data files. Fails (exit 1) when data moved and the sheet was not rebuilt —
 the same contract as stamp_assets.py --check. Stdlib only."""
+
 import json
-import os
 import sys
+from pathlib import Path
 
-sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+sys.path.insert(0, str(Path(__file__).resolve().parent))
 from build_site_facts import compute_facts, A
 
 
 def main():
     path = A("site_facts.json")
-    if not os.path.exists(path):
+    if not path.exists():
         print("FAIL assets/site_facts.json missing — run scripts/build_site_facts.py")
         return 1
-    with open(path, encoding="utf-8") as f:
+    with path.open(encoding="utf-8") as f:
         current = json.load(f)
     fresh = compute_facts()
     if current.get("facts") == fresh["facts"]: