diff --git a/Userland/Services/ChessEngine/MCTSTree.cpp b/Userland/Services/ChessEngine/MCTSTree.cpp
index a06fa62d41..a1a65af4d0 100644
--- a/Userland/Services/ChessEngine/MCTSTree.cpp
+++ b/Userland/Services/ChessEngine/MCTSTree.cpp
@@ -98,7 +98,15 @@ void MCTSTree::apply_result(int game_score)
 
 void MCTSTree::do_round()
 {
-    auto& node = select_leaf().expand();
+
+    // Note: Limit expansion to spare some memory
+    //       Efficient Selectivity and Backup Operators in Monte-Carlo Tree Search.
+    //       Rémi Coulom.
+    auto* node_ptr = &select_leaf();
+    if (node_ptr->m_simulations > s_number_of_visit_parameter)
+        node_ptr = &select_leaf().expand();
+
+    auto& node = *node_ptr;
 
     int result;
     if constexpr (s_eval_method == EvalMethod::Simulation) {
diff --git a/Userland/Services/ChessEngine/MCTSTree.h b/Userland/Services/ChessEngine/MCTSTree.h
index 8c7c0dd776..15ce2943c8 100644
--- a/Userland/Services/ChessEngine/MCTSTree.h
+++ b/Userland/Services/ChessEngine/MCTSTree.h
@@ -37,6 +37,7 @@ private:
     // While static parameters are less configurable, they don't take up any
     // memory in the tree, which I believe to be a worthy tradeoff.
     static constexpr double s_exploration_parameter { M_SQRT2 };
+    static constexpr int s_number_of_visit_parameter { 1 };
     // FIXME: Optimize simulations enough for use.
     static constexpr EvalMethod s_eval_method { EvalMethod::Heuristic };