Pipeline Aggregations with buckets_path more than two

Elastic Stack Elasticsearch
4

So if I'm reading the example right it is trying to do some kind of standard deviation where each day's total is expressed as a multiple of the average sales for all days.

To do this you'd need something like this:

DELETE test
POST test/sales
{
	"amount":10,
	"date":"2012-09-21"
}
POST test/sales
{
	"amount":20,
	"date":"2012-09-21"
}
POST test/sales
{
	"amount":5,
	"date":"2012-09-22"
}
POST test/sales
{
	"amount":15,
	"date":"2012-09-22"
}

GET test/sales/_search
{
   "size": 0,
   "aggs": {
	  "by_date": {
		 "date_histogram": {
			"field": "date",
			"interval": "day"
		 },
		 "aggs": {
			"Total": {
			   "sum": {
				  "field": "amount"
			   }
			},
			"normalized_sales": {
			   "bucket_script": {
				  "buckets_path": {
					 "dailySalesTotal": "Total",
					 //!!!! THIS DOES NOT WORK.... !!!!!
					 "allDaysAverage": "../global_stats>avg"
				  },
				  "script": "params.dailySalesTotal / params.allDaysAverage"
			   }
			}
		 }
	  },
	  "global_stats": {
		 "stats_bucket": {
			"buckets_path": "by_date>Total"
		 }
	  }
   }
}

But notice this won't work because inside the array of by_date results we can't refer to a parent-level stat (attempted here using the illegal .. syntax). The docs state:

Paths are relative from the position of the pipeline aggregation; they are not absolute paths, and the path cannot go back "up" the aggregation tree