diff --git a/libecole/CMakeLists.txt b/libecole/CMakeLists.txt
index 55ea583e..d92962ce 100644
--- a/libecole/CMakeLists.txt
+++ b/libecole/CMakeLists.txt
@@ -28,6 +28,7 @@ add_library(
 	src/reward/lp-iterations.cpp
 	src/reward/solving-time.cpp
 	src/reward/n-nodes.cpp
+	src/reward/tree-size-estimate.cpp
 	src/reward/bound-integral.cpp
 
 	src/observation/node-bipartite.cpp
diff --git a/libecole/include/ecole/reward/tree-size-estimate.hpp b/libecole/include/ecole/reward/tree-size-estimate.hpp
new file mode 100644
index 00000000..3b108d8e
--- /dev/null
+++ b/libecole/include/ecole/reward/tree-size-estimate.hpp
@@ -0,0 +1,23 @@
+#pragma once
+
+#include <cstdint>
+
+#include "ecole/export.hpp"
+#include "ecole/reward/abstract.hpp"
+#include "scip/event_estim.h"
+#include "scip/scip_event.h"
+
+#define EVENTHDLR_NAME "estim"
+
+namespace ecole::reward {
+
+class ECOLE_EXPORT TreeSizeEstimate {
+public:
+	ECOLE_EXPORT auto before_reset(scip::Model& model) -> void;
+	ECOLE_EXPORT auto extract(scip::Model& model, bool done = false) -> Reward;
+
+private:
+	SCIP_Real tree_size_estimate = 0.0;
+};
+
+}  // namespace ecole::reward
diff --git a/libecole/src/reward/tree-size-estimate.cpp b/libecole/src/reward/tree-size-estimate.cpp
new file mode 100644
index 00000000..6e87e937
--- /dev/null
+++ b/libecole/src/reward/tree-size-estimate.cpp
@@ -0,0 +1,16 @@
+#include "ecole/reward/tree-size-estimate.hpp"
+
+#include "ecole/scip/model.hpp"
+#include "scip/def.h"
+
+namespace ecole::reward {
+
+void TreeSizeEstimate::before_reset(scip::Model& /* model */) {}
+
+Reward TreeSizeEstimate::extract(scip::Model& model, bool /* done */) {
+	// getTreeSizeEstimation returns -1 when no estimation has been made yet.
+	tree_size_estimate = SCIPgetTreesizeEstimation(model.get_scip_ptr());
+	return tree_size_estimate;
+}
+
+}  // namespace ecole::reward
diff --git a/python/ecole/src/ecole/core/reward.cpp b/python/ecole/src/ecole/core/reward.cpp
index 984c7b34..0173dcf6 100644
--- a/python/ecole/src/ecole/core/reward.cpp
+++ b/python/ecole/src/ecole/core/reward.cpp
@@ -11,6 +11,7 @@
 #include "ecole/reward/lp-iterations.hpp"
 #include "ecole/reward/n-nodes.hpp"
 #include "ecole/reward/solving-time.hpp"
+#include "ecole/reward/tree-size-estimate.hpp"
 #include "ecole/scip/model.hpp"
 
 #include "core.hpp"
@@ -147,6 +148,20 @@ void bind_submodule(py::module_ const& m) {
 		The difference in number of nodes is computed in between calls.
 		)");
 
+	auto treesizeestimate = py::class_<TreeSizeEstimate>(m, "TreeSizeEstimate", R"(
+		Estimate the size of a tree.
+
+		The reward is defined as the total number of nodes processed since the previous state.
+	)");
+	treesizeestimate.def(py::init<>());
+	def_operators(treesizeestimate);
+	def_before_reset(treesizeestimate, "Reset the internal node count.");
+	def_extract(treesizeestimate, R"(
+		Update the internal node count and return the difference.
+
+		The difference in number of nodes is computed in between calls.
+		)");
+
 	auto solvingtime = py::class_<SolvingTime>(m, "SolvingTime", R"(
 		Solving time difference.