diff --git a/src/ranked_match.cc b/src/ranked_match.cc
index 03cb2fcd..9acaa8b4 100644
--- a/src/ranked_match.cc
+++ b/src/ranked_match.cc
@@ -1,65 +1,113 @@
 #include "ranked_match.hh"
 
+#include "unit_tests.hh"
+
 namespace Kakoune
 {
 
-static bool match_rank(StringView candidate, StringView query)
+static int count_word_boundaries_match(StringView candidate, StringView query)
 {
-    int rank = 0;
-    auto it = candidate.begin();
+    int count = 0;
+    auto it = query.begin();
     char prev = 0;
-    for (auto c : query)
+    for (auto c : candidate)
     {
-        if (it == candidate.end())
-            return 0;
-
-        const bool islow = islower(c);
-        auto eq_c = [islow, c](char ch) { return islow ? tolower(ch) == c : ch == c; };
-
-        if (eq_c(*it)) // improve rank on contiguous
-            ++rank;
-
-        while (!eq_c(*it))
-        {
-            prev = *it;
-            if (++it == candidate.end())
-                return 0;
-        }
-        // Improve rank on word boundaries
-        if (prev == 0 or prev == '_' or
-            (islower(prev) and isupper(*it)))
-            rank += 5;
-
+        const bool is_word_boundary = prev == 0 or
+                                      (ispunct(prev) and is_word(c)) or
+                                      (islower(prev) and isupper(c));
         prev = c;
-        ++rank;
+
+        if (not is_word_boundary)
+            continue;
+
+        const char lc = tolower(c);
+        for (; it != query.end(); ++it)
+        {
+            const char qc = *it;
+            if (qc == (islower(qc) ? lc  : c))
+            {
+                ++count;
+                ++it;
+                break;
+            }
+        }
+        if (it == query.end())
+            break;
+    }
+    return count;
+}
+
+static bool smartcase_eq(char query, char candidate)
+{
+    return query == (islower(query) ? tolower(candidate) : candidate);
+}
+
+static bool subsequence_match_smart_case(StringView str, StringView subseq)
+{
+    auto it = str.begin();
+    for (auto& c : subseq)
+    {
+        if (it == str.end())
+            return false;
+        while (not smartcase_eq(c, *it))
+        {
+            if (++it == str.end())
+                return false;
+        }
         ++it;
     }
-    return rank;
+    return true;
 }
 
 RankedMatch::RankedMatch(StringView candidate, StringView query)
 {
-    if (candidate.empty() or query.empty())
+    if (candidate.empty() or query.length() > candidate.length())
+        return;
+
+    if (query.empty())
     {
         m_candidate = candidate;
         return;
     }
 
-    m_match_rank = match_rank(candidate, query);
+    if (not subsequence_match_smart_case(candidate, query))
+        return;
+
+    m_candidate = candidate;
+
+    m_first_char_match = smartcase_eq(query[0], candidate[0]);
+    m_word_boundary_match_count = count_word_boundaries_match(candidate, query);
+    m_only_word_boundary = m_word_boundary_match_count == query.length();
+    m_prefix = std::equal(query.begin(), query.end(), candidate.begin(), smartcase_eq);
 }
 
 bool RankedMatch::operator<(const RankedMatch& other) const
 {
-    if (m_match_rank == other.m_match_rank)
-       return std::lexicographical_compare(
-           m_candidate.begin(), m_candidate.end(),
-           other.m_candidate.begin(), other.m_candidate.end(),
-           [](char a, char b) {
-               const bool low_a = islower(a), low_b = islower(b);
-               return low_a == low_b ? a < b : low_a;
-           });
+    if (m_only_word_boundary or other.m_only_word_boundary)
+        return m_only_word_boundary and other.m_only_word_boundary ?
+            m_word_boundary_match_count > other.m_word_boundary_match_count
+          : m_only_word_boundary;
 
-    return m_match_rank < other.m_match_rank;
+    if (m_prefix != other.m_prefix)
+        return m_prefix;
+
+    if (m_word_boundary_match_count != other.m_word_boundary_match_count)
+        return m_word_boundary_match_count > other.m_word_boundary_match_count;
+
+    if (m_first_char_match != other.m_first_char_match)
+        return m_first_char_match;
+
+    return std::lexicographical_compare(
+        m_candidate.begin(), m_candidate.end(),
+        other.m_candidate.begin(), other.m_candidate.end(),
+        [](char a, char b) {
+            const bool low_a = islower(a), low_b = islower(b);
+            return low_a == low_b ? a < b : low_a;
+        });
 }
 
+UnitTest test_ranked_match{[] {
+    kak_assert(count_word_boundaries_match("run_all_tests", "rat") == 3);
+}};
+
 }
diff --git a/src/ranked_match.hh b/src/ranked_match.hh
index d6367453..2a805002 100644
--- a/src/ranked_match.hh
+++ b/src/ranked_match.hh
@@ -19,7 +19,10 @@ struct RankedMatch
 
 private:
     StringView m_candidate;
-    int m_match_rank = 0;
+    bool m_first_char_match = false;
+    bool m_prefix = false;
+    int m_word_boundary_match_count = 0;
+    bool m_only_word_boundary = false;
 };
 
 using RankedMatchList = Vector<RankedMatch>;