Which Bullpens Won the Close Ones?by Dave Studeman
April 06, 2006
How do teams consistently win close games? Well, the first answer is that they usually don't. Close games are fickle. There is no better example in baseball history than last year's Nationals, who won 12 consecutive one-run games, then lost 13 of them in a row. Close games are the random outcome generator of baseball.
Still, some teams manage to sustain winning records in close games throughout a year. Last year's White Sox, for example, were 61-34 in games decided by one or two runs. And sometimes teams manage to maintain a level of close game performance over several years. The Yankees are 241-163 in close games (one- or two- run games) over the last five years, while, perversely, the Kansas City Royals have been 168-243 in close games during the same time.
There have been many times in baseball's past when winning the close ones was the difference between winning it all and not even making it to the postseason. One example: the 1975 Red Sox famously almost won it all against the Reds, but they wouldn't have even qualified for the postseason if they hadn't gone 46-28 in close games during the regular season, while the Yankees were 33-45 and the Orioles were 37-40.
So baseball managers and analysts keep pursuing the close game edge, hoping to find the secret difference between glory and whatever the opposite of glory is. A few years ago, Baseball Prospectus' Rany Jazayerli and Keith Woolner found that bullpen strength makes a difference in close games, though the relationship is weaker than you might expect.
You see, there are many ways a team can win a close game. They can get a big lead early, then watch the bullpen fritter it away to just hang on at the end. Or, they can win with a dramatic home run in the ninth. Or, the bullpen can take over with a one-run lead in the sixth and hold the other team scoreless for the rest of the game. Or...
You get the picture. Finding the key to close-game success (if there is any key at all) requires a detailed review of all close games to see what exactly happened in each one. So my THT mates John Walsh and Bryan Donovan applied a little something called Win Probability Added to each 2005 game to see what we could see.
I've used Win Probability before to rank bullpens and I'm going to use it in this article to assess the importance of each team's bullpens in last year's nailbiters. The idea is pretty simple. On every play, the probability that a team will win rises and falls. Each team starts with a 50% probability of winning, and the winner eventually accumulates 50 net points of WPA while the loser eventually loses 50 points of WPA. That way, the winner sits at 100% at the end of the game and the loser is at 0%.
The difference between the win probability before and after a specific play is the Win Probability Added of that play. If a batter hits a home run in the first inning, his team's chances of winning rise. If he hits it in the ninth inning of a tie game, they rise a lot. So WPA is the confluence of both circumstance and performance, while winning close games is often the result of performing well when the game is on the line. That's why WPA is a very useful stat for interpreting baseball's cliffhangers.
Let me give you a few examples of how WPA plays out in close games:
- On April 9 last year, the Texas Rangers beat the Seattle Mariners 7-6 by scoring four runs in the ninth after the Mariners had scored four in the eighth. Because of the Mariners' four-run eighth, the Rangers' win probability was 74% in the top of the eighth, 4% in the top of the ninth and 100% at the end of the ninth. As you can imagine, neither bullpen ranked highly in WPA: -0.421 for the Rangers and -0.580 for the Mariners. The WPA credit went to the hitters of both teams.
- On May 9, the Padres scored four in the top of the ninth to tie the Reds 5-5, then scored again in the 13th to win the game. Because both teams were scoreless for three extra innings, the bullpens accrued decent WPA totals: 0.831 for the Padres and 0.343 for the Reds (despite the four runs allowed in the ninth by Ryan Wagner).
- On August 20, the Mets had a 8-0 lead over the Nationals (Win Probability of 99%), but their bullpen gave up eight runs in the next three innings. The Mets scored a run in the bottom of the ninth to win. In this case, both bullpens had negative WPA totals (-0.155 for the Mets and -0.154 for the Nationals) because the Mets' bullpen gave up a lot of runs but the Nationals' bullpen gave up the winning run.
To give you a bit more feel for the yin and yang of WPA, here is a table of all games in our 2005 data set that went to at least the seventh inning, listed by the WPA of the eventual winning team at the top of the seventh. I'll also include the number of games in each category, the eventual bullpen WPA of the winning team and the eventual bullpen WPA of the losing team. A good benchmark is that any team with a three-run lead at the top of the seventh has approximately a 90% probability of winning the game (according to WPA theory) and one with a two-run lead has about an 80% win probability.
