Sunday, January 6, 2013

Anomaly, Or Not To Be

This inquiry began when my good friend Dan Heaton sent me an article from the New York Times about the game-winning patterns of World Series winners. The geekish article told us, for instance, that apart from sweeps, the most common pattern is for the winning team to win Games 1, 3, 4, and 5. That has happened ten times, compared to the three times that winner has won all four home games (1, 2, 6, 7). Only once has a team won Games 1, 2, 5, and 7, while no winner (surprisingly) has ever won Games 1, 4, 6 and 7. And so on. More relevant to the discussion here, no team has ever needed six or seven games to capture the title after winning the first three games. The Red Sox may have come back from a 3-0 deficit in the LCS, but in the World Series, no team has won even twice after going down 3-0.

Here's what Dan wrote that started a rather heated e-mail debate: "What I find stunning. . .is that teams that start 3-0 are 21-3 in game 4! Is this statistically significant--can we really say that teams that start 3-0 are typically that much better than their opponents? Clearly not--the best team in baseball isn't enough better than the worst to play .875 against them. So it's an anomaly--but what an anomaly!"

My instinctive response was that there was most likely a strong psychological factor making the trailing teams so hapless in Game 4. The weight of historical precedent would make the losing team realize that they had no chance, and at some level they would be more mentally prepared to get to the golf course a day sooner than to play ball. Dan dismissed this as nonsense, insisting that a team that had come that far already and had fought its way through two previous postseason series would not suddenly succumb to collective despair.

We didn't settle anything on that front, so I turned to the numbers themselves. My feeling was that even though 21-3 looked like an anomaly and sounded like an anomaly, it didn't smell to me like an anomaly. It didn't automatically have to be dumped into the "shit happens" file or dismissed because statistically speaking it is the kind of small sample where any result could occur. Maybe it wasn't inexplicable. Maybe there was a common factor or thread which might explain at least most of those results. In fact, I started thinking, maybe it wasn't even an unexpected result under those circumstances.

I decided to look at the other two professional sports which have a best-of-seven championship series. First I went through NBA final series going back to 1951. That year, the Rochester Royals led the New York Knicks in the finals by a 3-0 margin, but the Knicks fought back to take the next three games before the Royals won Game 7. Since then, there have been nine final series in which one team led, 3-0. In eight of the nine, that team has swept. So it's 8-2 overall in the NBA, not as extreme as baseball but still quite tilted toward sweeps.

Next up was the NHL, a quest that took me back to 1919, when Montreal led Ottawa 3-0 in the finals, lost Game 4, and wrapped up the series in five games. There were no other 3-0 leads until 1941 (in many of the intervening seasons, the finals were not best-of-7), which began a stretch of six straight years when one team took a 3-0 lead in the finals. Those teams split the Game 4 results; in 1942, as hockey nuts can tell you, the Toronto Maple Leafs came back from trailing the Detroit Red Wings to win the Stanley Cup in seven games, while in 1945 the Maple Leafs almost blew a 3-0 lead, needing seven games to defeat the Red Wings.

Through 1960, teams leading 3-0 in the Stanley Cup finals were only 7-5 in Game 4. That didn't look good at all for my anti-anomaly theory, but a funny thing happened once the NHL doubled in size in 1967-1968 and added preliminary playoff series, as baseball did one year later. Since 1967, NHL teams leading 3-0 in the Stanley Cup Finals have gone 13-2 in Game 4. In the NBA, the figure since that time is 7-1, and in baseball, sports fans, it is 9-1.

If you look at the overall totals for the three sports, the record in Game 4 for teams ahead 3-0 is 49-12, a winning percentage of .803. From 1967 through 2012, however, it is 29-4, a winning percentage of .879. In that light, baseball's overall .875 seems to be right on the money in terms of what you would expect from a professional team. Of course, the three sports have very different formats, and the nature of the sports themselves are different enough that the reasons for each sport's result might be unique. The fact is, however, that when pro teams in a championship series find themselves down 3-0, they almost always lose. The 21-3 record in the World Series is not an anomaly. We should expect a sweep.

