Turnover

The question: Does keeping a team together after a World Championship provide more wins than we would otherwise expect? There are several caveats to point out before answering:

1. Any team that wins the World Series is probably pretty good to begin with. So if they kept most of the players that got them there, we would expect them to do well. What we're looking for is whether keeping a team together provides more success, by and large, than making changes.
2. What we're looking for is something that cannot be explained by the performance of the players. The argument is that the team is greater than the sum of its parts. The theory we are testing states that even if you are able to replace your players with others of equal or better skill, it's won't be as successful as retaining the core of the World Championship team.
3. I'm going to look at the teams as they were the year they won the Series, and as they were the subsequent year. I'll see which players they lost and which players they gained. I'll compare their won-lost records, as well as their Pythagorean won-lost records. The Pythagorean Won-Lost Theory, if you're not familiar with it, states that a team's W-L record can be accurately predicted from the number of runs they score and number of runs they allow. While an actual W-L record can benefit greatly from luck and other outside forces, the Pythagorean record gives us a better idea of how well the team performed, as opposed to how many games they won. This may help us explain a rise or fall in a team's wins. If a team's W-L record is 96-66 and their Pythagorean record is 90-72, the team has probably been lucky. We would expect their W-L record to fall the next season. This will help us determine if a team is actually performing better, or if perhaps they just got lucky.
4. I'm confining my study to modern teams. A study of teams around before free agency would not really be relevant to this study, as teams had complete control over every player gained and lost. The circumstances of player movement between teams, not to mention the singular challenges in building a team, make modern teams incomparable to teams in the past.

So let's look, team-by-team, at some of the World Champions of the past 15 years. The Rough Level of Turnover is a rough figure representing what percent of the team's At Bats and Innings Pitched were lost after their World Series victory. Pythag. Win-Loss Record is represented by pW-pL.

2004 Boston Red Sox
Key players lost: Pedro Martinez, Derek Lowe, Orlando Cabrera, Nomar Garciaparra
Key players gained (2005): Edgar Renteria, David Wells, Matt Clement, John Olerud
2004 W-L: 98-64; 2005 W-L: 95-67; Net Gain/Loss: -3 Wins
2004 pW-pL: 96-66; 2005 pW-pL: 90-72; Net Gain/Loss: -6 Wins
Rough Level of Turnover: 21% --Moderately High
Conclusion: The Red Sox suffered a good deal of turnover and failed to repeat. This could support the "chemistry" argument. However, the truth is that the players the Sox added in 2005 produced more in quality and quantity than the 2004 players they replaced. This may seem hard to believe. But Matt Clement and David Wells together were more valuable than Pedro and D-Lowe were in 2004. And Edgar Renteria was more productive than the Sox' 3 SS in 2004.
The reason the Sox performed worse in 2005 was due to the huge drop in quality by the players they kept. If Curt Schilling, Keith Foulke, and others such as Kevin Millar had even come close to their 2004 numbers, the Sox very well could have repeated. So it would seem that the 2004 Sox don't support the "chemistry" theory, unless you believe that the reason Schilling, et al didn't perform is because they missed Pedro and Nomar.

2003 Florida Marlins
Key players lost: Derrek Lee, Ivan Rodriguez, Mark Redman, Braden Looper, Ugueth Urbina
Key players gained (2004): Armando Benitez, Hee Seop Choi, Damion Easley, Paul Lo Duca
2003 W-L: 91-71; 2004 W-L: 83-79; Net Gain/Loss: -8 Wins
2003 pW-pL: 83-79; 2004 pW-pL: 83-79; Net Gain/Loss: Even
Rough Level of Turnover: 28.6% -- High
Conclusion: It looks like the Marlins got worse in 2003, but I would argue that it was simply a case of their luck running out, as the Pythagorean record would seem to indicate. The 2003 Marlins are, in my opinion, one of the worst teams to win the World Series.

