A Final Word on the Astros, Rangers and Baseball Clutchiness

This past season was a lot of things for Astros fans - the first word that comes to mind, without blame, is "frustrating". It was frustrating to watch the team muddle through the second half with a depleted lineup. Frustrating to watch the starting rotation implode while the offense scored runs, frustrating to watch the offense shut down while the starters dealt, and to cap it off, frustrating to watch the reliable, shut-down bullpen arms falter in must-win contests against Los Angeles and Seattle in the last week of the season.

Most of all, it was frustrating to watch the team find every which way to lose to the Rangers. But while it was that and more, from a high-level MLB perspective, it was also fascinating. The Rangers fared extremely well against teams over .500, and played poorly against teams under .500. Their bullpen wasn't 2012 Orioles-esque, and #NeverEverQuit isn't enough to explain away 36-11 in one-run games. Explanations of the Rangers; success were discussed at length by really good ball writers everywhere: Google "Rangers Run Differential" and you'll find posts from Fangraphs, Five Thirty Eight, and even the Wall Street Journal.

Pretty much every outlet supported the following claim: The Texas Rangers were a quality baseball team, but also needed a historic bit of luck to reach 95 wins. As the seasons of both AL West rivals have now reached a close, I will offer a final word (and Fangraphs data) that will hopefully put the entire idea of baseball "clutchiness" into a season-long perspective.

The origin of the graph below comes from a 2014 Fangraphs article by Jeff Sullivan. In it, he plots each MLB team's Win percentage and BaseRuns win percentage as a differential (y-axis) against each team's Clutch score (x-axis). These are defined below:

BaseRuns: Measure of a team's record based on underlying performances from raw stats (singles, doubles, home runs, etc.). The entire BaseRuns explanation in the link offers a more comprehensive look than that one-sentence definition.

Clutch Score: Measures a team's hitting and pitching performances in high leverage situations. This is essentially a measure of timeliness and will also be referred to as "sequencing".

I recreated this graph for 2016 - and here's why. The Astros and Rangers finished the season with the exact same BaseRuns record (82-80). So based on raw stats from both pitchers and hitters, they performed equally well across 162 games. So how does this discrepancy in actual record happen?

"Ah yes", you say. "Clutch, again". Here's what the above graph tells us: Clutch score (or sequencing) can mean something for season results; in extreme cases, it can mean a lot. Teams that finished the season with significant clutch scores were able to either add more wins over what BaseRuns projected (again, expected record based on raw stats), or drop more games than expected from BaseRuns due to better sequencing than other teams.

But does clutch really tell us anything about a team's talent? The answer is definitely no. Let's take our favorite out-performing team up I-45 (Texas) and Washington as a case study. Both teams won 95 games. The Nationals fell into the large cluster of teams near the middle in the matrix above - they registered a modestly negative clutch score around -3.0 and underperformed their BaseRuns record by three games. They were a quality, balanced team this year, but didn't receive an inordinate amount of "clutch" production on the offensive side. In fact, Washington recorded a .705 OPS and a wRC+ of just 84 in high leverage scenarios.

The Rangers, on the other hand, registered an .828 OPS and a wRC+ of 115 in high leverage situations. Though many other statistics can be cited to explain good baseball fortune, that discrepancy could be one of many small factors that tilted Texas toward the upper right corner of the matrix, and Washington toward the "not clutch, unlucky" cluster where many teams end up falling (because "clutch" team performances on their own should not be expected to cause a winning season). A thought experiment: Though the 2016 season has been spoken for, what would the Nationals' record look like with their deep, talented roster if they switched places on the matrix with Texas and simply hit as well in clutch situations?

A second and final graph below illustrates the relationship between the same Clutch score and actual winning percentage.

There's pretty much zero relationship between a team's clutch score and actual overall record in a given season. To reiterate: Good sequencing can drive team luck, and can show how us much that lucky really benefited a team in hindsight. The Rangers won a ton of games that were largely decided in high-leverage. But it's not likely that a team can depend on those high-leverage performances every night toward a good record. In this sense, the Rangers weren't as much of an outlier as they were an extreme case of hitting and pitching historically better in tight games relative to what they did in medium and low leverage situations.

As for the Astros? They weren't affected much at all due to luck this season. In the first graph, their clutch score was essentially a neutral zero (suggesting that no, they were not terrible in high leverage situations with RISP or late in games like many might believe). Instead, they were unable to create as much good luck from sequencing than their division rival.

That being said, better luck or clutch on the Astros' end coupled with "normal" luck from the Rangers probably wouldn't have swung the division the other way. As shown above, Houston actually outperformed BaseRuns by two wins, and even with normalized luck, the Rangers had a healthier, more productive roster down the stretch that was aided by aggressive deadline acquisitions. And amidst the midseason banter between this site and Ranger fans, Seattle actually ended up with the best case for the most "talented" division team, as they outperformed both Houston and Texas on run-scoring records.

So, looking at the relationship between sequencing and team records is a "fun" (or infuriating) exercise, but can only be done in hindsight - like right now. The only certainties with this graph in 2017 are most teams will end up clustered near the middle, without enough sequencing to drastically swing an actual record in either direction. Unless Jeff Banister sells his soul to the devil again.

(All stats via Fanngraphs)