diff --git a/pgx/_src/dwg/tictactoe.py b/pgx/_src/dwg/tictactoe.py
index b3cc46fef..b8cc66748 100644
--- a/pgx/_src/dwg/tictactoe.py
+++ b/pgx/_src/dwg/tictactoe.py
@@ -77,7 +77,7 @@ def _make_tictactoe_dwg(dwg, state: TictactoeState, config):
             )
         )
 
-    for i, mark in enumerate(state._board):
+    for i, mark in enumerate(state._x._board):
         x = i % BOARD_WIDTH
         y = i // BOARD_HEIGHT
         if mark == 0:  # 先手
diff --git a/pgx/tic_tac_toe.py b/pgx/tic_tac_toe.py
index 08897dd4d..135cc331a 100644
--- a/pgx/tic_tac_toe.py
+++ b/pgx/tic_tac_toe.py
@@ -23,6 +23,15 @@
 TRUE = jnp.bool_(True)
 
 
+@dataclass
+class GameState:
+    _turn: Array = jnp.int32(0)
+    # 0 1 2
+    # 3 4 5
+    # 6 7 8
+    _board: Array = -jnp.ones(9, jnp.int32)  # -1 (empty), 0, 1
+
+
 @dataclass
 class State(core.State):
     current_player: Array = jnp.int32(0)
@@ -32,12 +41,7 @@ class State(core.State):
     truncated: Array = FALSE
     legal_action_mask: Array = jnp.ones(9, dtype=jnp.bool_)
     _step_count: Array = jnp.int32(0)
-    # --- Tic-tac-toe specific ---
-    _turn: Array = jnp.int32(0)
-    # 0 1 2
-    # 3 4 5
-    # 6 7 8
-    _board: Array = -jnp.ones(9, jnp.int32)  # -1 (empty), 0, 1
+    _x: GameState = GameState()
 
     @property
     def env_id(self) -> core.EnvId:
@@ -79,8 +83,8 @@ def _init(rng: PRNGKey) -> State:
 
 
 def _step(state: State, action: Array) -> State:
-    state = state.replace(_board=state._board.at[action].set(state._turn))  # type: ignore
-    won = _win_check(state._board, state._turn)
+    state = state.replace(_x=state._x.replace(_board=state._x._board.at[action].set(state._x._turn)))  # type: ignore
+    won = _win_check(state._x._board, state._x._turn)
     reward = jax.lax.cond(
         won,
         lambda: jnp.float32([-1, -1]).at[state.current_player].set(1),
@@ -88,10 +92,10 @@ def _step(state: State, action: Array) -> State:
     )
     return state.replace(  # type: ignore
         current_player=(state.current_player + 1) % 2,
-        legal_action_mask=state._board < 0,
+        legal_action_mask=state._x._board < 0,
         rewards=reward,
-        terminated=won | jnp.all(state._board != -1),
-        _turn=(state._turn + 1) % 2,
+        terminated=won | jnp.all(state._x._board != -1),
+        _x=state._x.replace(_turn=(state._x._turn + 1) % 2),  # type: ignore
     )
 
 
@@ -103,13 +107,13 @@ def _win_check(board, turn) -> Array:
 def _observe(state: State, player_id: Array) -> Array:
     @jax.vmap
     def plane(i):
-        return (state._board == i).reshape((3, 3))
+        return (state._x._board == i).reshape((3, 3))
 
     # flip if player_id is opposite
     x = jax.lax.cond(
         state.current_player == player_id,
-        lambda: jnp.int32([state._turn, 1 - state._turn]),
-        lambda: jnp.int32([1 - state._turn, state._turn]),
+        lambda: jnp.int32([state._x._turn, 1 - state._x._turn]),
+        lambda: jnp.int32([1 - state._x._turn, state._x._turn]),
     )
 