2002 Anaheim Angels
Key players lost: Orlando Palmeiro
Key players gained (2003): (none of note)
2002 W-L: 99-63; 2003 W-L: 77-85; Net Gain/Loss: -22 Wins
2002 pW-pL: 101-61; 2003 pW-pL: 80-82; Net Gain/Loss: -21 Wins
Rough Level of Turnover: 6.8% -- Very Low
Conclusion: The Angels kept everyone from their World Series roster except Palmeiro, their 4th outfielder. And it got them a loss of 22 wins. The Angels are the strongest and most obvious argument against keeping your roster together for the sake of chemistry. Their level of turnover is among the lowest in the free agent era, and yet they simply sank to the bottom. It's very, very hard to explain this Angel team if you believe in keeping your team together for the sake of chemistry.

2001 Arizona Diamondbacks
Key players lost: Reggie Sanders, Albie Lopez
Key players gained (2002): Quinton McCracken, Rick Helling
2001 W-L: 92-70; 2002 W-L: 98-64; Net Gain/Loss: +6 Wins
2001 pW-pL: 95-67; 2002 pW-pL: 95-67; Net Gain/Loss: Even
Rough Level of Turnover: 12.5% -- Low
Conclusion: The Diamondbacks kept their team together and performed almost exactly as well. But they didn't even make it to the World Series, losing in the NLDS. This is another wrinkle in this study; winning the World Series isn't always a question of being the best team. So planning a team to win the World Series is problematic at best; the easiest thing is to try and create the best team, which would then have a better chance to make the World Series. But repeating as World Series Champions requires a whole lot of luck, perhaps even as much luck as talent.

2000 New York Yankees
Key players lost: Jeff Nelson, Dwight Gooden, Glenallen Hill
Key players gained (2001): (none of note)
2000 W-L: 87-74; 2001 W-L: 95-65; Net Gain/Loss: +8 Wins
2000 pW-pL: 85-76; 2001 pW-pL: 89-71; Net Gain/Loss: +4 Wins
Rough Level of Turnover: 18.2% -- Moderate
Conclusion: The Yankees went from a bad team (relatively speaking) winning the World Series to a much better team losing the World Series. Such is baseball. But there's nothing here out of the ordinary.

1999 New York Yankees
Key players lost: Chili Davis, Hideki Irabu, Chad Curtis
Key players gained (2000): David Justice, Glenallen Hill, Dwight Gooden
1999 W-L: 98-64; 2000 W-L: 87-74; Net Gain/Loss: -11 Wins
1999 pW-pL: 96-66; 2000 pW-pL: 85-76; Net Gain/Loss: -11 Wins
Rough Level of Turnover: 15.6% -- Fairly Low
Conclusion: The Yankees kept almost almost all of the key players from their 1999 World Series sweep of the Braves and ended up a significantly worse team. But they still repeated as World Champions. I'm starting to sense that luck has about as much to do with winning the World Series as roster moves, aren't you?

1998 New York Yankees
Key players lost: David Wells, Tim Raines
Key players gained (1999): Roger Clemens, Jason Grimsley
1998 W-L: 114-48; 1999 W-L: 98-64; Net Gain/Loss: -16 Wins
1998 pW-pL: 108-54; 1999 pW-pL: 96-66; Net Gain/Loss: -12 wins
Rough Level of Turnover: 10.7% -- Low
Conclusion: If ever there was a team that should be kept together, it's the '98 Yankees, right? Well, they kept it together and lost 16 wins. But this isn't a lesson about the loss of David Wells' "chemistry," it's a lesson about regression to the mean; whereby any extreme performance (good or bad) will tend to regress back toward the average in the future.

