When to play the statistical long game

The game show "Let's Make a Deal" debuted in 1963 and was hosted for 30 years by Monty Hall. The game show was known for the crazy costumes the contestants wore and is famous in the statistics world for the "Big Deal" segment. In the Big Deal segment of the show, contestants are shown 3 numbered doors and are told about the fabulous/expensive prize behind 1 of the three doors. One of the other doors has an ok prize, and the third door typically would have a lousy prize, like a goat (the farm animal kind, not the Tom Brady kind).

The contestant chooses one door, let us say they choose door #1. At that point one of the two other doors is exposed, typically the one with the "ok" prize, let us say door #2. So now the contestant has chosen door #1, door #2 has been exposed as the "ok" prize, and now the Big Deal prize is behind either door #1 or door #3. The contestant is now offered the choice: "do you want to stay with door #1, or would you like to change your choice to door #3?" In statistics, this has become known as the "Monty Hall" problem, and to statisticians, the choice is clear, you should ALWAYS switch to door #3. Should you?

Well, yes, if you are playing the long game. Or no, maybe, if you're playing the short game.

Michael Lewis published the book "Moneyball" in 2003 about the 2002 Oakland A's baseball season. "Moneyball" was made into a movie in 2011, and Brad Pitt starred as the Oakland A's General Manager Billy Beane. Moneyball details how the A's, with their low payroll, used data and analytics instead of the "institutional knowledge" type of decision-making that was typical at the time to put together a baseball team that was amazingly effective across a 162-game baseball season. Across a 162-game season, analytical projections will work themselves out across players' good days, bad days, and all.

The ultimate prize in baseball is not the regular season – it's winning in the playoffs. In that "magical" 2002 season the A's lost in the first round of the playoffs to the Minnesota Twins, a team in which the A's were 9 games better than in the regular season. In the "short game" of the playoffs, the sample size was not big enough to help the statistically better A's get past the Twins.

What does this have to do with sewer pipes? There are nearly 1 million miles of gravity sewer main pipes in the US. They all have the potential to become blocked and back up, creating an SSO. The maintenance to avoid SSO's is to either repair a broken or flawed pipe or to clean the pipe and remove any roots, FOG (fats, oils, grease) or debris. Is this a "long game" or a "short game" challenge?

The easiest way to look at the Monty Hall problem is: if you are offered a choice of one door, and immediately after making that choice, you're offered the two OTHER doors; statistically, you'd realize that two doors give you a 2/3 chance and one door a 1/3 chance. The exposing of one door in the middle doesn't change those statistical odds. If you stay with your original choice, you have a 1/3 chance of winning; if you switch, you have a 2/3 chance (it's not 50/50!). If you were allowed to play this game 10 or 100 times, you would win the most by always changing your choice when it's offered. That is the "long game" approach.

But in an emotional game show, the contestant might have a hunch, or maybe noticed a slight movement behind one of the doors, or perhaps has some "institutional knowledge" about which door tends to be the big prize on that day of the week, etc. In that sense, they might view it more like the baseball playoffs where in that one opportunity; they leverage institutional knowledge to stay with their original choice – taking a short game approach.

The U.S. EPA estimates that there are 23-75k SSO events per year across all collection systems. That's a lot of SSOs, but statistically, those occur on a very small % of all gravity main segments (less than 5% of mains typically). Is gravity main maintenance a short game or a long game challenge?

SSO events on gravity mains present a good short game challenge. Mains that have had SSOs in the past are the most likely candidates to have SSO events in the future unless they are put on a Hot Spot cleaning list and cleaned frequently. The institutional knowledge from those past SSOs gives a good reason for frequent cleaning.

But for the other 95+% of the gravity mains? That's where playing the analytics long game can bring significant benefit. Across many miles of gravity mains and many months/years of maintenance and cleaning, the most efficient approach is to use analytics to help prioritize which mains to clean first and how often. Of course, there are logistics to factor in ground crew and equipment management, but adding a data-driven opinion to the process will become the norm, just like all baseball teams have now adopted some version of the Moneyball tactics for their regular season planning.