     return jnp.stack(plane(x), -1)
diff --git a/tests/test_tic_tac_toe.py b/tests/test_tic_tac_toe.py
index d60cad8ba..c9fd63287 100644
--- a/tests/test_tic_tac_toe.py
+++ b/tests/test_tic_tac_toe.py
@@ -19,13 +19,13 @@ def test_step():
     key = jax.random.PRNGKey(1)
     state = init(key=key)
     assert state.current_player == 1
-    assert state._turn == 0
+    assert state._x._turn == 0
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([1, 1, 1, 1, 1, 1, 1, 1, 1], jnp.bool_)
     )  # fmt: ignore
     assert jnp.all(
-        state._board == jnp.int32([-1, -1, -1, -1, -1, -1, -1, -1, -1])
+        state._x._board == jnp.int32([-1, -1, -1, -1, -1, -1, -1, -1, -1])
     )
     assert not state.terminated
     # -1 -1 -1
@@ -35,13 +35,13 @@ def test_step():
     action = jnp.int32(4)
     state = step(state, action)
     assert state.current_player == 0
-    assert state._turn == 1
+    assert state._x._turn == 1
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([1, 1, 1, 1, 0, 1, 1, 1, 1], jnp.bool_)
     )  # fmt: ignore
     assert jnp.all(
-        state._board == jnp.int32([-1, -1, -1, -1, 0, -1, -1, -1, -1])
+        state._x._board == jnp.int32([-1, -1, -1, -1, 0, -1, -1, -1, -1])
     )
     assert jnp.all(state.rewards == 0)  # fmt: ignore
     assert not state.terminated
@@ -52,12 +52,12 @@ def test_step():
     action = jnp.int32(0)
     state = step(state, action)
     assert state.current_player == 1
-    assert state._turn == 0
+    assert state._x._turn == 0
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([0, 1, 1, 1, 0, 1, 1, 1, 1], jnp.bool_)
     )  # fmt: ignore
-    assert jnp.all(state._board == jnp.int32([1, -1, -1, -1, 0, -1, -1, -1, -1]))
+    assert jnp.all(state._x._board == jnp.int32([1, -1, -1, -1, 0, -1, -1, -1, -1]))
     assert jnp.all(state.rewards == 0)  # fmt: ignore
     assert not state.terminated
     #  1 -1 -1
@@ -67,12 +67,12 @@ def test_step():
     action = jnp.int32(1)
     state = step(state, action)
     assert state.current_player == 0
-    assert state._turn == 1
+    assert state._x._turn == 1
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([0, 0, 1, 1, 0, 1, 1, 1, 1], jnp.bool_)
     )  # fmt: ignore
-    assert jnp.all(state._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, -1, -1]))
+    assert jnp.all(state._x._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, -1, -1]))
     assert jnp.all(state.rewards == 0)  # fmt: ignore
     assert not state.terminated
     #  1  0 -1
@@ -82,12 +82,12 @@ def test_step():
     action = jnp.int32(8)
     state = step(state, action)
     assert state.current_player == 1
-    assert state._turn == 0
+    assert state._x._turn == 0
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([0, 0, 1, 1, 0, 1, 1, 1, 0], jnp.bool_)
     )  # fmt: ignore
-    assert jnp.all(state._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, -1, 1]))
+    assert jnp.all(state._x._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, -1, 1]))
     assert jnp.all(state.rewards == 0)  # fmt: ignore
     assert not state.terminated
     #  1  0 -1
@@ -97,12 +97,12 @@ def test_step():
     action = jnp.int32(7)
     state = step(state, action)
     assert state.current_player == 0
-    assert state._turn == 1
+    assert state._x._turn == 1
     assert jnp.all(
         state.legal_action_mask
         == jnp.array([1, 1, 1, 1, 1, 1, 1, 1, 1], jnp.bool_)
     )  # fmt: ignore
-    assert jnp.all(state._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, 0, 1]))
+    assert jnp.all(state._x._board == jnp.int32([1, 0, -1, -1, 0, -1, -1, 0, 1]))
     assert jnp.all(state.rewards == jnp.int32([-1, 1]))  # fmt: ignore
     assert state.terminated
     #  1  0 -1