I emphasize "championship series" because I also looked at the various preliminary series for these sports, which historically are far more plentiful in hockey, where the deeper playoff field allows more mediocre teams to compete, thereby increasing the odds of a sweep in earlier series. Before I starting my research, I thought about what I might expect the Game 4 result to be between any two teams in any situation when one team has won the first three contests. Figuring that the majority of those winning teams would most likely be superior in talent, I guessed it would be upwards of two-thirds of the time. The numbers from preliminary series confirm that suspicion. In the NHL, the leading teams have gone 89-56 in Game 4 (.618), and in the NBA they are 54-35 (.607). In baseball, which has conducted far fewer best-of-7 playoff series, only nine times has one team in an LCS led, 3-0. They have gone 6-3, including of course the 2004 Yankees, the only baseball team ever to lose a post-season series after being ahead 3-0. Take the three sports together and the record is 149-94, a winning percentage of .613.

There's a very big difference between .613 and .803, especially when you consider (as Dan emphasized) that the team making the finals has already passed difficult tests in earlier series. They are presumably stronger, more talented, and more resilient than the teams that have fallen by the wayside, but that only increases my belief that there must be some rational way of accounting for the extreme results we see in final series. The much higher percentage since the advent of tiered playoffs suggests that playing all those extra games provides that many more chances for catastrophic injuries or cumulative fatigue. A team playing grueling preliminary series might simply run out of gas in the finals, especially if the penultimate series was a tough one.

In recent years, when the major league team with the best record during the regular season has hardly ever won the World Series, it has become clear that simply getting to the Series is an accomplishment, a matter of survival for even the best team. In fact, more than a few times I've heard players from the winning LCS series declare, in effect, that "I'm just happy to be playing in the Series. It's all gravy from here. It doesn't even matter if we win." With that prevailing sense of relief at merely making the Big Show, it doesn't seem inconceivable to me that a team falling behind 3-0 would feel, individually if not collectively, "Oh well, we're toast, but at least we made it this far." This isn't like flipping coins, where each event is an isolated, fresh opportunity. Post-season factors are cumulative, and nearly 20% of all World Series have been sweeps.

Let's take a closer look at the 24 times a team has gone up 3-0 in the World Series. How many times has it truly been a mismatch? Is there a pattern or a persistent factor that accounts for many of the results? I looked at a number of things which might have contributed, notably regular-season record, run differential (runs scored minus runs allowed), pitching matchup, playoff pathways, and Game 3 results. We can get six of the 24 out of the way at the outset.

 There have been three huge World Series upsets that were sweeps. In 1914, the "Miracle Braves" swept Connie Mack's first-dynasty Athletics, who won five more games during the season and had a big advantage in run differential. However, the Braves, after being in last place on the Fourth of July, got as hot as any team ever, going 68-19 after that and winning 25 of their last 30 games to build up huge momentum that not even the Athletics could withstand. In Game 4, the Braves started Dick Rudolph, a 26-game winner, and he won, 3-1.

The second huge upset occurred in 1954, when the New York Giants shocked the Cleveland Indians, whose 111-43 record that season is still the American League record for winning percentage. However, Indians manager Al Lopez overworked his starters during the final week, including letting Mike Garcia pitch 12 innings on the final day in a futile attempt to get him a 20th win. In the Series, Dusty Rhodes burned the Indians in the first two games with clutch pinch-hits, mainly a 10th-inning home run off Bob Lemon to win Game 1. Lopez brought Lemon back on two days' rest to start Game 4, and the Giants countered with a well-rested Don Liddle, whose only Series appearance before that was pitching to one batter in Game 1, inducing Vic Wertz to blast the ball on which Willie Mays made "The Catch." Lemon, clearly out of gas, gave up seven runs in four innings, and the Giants had their sweep.