1997 Florida Marlins
Key players lost: Kevin Brown, Moises Alou, Robb Nen, Jeff Conine, Devon White, Kurt Abbott, Al Leiter, Darren Daulton, Dennis Cook
Key players gained (1998): Derrek Lee, Todd Zeile*, Dave Berg
* -- the Marlins acquired Zeile on May 14 and traded him away on July 31.
1997 W-L: 92-70; 1998 W-L: 54-108; Net Gain/Loss: -38 Wins
1997 pW-pL: 88-74; 1998 pW-pL: 58-104; Net Gain/Loss: -30 Wins
Rough Level of Turnover: 35.3% -- Obscenely High
Conclusion: This isn't really relevant to the argument, but I have to talk about the complete decimation of the 1997 World Champion Florida Marlins. When the 1999 season opened, the only player left from the '97 starting lineup was 2B Luis Castillo. Only Livan Hernandez and Alex Fernandez were left from the starting rotation. Not only that, but the Marlins also traded away all of their marginal players. Only Craig Counsell, Cliff Floyd, and Mark Kotsay were left (when 1999 opened) from the '97 bench, and only Rob Stanifer, Kirt Ojala, and Antonio Alfonseca were left from the bullpen.
I also have to point out just how little the Marlins got in exchange for all these players. Here's a look at who the Marlins traded away, and who they got in return:
TRADED AWAY: Mike Piazza, Moises Alou, Kevin Brown, Gary Sheffield, Robb Nen, Edgar Renteria, Al Leiter, Todd Zeile, Bobby Bonilla, Devon White, Charles Johnson, Jeff Conine, Kurt Abbott, Jim Eisenreich, Dennis Cook, Manuel Barrios, Jay Powell, Scott Makarewicz
RECEIVED IN RETURN: Derrek Lee, A.J. Burnett, Preston Wilson, Braden Looper, Pablo Ozuna, Oscar Henriquez, Mark Johnson, Jesus Martinez, Joe Fontenot, Mike Pageler, Mike Villano, Blaine Mull, Rafael Medina, Steve Hoff, Fletcher Bates, Scott Comer, Eric Ludwick, Jesus Sanchez, Robert Stratton, Ed Yarnall, Geoff Goetz, Ramon Castro, Daniel DeYoung, Jose Santo, Armando Almanza
Never has any organization made such a mockery of baseball as the 1997 Marlins. Not only did they trade away a horde of excellent players, they didn't even bother to get crap in return.

But back to the point: After one year, the Marlins had lost 35.3% of their 1997 World Series production. When the 1999 season began, they had lost 71%. The 2-year loss is a record, but would you believe that the '97 Marlins do not hold the record for highest percentage of AB/IP lost in one year?
So who does hold that record? Well, let's just say that they prove my point and render any defense of the "chemistry" theory useless. So I'll just go ahead and skip to them.

1992 Toronto Blue Jays
Key players lost: Dave Winfield, Candy Maldonado, Jimmy Key, Tom Henke, Manuel Lee, Kelly Gruber, David Cone, Derek Bell
Key players gained (1993): Paul Molitor, Tony Fernandez, Dave Stewart, Danny Cox, Tony Castillo
1992 W-L: 96-66; 1993 W-L: 95-67; Net Gain/Loss: -1 Win
1992 pW-pL: 91-71; 1993 pW-pL: 91-71; Net Gain/Loss: Even
Rough Level of Turnover: 40.23% -- Very, Very High
Conclusion: Would you believe that the team with the highest post-championship turnover I've ever seen managed to repeat? The Blue Jays lost 2210 of their 5536 ABs and 589 of their 1440 IP. It would seem that any team so thoroughly gutted would be completely unable to make it back the next year. How much money would it cost to replace all those ABs and IP with anything but minimum-wage replacement players? The Blue Jays did it somehow and did just as well as they did in 1992.

