A successful stolen base back on May 18 ignited a minor controversy recently in Twins territory. Holding a four run lead in the top of the 7th, the Detroit Tigers were playing it safe against the left handed batter, Joe Mauer. There were two outs, and the Tigers infielders were favoring the right side of the diamond, where any ground ball off Mauer's bat was likely to be hit. Tigers third baseman Nick Castellanos was playing well off third. Twins outfielder Eddie Rosario, who was the runner on second, saw an opportunity.

Here's how Twins announcer Dick Bremer described what happened next:

"And now running to third, with Rosario doing what Trevor Plouffe did the other day, taking advantage of Castellanos playing off the bag, and Rosario ends up at third base anyway with a stolen base before the pitch was thrown."

Justin Verlander, the Detroit pitcher, was still in the set position when he noticed Rosario breaking for third base. Verlander quickly stepped off and threw accurately to Castellanos (who himself was running to third to take the throw), but the speedy Rosario, sliding head first into the bag, easily beat them both.

Yet Rosario's stolen base was Twins manager Paul Molitor's last straw. After the inning ended, reserve outfielder Darin Mastroianni trotted out to left field instead of Rosario. Molitor had taken Rosario out of the game.

"The risk — 100 fold — is greater than the reward," Molitor told reporters afterwards. "Being safe doesn't make it right for me." Interestingly, two days earlier, Trevor Plouffe had attempted almost the same steal of third base under similar conditions — in a middle inning, down several runs. According to Bremer's call of the earlier play, Plouffe should have been out by five feet had the throw from the catcher not skipped wildly into left field. But baseball is a funny game. Plouffe was not only credited with the stolen base, but scored on the error, and perhaps sparked a rally, as the Twins would score three more runs in the inning.

How much risk did Rosario's steal entail? Were the Plouffe and Rosario situations all that different? Perhaps the manager was thinking primarily of that old baseball rule of thumb: Never make the final out at third. Here are three more ways to evaluate those steal attempts of third base: Run Expectancy, Win Probability, and Leverage.

Run Expectancy

The statistic RE24 shows the number of runs a team can expect to score in an inning, given the number of outs, and placement of base runners. A more in depth explanation of this metric can be found here.

As Plouffe stood on second base with one out in the May 16 game, the Twins could expect to score 0.644 additional runs before that inning ended. In Rosario's situation two days later, with two outs and a trailing runner on first, 0.343 more runs could be expected.

Obviously, no one scores a decimal or fraction of a run. This is generalized data that gives approximates what to expect over the course of a season.

What were the run expectancies after Plouffe and Rosario each stole third? Also, what if they had been thrown out? The following chart shows the run expectancy after those situations:

If Rosario had been thrown out, the inning would have ended with no runs scored. But if Plouffe had been thrown out, that also would have left the Twins without much expectation of scoring — a negative expectation of -.549, or half a run.

This change in run expectancy is essential to finding a break even percentage for evaluating the risk of stealing a base. More on that here. Finding that break-even point can provide valuable insight — for a player, manager, or fan — to determine for themselves if a steal attempt is worth the trouble. It would be important to weigh this break even point with other variables, such as the throwing ability of the catcher in Plouffe's case, or the speed and guile of the runner himself (as well as the same throwing ability of the same catcher) in Rosario's case. Indeed, the runners themselves are making these calculations intuitively throughout a game.

In Plouffe's base-out situation, a steal attempt should have a success rate of 71.3%. In Rosario's situation, a 72.8% success rate is needed. Taking into account other variables, an observer can then form an opinion on any given steal attempt — an imperfect science, to be sure.

While RE24 might tell you how advancing an extra base impacts an inning, it doesn't tell you anything about what's happening outside of the inning, or how the steal attempt will affect the score of the game.

Win Probability

An explanation of win probability can be found here. The Twins were trailing by several runs when both Plouffe and Rosario each stole their base. The chances of winning those games were not high; four- and five-run deficits are hard to overcome.

At the moment of Plouffe's steal attempt, the Twins were trailing 8-3 in the 4th inning and had an 8.4% chance of winning the game. At the moment of Rosario's attempt, they were trailing 5-1 in the top of the 7th and their win chances were 5.4%. Here is how those win probabilities changed with their steals, and how they would have changed had they been thrown out.

In this case, Molitor's intuition was correct. In terms of win expectancy, having Rosario on third base instead of second increases the win probability only from 5.4% to 5.5%. It does virtually nothing to improve the Twins chances of winning the game.

By the way, when the Detroit catcher committed the throwing error allowing Plouffe to score, the Twins win percentage increased, to 10.3%. (Again, these are benchmark numbers which may vary slightly when examining specific game logs on sites like Fangraphs.)

For another comparison, here are some other recent base running plays that impacted their games much more than the Plouffe or Rosario steals.

What this chart shows is that there are much higher leveraged base running situations which will impact a ballgame than the stolen bases of Plouffe and Rosario.

Leverage

The Leverage Index (LI) attempts to quantify the importance of a given game situation. More information here.

Fangraphs presently recognizes three types of leverage: low, medium, high. Any leverage index number lower than 0.85 is considered low, and above 2.00 is considered high. As a benchmark, the leadoff hitter in the top of the first comes to bat in a leverage situation of .87 but again, these numbers can change slightly and are regularly updated.

Here are the baserunning plays from the previous example, as measured by Leverage Index, with a dotted line to indicate low, medium, and high leverage situations:

You see how the stolen bases of Plouffe and Rosario are fairly insignificant when it comes to other moments in a ballgame that can swing the result.

The Twins have done a good job of stealing third base during Molitor’s tenure. From the start of 2015 through the end of May 2016, the Twins have stolen third base in 14 out of 17 tries, for an 82.4% success rate, where the American League average is roughly 16 steals in 20 tries.

Here is a list of all attempted steals of third base, for both the Twins and their opponents, since the start of 2015. The dashed lines indicate markers between low, medium, and high leverage situations.

Plotting these leverage numbers onto a graph shows that, consistent with Molitor’s statements above, the Twins have a greater sensitivity to risk when it comes to stealing third, at least since Molitor has been manager. Granted, there is not enough data here for a firm conclusion to be made, nor is there a strong enough correlation to support the regression lines drawn in. Nevertheless, I do feel a story starts to emerge.

As close ballgames advance into the later innings, the leverage of each play naturally increases. While opponents continue to steal third against the Twins in the late innings, more of the Twins attempts seem to come in the earlier innings. With more data, the risk aversion Molitor hinted at might become clearer.

There are four dots almost off the chart in the upper right corner that skew the data a bit.

Two of the three red dots come from the exciting game versus Kansas City back on September 9th, when Jarrod Dyson and Terrance Gore both stole third in the late innings, but did not score. The Twins eventually won that game in the 12th inning on a Miguel Sano home run. The other red dot is Oakland’s Billy Burns stealing third base and scoring the winning run in the 10th inning in a game last July.

The black dot up in the corner? That’s Eddie Rosario again, stealing third base with two outs in the bottom of the 9th in a tie game against the Seattle Mariners last August. He stole the base without a throw, and then scored the winning run on a walk-off single to left from Suzuki.

There might be some interesting information hiding in all this scattered data, or there might not. So far, the jury is still out. Neither the performances of Rosario or Molitor this season give us enough information to take hard sides on the contentious stolen base back on May 18.

One thing that does seem apparent, however, is a difference in game philosophy with the manager and playing style with at least one of the young prospects the Twins are counting upon. In order to get the best performance from their players and return to the successes they have known in the past, the Twins will probably need to reconcile these differences — not only between Molitor and Rosario, but also between others on the field and in the organization.