7th Inn WP Games Win WPA Lose WPA 0-0.1 7 0.293 -0.581 0.1-0.2 83 0.240 -0.514 0.2-0.3 3 0.170 -0.606 0.3-0.4 120 0.246 -0.342 0.4-0.5 7 0.265 -0.332 0.5-0.6 246 0.272 -0.151 0.6-0.7 270 0.268 -0.022 0.7-0.8 10 0.212 -0.019 0.8-0.9 338 0.177 -0.011 0.9-1 76 0.042 -0.055 Totals 1160 0.223 -0.123The first row says that there were seven games in which the eventual winning team had less than a 10% probability of winning at the top of the seventh. Obviously, their offense scored a lot of runs in the last innings and their bullpen held on to the lead. You can see this in the WPA results for each bullpen: winners received 0.293 WPA points, while the losing bullpen received a whopping -0.581 WPA points for giving up those runs.
The WPA for winning bullpens is relatively steady from the top row until WP reaches about 70% probability, when it starts falling. In other words, when a team has a two-run lead or better in the seventh, the bullpen isn't going to add much to the winning effort. Of course, the reverse is also true: the bullpens of losing teams aren't usually at fault when the eventual margin was set earlier in the game, so their negative totals also move toward zero at the bottom of the table. The real action occurs when seventh-inning WP is between 30% and 70%, one-run margins and tie games.
Next, let's look at some general averages for close games. For all one- and two-run games in 2005, we allocated WPA to relievers and everyone else (batters and starting pitchers, no WPA credit was split with fielders in this analysis). Here's a simple table of what we found:
Relief Others Games WPA 115.57 -115.57 2316 WPA/G 0.05 -0.05There are 1,158 games in the data, which is doubled to 2,316 when you count each team separately. For the record, there were actually 1,163 close games last year, so our dataset is missing five games. I don't know what happened to those games, but I doubt it undermines the analysis.
When you add the totals for the winners and losers, your net result is zero. But, as you can see, relievers actually added about .05 WPA for every close game, regardless of whether their team won or lost. After thinking about it a bit, I think this makes sense.
When your team is in a close game, a manager more likely to bring in his best relievers, hurlers who are more likely to retire the side without allowing a run. So, they will be more likely to perform well in important circumstances. In fact, my guess is that, if we were to perform this analysis for all games (regardless of whether they were close or not), relievers would have positive overall WPA totals and starters would have negative overall WPA totals. This is because managers can choose when to bring in their best relievers and when to bring in their less-than-best arms. They control the circumstances in which their relivers pitch. Starting pitchers don't have that luxury.
Anyway, as you can imagine, relievers gained more WPA credit when their team won. Here is a table of WPA/Game for victors and losers:
WPA/Game Relief Others Total Victory 0.22 0.28 0.50 Loss -0.12 -0.38 -0.50Relievers receive 44% of the credit for every victory and only 24% of the "debit" for every loss. Relievers are more likely to help win a close game than blow a losing one. Let me say it a bit differently at the risk of getting it wrong: given how bullpens are used, relievers are more likely to hang onto a close lead than give one up.
Here's another angle: home teams have a decided advantage in one-run games. Last year, home teams had a .547 winning percentage in one- and two-run games (668 wins and 495 losses). Home teams tend to win more often in general (.537 winning percentage in all games last year) and they will naturally win more close games because they bat last.
For example, if a game is tied in the ninth inning and the home team leads off with a home run, then the game is over without playing the rest of the inning. This sort of thing leads to more close game victories by the home team. So here's a WPA table that splits home and road games into wins and losses:
WPA/Game Relief Others Total Home Victory 0.19 0.31 0.50 Loss -0.09 -0.41 -0.50 Road Victory 0.26 0.24 0.5 Loss -0.15 -0.35 -0.5The difference in WPA credit for bullpens is 28 WPA points at home (0.19 for wins vs. -0.09 for losses) but 41 points on the road (0.26 for wins vs. -0.15 for losses). In other words, bullpens have a bigger impact on the road than at home, and they average more than half of the team's total WPA credit for road wins.
