diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate2/aggregates.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate2/aggregates.scala
index b18fe40855673..ec39ec68e72ff 100644
--- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate2/aggregates.scala
+++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate2/aggregates.scala
@@ -96,6 +96,8 @@ abstract class AlgebraicAggregate extends AggregateFunction2 with Serializable{
   /** Must be filled in by the executors */
   var inputSchema: Seq[Attribute] = _
 
+  def offsetExpressions: Seq[Attribute] = Seq.fill(bufferOffset)(AttributeReference("offset", NullType)())
+
   lazy val rightBufferSchema = bufferSchema.map(_.newInstance())
   implicit class RichAttribute(a: AttributeReference) {
     def left = a
@@ -112,8 +114,11 @@ abstract class AlgebraicAggregate extends AggregateFunction2 with Serializable{
   }
 
   lazy val boundUpdateExpressions = {
-    val updateSchema = inputSchema ++ bufferSchema
-    updateExpressions.map(BindReferences.bindReference(_, updateSchema)).toArray
+    val updateSchema = inputSchema ++ offsetExpressions ++ bufferSchema
+    val bound = updateExpressions.map(BindReferences.bindReference(_, updateSchema)).toArray
+    println(s"update: ${updateExpressions.mkString(",")}")
+    println(s"update: ${bound.mkString(",")}")
+    bound
   }
 
   val joinedRow = new JoinedRow
@@ -126,20 +131,27 @@ abstract class AlgebraicAggregate extends AggregateFunction2 with Serializable{
   }
 
   lazy val boundMergeExpressions = {
-    val mergeSchema = bufferSchema ++ rightBufferSchema
+    val mergeSchema = offsetExpressions ++ bufferSchema ++ offsetExpressions ++ rightBufferSchema
     mergeExpressions.map(BindReferences.bindReference(_, mergeSchema)).toArray
   }
   override def merge(buffer1: MutableRow, buffer2: InternalRow): Unit = {
     var i = 0
+    println(s"Merging: $buffer1 $buffer2 with ${boundMergeExpressions.mkString(",")}")
+    joinedRow(buffer1, buffer2)
     while (i < bufferSchema.size) {
-      buffer1(i + bufferOffset) = boundMergeExpressions(i).eval(joinedRow(buffer1, buffer2))
+      println(s"$i + $bufferOffset: ${boundMergeExpressions(i).eval(joinedRow)}")
+      buffer1(i + bufferOffset) = boundMergeExpressions(i).eval(joinedRow)
       i += 1
     }
   }
 
-  lazy val boundEvaluateExpression = BindReferences.bindReference(evaluateExpression, bufferSchema)
+  lazy val boundEvaluateExpression =
+    BindReferences.bindReference(evaluateExpression, offsetExpressions ++ bufferSchema)
   override def eval(buffer: InternalRow): Any = {
-    boundEvaluateExpression.eval(buffer)
+    println(s"eval: $buffer")
+    val res = boundEvaluateExpression.eval(buffer)
+    println(s"eval: $buffer with $boundEvaluateExpression => $res")
+    res
   }
 }
 
@@ -171,7 +183,7 @@ case class Average(child: Expression) extends AlgebraicAggregate {
       Add(
         currentSum,
         Coalesce(Cast(child, intermediateType) :: Cast(Literal(0), intermediateType) :: Nil)),
-    /* currentCount = */ If(IsNotNull(child), currentCount, currentCount + 1L)
+    /* currentCount = */ If(IsNull(child), currentCount, currentCount + 1L)
   )
 
   val mergeExpressions = Seq(
@@ -179,7 +191,7 @@ case class Average(child: Expression) extends AlgebraicAggregate {
     /* currentCount = */ currentCount.left + currentCount.right
   )
 
-  val evaluateExpression = Cast(currentCount, resultType) / Cast(currentSum, resultType)
+  val evaluateExpression = Cast(currentSum, resultType) / Cast(currentCount, resultType)
 
   override def nullable: Boolean = false
   override def dataType: DataType = resultType
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/execution/Aggregate2Suite.scala b/sql/core/src/test/scala/org/apache/spark/sql/execution/Aggregate2Suite.scala
index a82cf9d7b47bd..b65872dc898d2 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/execution/Aggregate2Suite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/execution/Aggregate2Suite.scala
@@ -61,12 +61,15 @@ class Aggregate2Suite extends QueryTest with BeforeAndAfterAll {
         |GROUP BY key
       """.stripMargin).queryExecution.executedPlan(3).execute().collect().foreach(println)
 
-    ctx.sql(
-      """
-        |SELECT key, avg2(value)
-        |FROM agg2
-        |GROUP BY key
-      """.stripMargin).show()
+    checkAnswer(
+      ctx.sql(
+        """
+          |SELECT key, avg2(value)
+          |FROM agg2
+          |GROUP BY key
+        """.stripMargin),
+      Row(1, 20.0) :: Row(2, -0.5) :: Row(3, null) :: Nil)
+
   }
 
   override def afterAll(): Unit = {