The third big upset was the Reds sweeping the A's in 1990. During the season, the A's won 103 games to the Reds' 91, and Tony LaRussa's team swept the Red Sox in the LCS to head into its third straight World Series as a big favorite. What happened? Five days off after the LCS may have flattened their momentum, but their main fault was a failure to handle three five players: Jose Rijo, Billy Hatcher, and the "Nasty Boys" of the Cincinnati bullpen--Randy Myers, Rob Dibble, and Norm Charlton. Rijo and two of the relievers combined for a shutout in Game 1, while Hatcher went 7-for-7 and scored five runs in the first two games. Game 4 was a rematch between Rijo and Dave Stewart, who surrendered the home run to Eric Davis in Game 1 which seemed to erase the A's invincible aura. Rijo allowed a run on two hits in the first inning of Game 4, then held the A's hitless until leaving in the ninth inning with a 2-1 lead that Myers finished off with no sweat, leaving the Reds bullpen unscored-on in the Series.

Now let's look at the three times that the losing team managed to win Game 4. The first time was in 1910, when the pitching-dominant Athletics and Cubs seemed evenly matched going on until the Athletics stormed ahead by drilling Chicago pitching to win by scores of 4-1, 9-3, and 12-5. The Game 1 winner, Chief Bender, went for the sweep and took a 3-2 lead to the ninth inning before the Cubs tied the game on player-manager Frank Chance's RBI triple.That's how close the Athletics came to a sweep. The Cubs beat Bender in the tenth inning, but in Game 5, Jack Coombs, who won 31 games during the season, added his third Series win to wrap up the title.

The second Game 4 comeback happened in 1937 and was the only game that prevented the Yankees from sweeping three consecutive World Series. The pitching matchup was the key in this one. The desperate Giants started their ace, 22-game winner Carl Hubbell, on two days' while the Yankees went with Bump Hadley, a modest 11-game winner during the season, giving Game 1 winner Lefty Gomez his normal rest. The Giants drilled Hadley for five runs in the second inning, and Hubbell coasted to a 7-3 victory. Gomez took care of business in Game 5, and that was that.

The final Game 4 comeback marked the only time it has happened in the last 16 3-0 situations. It happened in 1970 when the Orioles also came very close to sweeping the Reds. Both teams were juggernauts that year, though the Orioles had a much better run differential. The Orioles won 108 games during the season while the Reds won 102, and both teams swept their LCS battles. In the Series, the Orioles exposed the relative weakness of the Reds' starting pitching, and in Game 4, 20-game winner Jim Palmer took a 5-3 lead to the eighth inning. A walk and a single started the inning and ended Palmer's day, and Lee May blasted an Eddie Watt pitch for a three-run home run that gave the Reds a 6-5 win. The teams returned to form the next day as the Orioles knocked out Reds starter Jim Merritt in the second inning and won, 9-3.

Next up, we have nine teams that swept the World Series against demonstrably outmanned opponents. While you might quibble about whether these were total mismatches, I don't think you can say that any of these sweeps was a surprise. In the interest of conserving space, I'll summarize each of these in a sentence; for each team, I'll give the regular-season games won and run differential in parentheses:

