Washington Nationals turn to fifth best reliever in ninth again

For the second time in three days, the Washington Nationals entered the ninth inning trailing by one run. Just as they did in Sunday's loss to the Dodgers, the Nats had the top of the order due up in the bottom of the ninth inning. Bryce Harper, who has been the best hitter on the planet this season, was guaranteed an at bat in the ninth inning of each of those games. Unfortunately, just like he did on Sunday, Matt Williams handed the ball to the wrong man in the top of the inning.

The parallels between the Sunday and Tuesday games are certainly there for all to see. The names of the pitchers that Williams called upon in the ninth inning changed, but they turned close games into blowouts. The Nats failed to score in the bottom of the ninth in each game, but they did put two runners on base each time. If we were to use results oriented thinking, Williams' decision in the ninth inning meant nothing. A 1-0 loss and a 5-0 loss both count as a loss in the standings. The same goes for a 3-2 loss and a 7-2 loss.

Of course, using results oriented thinking generally isn't the best course of action. For starters, previous results don't necessarily have all that much predictive value of what's going to happen the next time that situation occurs. Secondly, people don't always tend to react the same under two unique sets of circumstances.

A pitcher throwing with a one run lead in the bottom of the ninth would seem more likely to tighten up, particularly once a runner (or two) gets on base. A pitcher with a five run lead in the bottom of the ninth would be expected to be more relaxed and confident in attacking the next batter.

A similar thing could be said about the opposing manager's mindset. On Tuesday, the Mets went with closer Jeurys Familia, who had been warming up anyway with a one run lead. On Sunday, the Dodgers reacted differently. Rather than using closer Kenley Jansen, who (results oriented thinking) had just allowed a two run homer to Bryce Harper on Saturday, they turned to lefty J.P. Howell instead.

Finally, the Nats could have approached the situation differently themselves. Had leadoff man Yunel Escobar still led off the ninth with a single, we could have seen two comically brilliant "managin'" moves behind it in a one run game. Matt Williams would almost surely have had Danny Espinosa bunt (he grounded into a double play), which probably would have led to Don Mattingly intentionally walking Bryce Harper to put the winning run on base. We were cheated of chaos!

Really, though. When a team is trailing by one run heading into the ninth, there's a much different feel to the game than when they're trailing by five. The crowd still has the energy to stay in the game. The dugouts seem more alive. The tension is much higher. We can't say, based on the fact that the Nats didn't score in the ninth inning of Sunday's game or Tuesday's game, that the same thing would have happened if they had still been one run games.

At the same time, we can't say with 100% accuracy that if Matt Williams had turned to Drew Storen instead of Blake Treinen or Tanner Roark, that Storen would have kept the score where it was heading into the bottom half. We don't know what would have happened if Williams had gone to Storen because he didn't go to Storen. What we can say about those three pitchers is that based on their production and performance this season, Storen was far more likely to keep the score where it was than Roark and Treinen were.

*These are the three pitchers' numbers solely as relievers. Roark's numbers don't include Tuesday night's outing yet. After Tuesday night's meltdown, Roark's reliever ERA (4.24) is about to match his FIP. His overall ERA (4.97) and FIP (4.83) have been worse than his performance in the bullpen.

What's gained by keeping the score at 1-0 or 3-2?

According to Win Probability Added, the Nationals improve their odds of winning a ballgame by nearly 18%. Using this Win Expectancy Calculator based on data from 1957-2013, home teams trailing by one run entering the bottom of the ninth inning have an 18.5% chance of winning. Teams trailing by five runs entering the bottom of the ninth have a 0.8% chance of winning. Even though we're already dealing with a situation where the Nats are more than four times more likely to lose than win, that's a pretty significant shift. Of course, win probability doesn't take into account who is due up either, so the fact that Bryce Harper is guaranteed an at bat (unless the first two hitters produce two runs) doesn't give them any extra credit by the win probability model.