This too makes a lot of sense. When a team wins a close game on the road, it has to shut down the home team in the bottom of the ninth without letting them take the lead. These are "high leverage" situations and bullpens that pull them off get a lot of WPA credit. So here's a hint: if you want to estimate how much a bullpen contributed (positively or negatively) to its team's record in close games, take a look at how that team performed on the road in close games. This isn't a hard-and-fast rule, just a "leading indicator."
So let's finally turn to the teams. Here is the payoff table, a list of the each team's WPA stats in close games last year. The columns list the total number of close games, total WPA in those games (remember, every win is worth .5 WPA points), the WPA gained per game by the bullpen and all others, and comments for each team.
When looking at this list, remember to compare the WPA totals to the averages listed above (.05 for relievers and -.05 for others). One other note: my comments apply to close game results only:
G Tot WPA Relief Others Comments CHA 95 13.5 0.10 0.05 Team effort: bullpen and others were both fourth best ANA 87 6.5 0.09 -0.02 Bullpen, .476 in close ones on the road (8th best) ARI 76 6.0 0.08 0.00 Second-best road record (.559) but bullpen stunk at home. BOS 63 5.5 0.00 0.09 Bullpen was second-worst. Others (Ortiz et al) were first. SLN 83 5.5 0.08 -0.01 Led by bullpen, played .550 on the road. NYA 64 5.0 0.03 0.05 Not the bullpen, the "Others" (third-best) SFN 82 3.1 0.06 -0.02 Edge to bullpen. .500 on the road (sixth-best record) ATL 70 2.0 -0.04 0.07 Worst bullpen. Think Andruw for second-best others. SDN 82 2.0 0.11 -0.08 The Pads had best bullpen. WAS 87 1.5 0.10 -0.08 Third-best bullpen. I'll look at the streaks later CHN 76 1.0 0.05 -0.04 Bullpen not bad. .512 on road but .514 at home. HOU 78 1.0 0.06 -0.05 Bullpen MIN 77 0.7 0.11 -0.10 Second-best bullpen, but only .405 in road games. TBA 79 0.5 0.02 -0.02 Not the bullpen. FLO 67 0.0 0.07 -0.07 Edge to the pen not supported by the others CIN 65 -0.5 0.04 -0.05 Reds were only .333 on the road. NYN 73 -0.5 0.03 -0.04 Second-worst road team: .297. Played .676! at home. CLE 90 -1.0 0.07 -0.08 Only team with better road record (.548 vs. .438 at home) OAK 72 -1.0 0.07 -0.08 Very good bullpen, .457 record on road (9th best) TEX 80 -1.0 0.02 -0.03 Bullpen was 25th-worst; others were 10th-best. COL 86 -1.1 0.02 -0.03 Bullpen not as bad as I expected. PHI 69 -1.5 0.04 -0.07 Bullpen decent but not on the road (.385 record). MIL 82 -2.0 0.01 -0.04 Bullpen had the third-worst WPA total. DET 83 -3.5 0.05 -0.09 Bullpen average but others were 23rd. SEA 80 -5.2 0.05 -0.11 Good bullpen, much worse on road (.357 vs. .556) BAL 64 -6.0 0.01 -0.10 Decent on the road (.394 vs. .438 at home) LAN 84 -6.0 0.04 -0.11 Dodger bullpen wasn't bad. Everyone else was. TOR 73 -7.5 0.06 -0.16 Good bullpen but "others" were last in the majors. PIT 74 -8.0 0.02 -0.12 Bad. KCA 76 -9.0 0.01 -0.13 Worse. Royals were .235 on road; .512 at home.Yes, bullpens help win close games. But winning close games is a team effort. A couple of last year's teams, the Red Sox and Braves, had very good records in close games despite having the two least-contributing close-game bullpens in the majors. On the other hand, the Twins and Padres had great bullpen contributions in close games, but their other players didn't support them.
We didn't really expect to find a "key" here, and we didn't. But we did get a better feel for how each team won or lost their close games. As this season progresses, we'll do the same thing in 2006, too.
References and Resources
None of this analysis would be possible without the generous contributions of Tangotiger's and Keith Woolner's research.
You can find the outcomes of all close games in baseball history at Baseball Reference.
Dave was called a "national treasure" by Rob Neyer. Seriously. Comments about this article can be sent to him through the miracle of e-mail.
<< Return to Article