25 July 2017

Games Back: It's Kind of Complicated

Yesterday, I asked the site's followers for their opinion of when the probabilities are high enough to reach out and start buying at the trade deadline.  Overwhelmingly, they choose 50% with the caveat that some had others ideas like 40%, but 50% was closest to their perspective.  In turn, I wanted to compare those preferences to the Orioles' situation and how we think about playoff likelihood in a more traditional way.

Over the past few weeks, feverish attention has been placed on how many Games Back the Orioles are of either the AL East crown or, more often, the second Wild Card slot.  This attention often appears to coincide with a demand that the club is not out of the hunt and should not scatter players into the transaction wire wind.  However, Games Back as simple as it sounds is actually a rather difficult metric to comprehend.  If you think of each club as a particle moving through space, the distance between and the speed of both your particle and the particle you wish to overtake is needed, but also every other particle in the mix.  Just focusing on you and the leader gives you an incomplete view.

But what exactly is that incomplete view?  In this column, we will ignore probability metrics.  Those are probably the best way to figure out whether or not the club you root for is alive.  And, well, if you have resisted looking at them for perceived faults in their talent assumptions, then, well, I doubt this column will change your mind.  Instead, I will show you a different metric,  This one is called summed Games Back.

Summed Games Back is literally what it claims to be.  You simply add the number of Games Back the club is to the number of Games Back every club ahead of them is.  For instance, here are the standings from July 21, 2017:

AL Wild Card Detailed Standings
Tm W L GB GBsum
Provided by Baseball-Reference.com: View Original Table
Generated 7/21/2017.

You will notice that the Orioles are shown as 8.5 GBsum vs. 3.5 GB.  That 8.5 summed GB number is the sum comes from adding the Orioles (3.5), Mariners (2.5), Royals (2.0), and Twins (1.0) Games Back values.  It is certainly a number that can fluctuate wildly.  For instance, if the Angels were a half game better, the Orioles would go from 8.5 to 11.5.  That inability to account for equal records certainly is a drawback (one that playoff probability metrics account for), but it is also a weak point in traditional Games Back approach.

So, let us compare GB vs. GBsum.  For this, I decided to look at all Wild Card races from 2012 until 2016 on July 25th.  I only looked at clubs who were five games or fewer Games Back.  I compared that placement with where they were at the end of the year.  That way we can decide what the probability of a placement is in getting a playoff entry.  I will do the same for those clubs, but use summed Games Back (regardless of that value for those teams that are 5 GB).  If GBsum is a better metric, then we will find that it will more accurately identify competitive teams.

When we batch numbers for Games Back from the second wild card from 2012-2016, something sticks out:

GB % PO GBsum % PO
0 50 0 50
1 and 2 21 1 to 3 21
3 and 4 9 4 to 9 13
5 11 10 to 14 0

Games back appears to be relatively useful in looking at how competitive a club might be until the data range between 3 and 5 Games Back.  At that point, it appears the probability sits around 10% without budging much while the 1 and 2 Games Back bucket sits comfortably twice as high in probability.  For the summed Games Back, there is a pretty consistent decrease in playoff probability.  Chances evaporate steadily through each bucket until you reach the bottom.

Where are the Orioles now, as I write this, they are 3.5 Games Back (suggesting ~10% likelihood) while they are 12.5 summed Games Back (suggesting 0% likelihood).  The larger take home would be this: 3.5 Games Back is a tough struggle to climb and a lot of data can be hidden in that number as you do not know how well other teams are doing.  Summed Games Back gives a better indication of who else is at play.  It suggests that the Orioles are in a situation where things are impossible.  If the Orioles do come back from 12.5 GBsum, it will be something no club has been able to do in the 2012 to 2016 window.

When it comes to that Twitter poll question, it seems fairly clear that maybe only about 1 in 20 Orioles fans think it is prudent for the club to buy and go for that Wild Card slot.  Most appear to think a club should only buy if it is sitting in the second wild card slot at the end of July.  A few more consider it practical to buy as long as a club is no more than two games back.  Then again, it may well be that most people and their perceived tolerance for probabilities are not in line with actual probabilities as they relate to Games Back.

Update: The Orioles won last night after I scheduled this for publication.  They currently are still 3.5 games out, but are now 9 summed Games Back.  This would put them squarely in the ~10% bracket.


Unknown said...

You shifted from saying we were 8.5 GBsum to 12.5 GBsum. According to your model, that makes the difference between us having a 10% to get in, or a zero chance to get in. Maybe if your 87 and you don't know how many more seasons you have left, you buy at 10% and hope the buy gets to the magic number.

Jon Shepherd said...

The table is from July 21st as noted in the article. The data for the model comes from standings on July 25th as noted in the article.

Boss61 said...

I like the method, when compared to traditional games back.

Chasing a bunch of teams, some of whom play one another, can be a maddeningly slow climb. Summed games back reflects this.