In sharp contrast, we can look at the game in between Sunday and Tuesday, in which Matt Williams did use Drew Storen. Storen, the Nationals best reliever, entered the game to pitch the top of the ninth inning on Monday with the Nats leading 7-2. The Nationals win probability at the start of the ninth inning in Monday's game was 99.4%. Even if Ian Desmond hadn't homered in the bottom of the eighth to pad the lead, the Nationals win probability with a three run lead would have been 97.3%.

We've seen Williams handle Storen this way all season long. He's pitched in 37 games so far this season. He's only entered a game three times when the Nats didn't already have the lead. One of those situations was on April 26, when the Nats were trailing by four in Miami and Storen only came in because he hadn't pitched in five days. The other two were in the ninth inning of tie games at home. Williams showed in those two tie games at home that he realizes that there's no (ridiculous) save situation to be had, so he went with his best reliever. Sometimes it makes sense to use your best reliever to ensure yourself a chance even when you're trailing by a run. Williams doesn't seem to think so.

The decision to (specifically) go to Roark

The Tanner Roark experiment hasn't worked particularly well for much of the year. It's unclear how much bouncing back and forth between the rotation and the bullpen has affected him, but he doesn't seem to have thrived in either role this season. This has been particularly true lately. Roark was lit up in his most recent attempt at starting, allowing 8 runs in 3.1 innings against the Phillies. Since returning to the bullpen, he had allowed at least a run in each of his three outings prior to Tuesday's game. This begs the question: If you're not going to go to your closer to keep it a one run game in the ninth inning at home, why turn to a reliever who has been a gas can lately?

With Aaron Barrett and Felipe Rivero having already been used, the bullpen may have been stretched a little thinner than Williams wanted. Casey Janssen had pitched on back to back days and the Mets had two righties surrounding a pinch-hitter due up, so maybe Roark made some sense if Williams' mindset was that he planned to absolutely refuse to use Storen. His only other (non-Storen) right-handed option in the bullpen was Abel de los Sanots, who eventually mopped up Roark's mess in his big league debut.

Of course, if there were to be a prolonged extra inning game (it was just a one run game), Roark would have been the guy the top choice to throw an extended outing though, right? Sammy Solis could probably throw a couple of innings, but part of the reason that Roark is the long man in the bullpen is because he's had success as a big league starter. In this scenario, Williams not only failed to extend the game (the goal), but he also spent his best option to work multiple innings if it turned into a marathon extra inning game.

Williams' other two big decisions

Williams' other two big decisions actually came in the same at bat, and I'm not going to criticize him for either of them. With runners on second and third and one out in the seventh inning, Williams turned to Aaron Barrett to replace Joe Ross. Barrett was brought on to protect a one run lead against pinch hitter Eric Campbell.

Removing Storen from the equation, Barrett seems like the best fit for this situation. The Mets were in a situation where contact likely scores the tying run. Barrett (12.0 K/9) is the best strikeout pitcher on the entire staff and had the platoon advantage over Campbell. He "got beat" on a slider down in the zone on the outer half that got poked over the drawn-in infield. Had the infield been playing at normal depth, it was a play that Danny Espinosa likely would have made for the second out... and the runners wouldn't have advanced. This leads us to the second decision, which was to play the infield in. I can see both sides of the argument.

On the one hand, with runners on second and third and one out, you want to limit the damage. Given that the Mets have scored the second least runs in all of baseball, you could concede the run on a ground ball and feel confident that you can outscore them the rest of the way. It's a conservative approach, but (assuming the at bat/next batter play out the same way) it would have actually allowed the Nats to maintain the 2-1 lead they had at the time.

On the flip side, it's the seventh inning. Wilmer Flores represents the tying run at third base, and he isn't exactly known for his speed. Eric Campbell is batting .179 and doesn't exactly inspire a ton of fear at the plate. It's late enough in the game so that playing the infield in against a mediocre hitter to try and keep the tying run from scoring makes sense.

In the case of playing the infield in, I can see both sides of the argument. Williams made a call. It backfired. Such is life. Going to your fifth or sixth best reliever in the ninth inning trailing by a run at home against your chief competition when you have the top of the order coming up in the bottom half? That's just asinine.