  1. 1907: The Cubs (107,+184) and the Tigers (92,+162) played a 12-inning tie in Game 1, after which the Cubs breezed to four straight wins by using superior starting pitching to limit the Tigers to three runs in the four games.
  2. 1927: The Yankees (110,+376) defeated the Pirates (94,+158) easily to cement the reputation of its formidable "Murderers Row" lineup.
  3. 1932: The Yankees (107,+278) destroyed the Cubs (90,+87) by scoring 37 runs in four games as Lou Gehrig, Babe Ruth, and Earle Combs combined to score 23 runs and drive in 18.
  4. 1938: The Yankees (99,+256) again trounced the Cubs (89,+116), outscoring them 22-9.
  5. 1939: The Yankees (106,+411!) embarrassed the Reds (97,+162) although they trailed by two runs going to the ninth inning of Game 4 and capitalized on Ernie Lombardi's "snooze" to win in the tenth.
  6. 1950: The Yankees (98,+223) had a tough time but still swept the Phillies (91,+98), winning 1-0 in Game 1 and scoring the winning run in their last at-bat in the next two games. 
  7. 1998: The Yankees (114,+309) finished off their winningest season in style against the Padres (98,+114) although the Padres did lead two of the games by three runs after six innings.
  8. 2005: The Yankees--no, wait, the White Sox (99,+96) swept the wild-card Astros (89,+84) even though it took 14 innings to win Game 2, with Game 4 a 1-0 thriller.
  9. 2007: The Red Sox (96,+210) cooled off the torrid Rockies (90,+102), bursting their bubble with a 13-1 drubbing at Fenway Park in Game 1 after the Rockies went from fourth place on September 15 to the World Series by winning 13 of their last 14 games, taking a one-game playoff to win the wild-card spot, then sweeping their first two playoff series. 
That leaves nine more, a lineup of diverse duels with the same outcome. We'll look at them chronologically, starting with the 1922 World Series, which like 1907 included a tie game. The Giants and Yankees were closely matched, winning 93 and 94 games respectively, and all five Series games were low-scoring and close. The key game was the first one, when Bullet Joe Bush blew a 2-0 lead in the eighth inning. Irish Meusel (136 RBI during the season) tied it with a bases-loaded single, and Ross Youngs' sacrifice fly won it. After a tie in the second game, wins of 3-0 and 4-3 put the Giants on the brink of a sweep. The Yankees had the pitching edge in Game 5, a reprise of the Game 1 pairing with the 26-7 Bush facing Art Nehf (19-13). In a case of deja vu, Bush again led in the eighth inning, 3-2, but another three-run rally did him in. The key factor in the sweep was that the Giants pitchers handled Babe Ruth, holding him to a .118 average and one RBI. The Giants had a deeper offense--they scored 95 more runs during the season--and by negating Ruth they gave themselves a better chance for the two come-from-behind victories.

The 1928 Yankees won six more games than their World Series opponents, the Cardinals, and had the same "Murderers Row" lineup that had swept the Pirates in 1927. So this was a borderline mismatch, and the first three games weren't close--4-1, 9-3, and 7-3. Game 4, at St. Louis, was a repeat of the Waite Hoyt-Bill Sherdel matchup of aces from Game 1. This brings up an interesting pattern in the 24 games studied here. In 15 of them, the team down 3-0 was at home in Game 4. In other words, after losing the first two games on the road, a team still returns home with high hopes, looking at three games with the home advantage. After all, ten times a team has gone down 2-0 in the World Series and come back to take the title. But when they lose Game 3, it seems to shatter what is left of their fragile confidence, and they have gone 2-13 in Game 4. So it was here, as the Cards led 2-1 after six innings before the roof caved in. Babe Ruth and Lou Gehrig homered in the seventh (Ruth's second of three in the game), and the final score was 5-2.

Jump ahead to 1963, when the Yankees had better credentials (104 wins,+167) than the Dodgers (99,+90). But the Dodgers had the big equalizer, Sandy Koufax, coming off his first Cy Young Award season and the pitching Triple Crown (25-5, 1.88 ERA, 306 Ks). He proved it was no fluke in Game 1, completely intimidating the Yankees by retiring the first 14 batters with nine strikeouts and only one fair ball. The Yankees didn't score a run before the seventh inning in any of the four games, and after Don Drysdale won Game 3, 1-0, it was Koufax again in Game 4, facing Whitey Ford, no slouch with a 24-7 record that season.  Ford outpitched Koufax the second time around, surrendering two hits to Koufax's six. One was a Frank Howard which was matched by Mickey Mantle in the seventh inning. In the home half, Yankees first baseman Joe Pepitone lost a throw from third base in the white-shirted crowd at Dodger Stadium, and the Dodgers cashed in an unearned run to gain the sweep. Chalk this one up to Hall of Famers Koufax and Drysdale logging 27 innings compared to two-thirds of an inning of work by the L.A. bullpen, and it isn't that surprising that the Yankees offense suffered, scoring just four runs in the Series.

