diff --git a/merge-performance/pirate-hex-words b/merge-performance/pirate-hex-words
index 816c6b6..21133a0 100644
--- a/merge-performance/pirate-hex-words
+++ b/merge-performance/pirate-hex-words
@@ -16,20 +16,20 @@ b100d	       blood
 
 acc057	       accost	    
 af10a7	       afloat	    
-a70115	       atolls	    
+a70115	       atolls	    !5
 ba771e	       battle	    
-befa11	       befall	    
-b1ade5	       blades	    
+befa11	       befall	    !3
+b1ade5	       blades	    !2
 b1a575	       blasts	    
 b1eed5	       bleeds	    
 b0771e	       bottle
-c0a575	       coasts	    
+c0a575	       coasts	    !1
 deface	       deface	    
-defea7	       defeat	    /
-fab1ed	       fabled	    
-fea575	       feasts	    /
+defea7	       defeat	    .
+fab1ed	       fabled	    !6
+fea575	       feasts	    .
 f1ee75	       fleets	    .
-1007ed	       looted	    (!)
+1007ed	       looted	    !4
 
 ab5ce55	       abscess	    
 acc0575	       accosts	    
@@ -52,9 +52,9 @@ acc057ed       accosted
 b0a710ad       boatload    
 c01055a1       colossal    
 decea5ed       deceased    
-0b57ac1e       obstacle    .
-5caff01d       scaffold    .
-5eac0a57       seacoast    .
+0b57ac1e       obstacle    X
+5caff01d       scaffold    X
+5eac0a57       seacoast    X
 
 ab5ce55ed      abscessed   
 ab5ce55e5      abscesses   
diff --git a/merge-performance/principles-to-cover.txt b/merge-performance/principles-to-cover.txt
index 3e7bbcf..ce81f27 100644
--- a/merge-performance/principles-to-cover.txt
+++ b/merge-performance/principles-to-cover.txt
@@ -19,22 +19,22 @@ Cleanups:
 
 Savings:
   * Avoiding rework based on understanding the algorithm:
-    * hash comparison for exact renames (and not throwing away that work)
-    * remove source file from comparisons if other side didn't modify
-    * use file basename to avoid exhaustive O(N^2) comparison
+    X hash comparison for exact renames (and not throwing away that work)
+    X remove source file from comparisons if other side didn't modify
+    X use file basename to avoid exhaustive O(N^2) comparison
     * re-use rename information while rebasing or cherry-picking
     * trivial tree merges (all match) and partial merges (side match)
     * avoid (dir) rename detection when traversal paths shows it's unnecessary
       (no directory removed, or none added: no dir rename)
     * avoid updating the index or working tree if not needed (esp. in rebase)
   * Brute force avoid unnecessary work or rework:
-    * tree traversals (old: {unpack:1(3), unique:2, diff:2(4),
+    X tree traversals (old: {unpack:1(3), unique:2, diff:2(4),
                              get_tree_entry:3*2*#renames},
                        new: {basic:1(3-partial), unpack:1(2)})
-    * repeatedly drop & reload cache (sometimes not necessary at all)
+    X repeatedly drop & reload cache (sometimes not necessary at all)
     * avoid re-comparing hashes (unpack_trees did it once, then we repeated)
     * memoize not-found results from previous runs of check_dir_renamed()
       (store dir_renames + non-renames in a strmap)
   * Avoid O(N^2):
-    * The idea behind remove_marked_cache_entries(); not used by merge-recursive
-    * string_list_insert (see NOTE in string-list.h)
+    X The idea behind remove_marked_cache_entries(); not used by merge-recursive
+    X string_list_insert (see NOTE in string-list.h)