1 files changed, 46 insertions, 0 deletions
diff --git a/src/eval_tune/compare_ecm_depth.py b/src/eval_tune/compare_ecm_depth.py
new file mode 100644
index 0000000..51329a7
--- /dev/null
+++ b/src/eval_tune/compare_ecm_depth.py
@@ -0,0 +1,46 @@
+#!/usr/bin/env python3
+"""Pair up per-problem achieved-depth (final iterated depth reached before
+the node-count budget ran out) from two sn=N ECM stdout logs (same input
+file, same order) and report win/loss/tie counts plus median depth delta.
+"""
+import re
+import statistics
+import sys
+
+TELL_RE = re.compile(r"tellothers d(\d+),")
+
+
+def parse(path):
+    """Return list of achieved-depth ints, one per completed problem, in order."""
+    depths = []
+    with open(path) as f:
+        for line in f:
+            m = TELL_RE.search(line)
+            if m:
+                depths.append(int(m.group(1)))
+    return depths
+
+
+def main():
+    base_path, cand_path = sys.argv[1], sys.argv[2]
+    base = parse(base_path)
+    cand = parse(cand_path)
+    n = min(len(base), len(cand))
+    base, cand = base[:n], cand[:n]
+
+    deltas = [c - b for b, c in zip(base, cand)]
+    wins = sum(1 for d in deltas if d > 0)
+    losses = sum(1 for d in deltas if d < 0)
+    ties = sum(1 for d in deltas if d == 0)
+
+    print(f"Paired problems: {n}")
+    print(f"  candidate deeper: {wins}")
+    print(f"  candidate shallower: {losses}")
+    print(f"  tied depth: {ties}")
+    print(f"  median depth delta (candidate - baseline): {statistics.median(deltas):+.1f}")
+    print(f"  mean depth delta: {statistics.mean(deltas):+.3f}")
+    print(f"  sum of depth delta: {sum(deltas):+d}")
+
+
+if __name__ == "__main__":
+    main()