The shoe was on the other foot in 1966, when the Dodgers' offense literally vanished after the third inning of Game 1. The Dodgers were still built around pitching and speed, and their hitters became so discombobulated by Moe Drabowsky that they never recovered. Drabowsky relieved Dave McNally in the third inning at Dodger Stadium with the Orioles leading, 4-1, and the bases loaded with one out. He walked in a run but held the Dodgers to one hit the rest of the way, striking out 11, including six in a row. Game 2 was the sad finale of Koufax's career, as the Dodgers committed five errors behind him, but it didn't matter because Jim Palmer pitched the first of three consecutive shutouts by the Orioles. The last two were identical 1-0 thrillers won by home runs, Paul Blair supporting Wally Bunker in Game 3 and Frank Robinson's fourth-inning blast off  Don Drysdale providing all the help McNally needed to polish off the sweep. I'm going to go out on a limb here and surmise that heading into Game 4, the Dodgers were not overconfident about scoring, much less winning.

The 1976 Reds are the only team in the playoff era to sweep all their post-season games. In retrospect, this one was a clear mismatch, as the Reds led the major leagues in runs, batting average, home runs, hit, walks, and fielding percentage, and were second in stolen bases. The "Big Red Machine" acted invincible, too, though Yankees manager Billy Martin refused to concede their superiority even after being swept in the World Series and outscored 22-8. The Reds ran wild on the weak arms of the Yankees outfield, and their bullpen stifled the Yankees for nine shutout innings, allowing only two hits. After an easy 6-2 win in Game 3 at Yankee Stadium, Gary Nolan took on Yankees ace Ed Figueroa in Game 4. The Yankees took their only lead of the Series with a first-inning run, but Johnny Bench's two-run home run in the fourth inning gave the Reds the lead for good, and his three-run blast in the ninth inning capped the 7-2 finale.

What can you say about the "Earthquake Series" of 1989 except that both Bay Area teams were glad to survive it? On paper it was a borderline mismatch between the 99-63 A's and the 92-70 Giants. Dave Stewart and Mike Moore held the Giants to one run in two games at Oakland, and after eleven days off because of earthquake damage, the Giants were still as shaken as the Candlestick Park infrastructure. The final two games were blowouts, with the A's leading 8-3 and 7-0 after five innings. The aggregate score for the Series was 32-14, but it wasn't that close, and there wasn't much suspense about Game 4.

One sweep that is tough to explain occurred in 1999. The Braves (103 wins,+179) had a better season than the Yankees (98,+169), they had plenty of offense (five players with 20+ home runs, led by Chipper Jones with 45, and a starting rotation featuring three future Hall of Famers plus Kevin Millwood, who outpitched them all with an 18-7 record and a 2.68 ERA. The Yankees were deep and balanced, of course, and had lost only one game in two preliminary series in their quest for a third title in four years. It figured to be a close Series, and in fact the Braves led after seven innings in two of the first three games (the Yankees drilled Millwood in Game 2). But Greg Maddux let a 1-0 lead get away in the eighth inning of Game 1, and in Game 3 Glavine blew a 5-2 lead, allowing three home runs, and Chad Curtis won it in the tenth inning with his second homer of the game. In Game 4, Joe Torre played his ace in the hole, Roger Clemens, who held the Braves scoreless through seven innings, and the Yankees won easily, 4-1. Pitching turned out to be the key to this sweep, as the Braves scored only nine runs and, apart from Bret Boone and Chipper Jones, batted .154 as a team. Let's face it--when you trot Clemens out there against a team whose offense has already gone south for three games, is the resulting sweep a surprise?

Next up is another Series in which the team with the better regular season, the Cardinals (105 wins,+196) got swept. I doubt that Dan would accept karma as a tangible factor in a sports competition, but from this vantage point it's hard to argue that the Red Sox (98,+181) were not destined to win their first World Series since 1918 after their unprecedented, emotionally charged comeback from being down 3-0 to the Yankees in the LCS. But that in itself does not account for the sweep, especially since both teams had to go seven tough games to win the LCS. However, only the first game of the Series was close, a slugfest in which the Red Sox blew an early 7-2 lead by the sixth inning, lost another lead in the eighth, and prevailed, 11-9, on Mark Bellhorn's late home run. After that, they cruised thanks to terrific starting pitching, holding the Cardinals to 13 hits over the final three games. Curt Schilling and Pedro Martinez didn't allow an earned run in Games 2 and 3, and suddenly the Cardinals were facing a sweep. Tony LaRussa gave the ball to Jason Marquis, 15-7 during the season but sporting a 7.36 ERA in two previous post-season starts. When Johnny Damon led off the game with a home run, it was all the support needed by Derek Lowe, whose charmed life already included clinching victories in the two earlier series. Trot Nixon doubled in two runs in the third inning, and Lowe duplicated Pedro's performance by scattering three hits in seven shutout innings. As with so many of the sweeps described here, this was a case of dominant starting pitching completely smothering the losing team's offense. After Game 1, the Cardinals batted a miserable .143. Game, set, and match.