I looked at every team from 1990-2004 and found no correlation between keeping a team together and subsequent success, other than that which can be explained by the quality of players. There's every indication that a significant level of turnover can help a team, if the new players can adequately replace those lost.
Not only did I not find any examples matching the archetype (a team that keeps its team virtually intact and performs better than expected), but I found two teams that demolished the archetype. There were the 2002 Anaheim Angels, a team that stayed intact and got significantly worse, and the 1992 Toronto Blue Jays, who lost a historic percentage of their championship team and were able to repeat as World Champions. I didn't list the teams in between, because I did not see them as strongly denying my central thesis. I'll go ahead and list the teams I studied not mentioned above here:

1996 Yankees: 9.6% turnover (low); +4 Wins, +12 pWins -- Lost in 1997 ALDS
1995 Braves*: 9.1% turnover (low); -3 Games; +1 pGame -- Lost in 1996 World Series
1993 Blue Jays*: 16.1% turnover (moderate); -16.5 Games, -11.5 pGames -- finished 3rd when 1994 season ended
1991 Twins: 22.2% turnover (fairly high); -5 Wins, -3 pWins -- finished 2nd in 1992
1990 Reds: 8.8% turnover (low); -17 Wins, -18 pWins -- finished 5th in 1991
* -- I used overall games instead of wins, because the 1995 and 1994 seasons were strike-shortened.

You may feel that the 1991 Twins support the "chemistry" theory. They had fairly high turnover and just missed the postseason. But I feel that the 1991 Twins represent two things: 1) regression to the mean, and 2) a team that got passed up. Another big hole we can poke in the "chemistry" theory is that it assumes that what was good enough to win one World Series is good enough to win the next. But this is absolutely false. In 1991, Minnesota's 95 wins (93 pW) were good enough to win the AL West. But in 1992, the Twins stayed roughly the same (falling to 90 Wins and 91 pW) while the Oakland A's won 96 games and shot past them to win the division. You may have to change your roster from year to year simply to keep up with the rest of the league.

So how does this reflect on the 2005 White Sox, which caused this study? The 2005 White Sox went 99-63, but their Pythagorean record was 91-71. This could be caused by Chicago's good bullpen (which can impact a team's Pythagorean record). Or it could mean that the Sox got really lucky in 2005 and are due for a fall in 2006. If I were Chicago GM Kenny Williams, I wouldn't risk it.
Even if the Sox do stay the same, is there any team that could zoom past them? I'd actually be surprised if the 2006 Cleveland Indians didn't zoom past the Sox. The 2005 Indians went 93-69, but their Pythagorean record was 96-66. This can't be explained away by bullpens; the Indians and Sox both have good relievers. What does this mean? It means that the Indians were a better team than the White Sox in 2005. And, considering their respective records, it would indicate that the White Sox will be the 2nd-best team going into 2006. So if I'm Kenny Williams, staring at a superior Cleveland team, am I going to sit still and hope for more extremely good luck (8 wins over your Pythagorean record is quite lucky)? Or am I going to go out and try to turn my 91-win team into a 96- or 97-win team? The latter choice is obvious, although the obvious can be surprisingly difficult for many baseball people to fathom.
There is, unfortunately, another factor here: public opinion. Kenny Williams may make the very shrewd move to keep his team together, thereby making him a fan favorite and getting the media (Peter Gammons included) to applaud him. And when the team inevitably falters next year, no one will blame Williams. He can point to the players and say, "I left you guys alone. I kept the same team that won the World Series last year, so it's Ozzie Guillen's fault and the players' fault if they can't repeat." And that explanation will be good enough for the fans and most of the media, with the notable exception of Rob Neyer, Baseball Prospectus, and the sabermetric minority. Many baseball executives (and managers, etc.) would much rather make the safe choice, the choice that will shield them from blame and protect their jobs, than to make the right choice. Because baseball rewards those who act according to tradition and custom, regardless of how effective those actions may be.
So let's hope, for Chicago's sake, that Williams is intelligent (and unselfish) enough to know what he is facing. Let's hope he realizes that Jose Contreras and possibly Jon Garland are due for a slump in 2006. Let's hope he realizes that postseason heroics aren't going to magically turn Joe Crede into a good hitter (303 career OBP in an AL hitter's park). If he knows these things, the Sox will be contenders. If not, the Indians can go ahead and put on those 2006 AL Central Champion t-shirts.