Finally we come to 2012, when the Giants swept the Tigers. After the Giants dispatched Tigers ace Justin Verlander in four innings in Game 1 thanks to the first two of Pablo Sandoval's three home runs, the Tigers knew they were in trouble. They managed two hits off Madison Bumgarner in Game 2 and improved to five hits in Game 3, but couldn't score a run. So there they were, another team helpless at home and ready to put their offense on a milk carton. The pitching matchup looked close, with Matt Cain (16-5) facing Max Scherzer (16-7). The game was close, too, tied 3-3 in the seventh inning when both starters departed and still 3-3 into extra innings. This one could have gone either way, but Marco Scutaro singled in a run in the top of the tenth inning and Sergio Romo capped an outstanding post-season by striking out the side.

So there's your answer, Dan. Given where these World Series stood after three games, you might make a strong case for a handful of different results for Game 4, but not more. Even if we let a few teams avoid a sweep, that only brings the overall record to 18-6, still a fair distance from the 15-9 it would take to approach the .613 winning percentage I calculated for all teams up 3-0 in any post-season series in the three big pro team sports.

Is there a common denominator? Yes:  great starting pitching. In the 21 sweeps studied here, the winning teams' starting pitchers in Game 3 averaged 7 1/3 innings per start with a cumulative ERA of 2.26 (2.01 without Andy Pettitte's start in 1999). They allowed no runs five times, and more than three earned runs only three times. By proving to the losing team that the first two wins were no fluke, by further frustrating their offense, and by putting them in a position from which no World Series contender has ever extricated itself, they set the stage for even greater domination in Game 4. In those 21 games, the starters also averaged 7 1/3 innings pitched, with a cumulative ERA of 1.40. Seven allowed no earned runs, and seven others allowed only one earned run. Some combination of talented pitchers and demoralized hitters gave the trailing team virtually no chance of winning Game 4. (The three times the trailing team won Game 4, the opposing starters posted a 3.04 ERA in Game 3 and 7.00 in Game 4.)

There you have it. Thank you for bearing with me through this lengthy dissection. I hope I've convinced you that you ought to be shocked the next time a team down 3-0 in the World Series wins Game 4, something that hasn't happened in the last eight opportunities. Good. I'm glad you're with me, because I've been thinking about Dan's assertion that "the best team in baseball isn't enough better than the worst to play .875 against them." Somehow, I still can't detect the bittersweet stench of an anomaly. I wonder if he believes that if they couldn't play .875 ball against the worst team, we ought to give them credit for playing .875 ball against any other (presumably better-than-the-last-place-) team. For instance, the 1927 Yankees were a mere 18-4 against the last-place Red Sox (.818), but went 21-1 (.955) against the seventh-place St. Louis. Let's see. . .


cheeseblab said...

Great recap. I really am exhausted discussing this, but one observation I would make is that the aggregate performance by the starters for the winning side in all WS finales is probably about as impressive as in the 21 sweeps. World Series champs tend to outpitch and outhit the teams they beat (1960 aside). That's why they win.

My only other comment is: "Well, I got my guy!"

Gabriel Schechter said...

You're pretty much right about ALL starters for the deciding game winners. It's 1.86. During the Deadball Era, it was 1.01. Since then, it is 2.07. Surprisingly, 15 times the starter hasn't made it past the 4th inning, and three times has gotten knocked out in the 1st inning. (The Pirates starter in Game 7 in 1960 went five innings and gave up three runs, so he was far from the worst.)