It had to end sometime. After sustaining a perfect record and a staggering 142-37 scoring margin over more than three weeks of play, the Cleveland Indians finally lost Friday night, dropping a tight contest to the Kansas City Royals. It was their first loss after winning 22 straight games. Now that The Streak is over, Cleveland can go back to focusing on the playoffs like any contending team.
Just because the Indians can put their streak in the rearview mirror, though, doesn’t mean that we can’t dwell on it a little more. It wasn’t the major league record for consecutive wins — if we include unofficial ties in between victories, the 1916 New York Giants’ 26-game mark still reigns supreme. But we can compare the Indians’ streak to that Giants’ run and determine exactly how difficult baseball’s best winning streaks were in general. (And, because I can’t resist, compare the Indians’ accomplishments with winning streaks in the NBA.)
Depending on how you measure the streak’s likelihood, the chances of a team like Cleveland pulling off their streak might have been as low as 1 in 65,000.
To judge this, I compared all the MLB streaks to one another, assuming they were done by the same, generic contending team. I set up a simulation under which a team with a fixed Elo rating — our method for determining how good a team is at a given moment — would take a crack at the particular opponents1 faced by every real MLB team who had a winning streak2 of at least 18 games since 1901.
A few more technical details of the simulation: I gave all the teams the same fixed rating, 1560, which is the average Elo of a World Series participant since 1903, when the first modern Fall Classic was staged.3 For comparison’s sake, the Indians’ rating at the beginning of their streak was 1555. I also assumed the streaking team had a five-man starting rotation, with the team’s rotation slot for the initial game of the streak randomized.4 (This matters because a team that goes into a potential streak with its No. 5 starter is much less likely to get off on the right foot than a team putting its ace on the mound.)
After running the first round of simulations, here were the odds of our generic contending team pulling off each streak:
|YEAR||TEAM||STREAK LENGTH||AVG. OPP. ELO||AVG. WIN PROB.||GENERIC TEAM ODDS|
|1916||New York Giants||26||1493.4||65.2%||1 in 76,702|
|1953||New York Yankees||18||1518.0||58.6||16,752|
|1947||New York Yankees||19||1506.2||60.8||13,297|
|1906||Chicago White Sox||19||1507.4||61.4||11,642|
|1904||New York Giants||18||1471.4||66.4||1,691|
According to this model, the hardest streak still belonged to the 1916 Giants — which isn’t too surprising, since they won four more games in a row than the Indians. And sorry, Billy Beane: the 2002 “Moneyball” A’s also fall behind lesser streaks because of the weak opposition they faced during their streak. But another thing that stands out are the odds, which are much more favorable than if we simply ran them on a .500 team.
(We’ll have to leave the impressiveness of the Dodgers’ feat — winning 52 out of 61 earlier in the season — for another time.)
The difference is because a 1560 Elo squad is (by design) no ordinary .500 team. Our generic team is going to be predisposed to running off a stretch of games like this, which only makes sense — average teams don’t go on these kinds of tears. And our simulation teams only got hotter as they won — that is, a team’s rating is fixed at 1560 before the streak begins, but then it gains steam with each victory, making the odds of winning again higher.
But there’s another layer we can add to the simulation to make it more reflective of the conditions under which each streak was actually compiled. Most clubs didn’t use the five-man rotation, for instance, until the 1970s or early ‘80s; likewise, the best teams of the past used to be much stronger Elo-wise, making it more likely we’d see such a run of greatness earlier in baseball history. We can account for these wrinkles by assigning a four-man rotation to teams before 1980, and adjusting our generic team’s fixed rating to be slightly higher in the past than in later seasons.5 After re-running the numbers with these two tweaks, here’s an amended list of the most difficult streaks:
|YEAR||TEAM||STREAK LENGTH||AVG. OPP. ELO||AVG. WIN PROB.||GENERIC TEAM ODDS (ADJ.)|
|2017||Cleveland Indians||22||1496.6||60.9%||1 in 65,566|
|1916||New York Giants||26||1493.5||67.2||34,720|
|1953||New York Yankees||18||1518.0||59.2||13,895|
|1947||New York Yankees||19||1506.2||61.7||10,223|
|1906||Chicago White Sox||19||1507.5||63.9||5,313|
|1904||New York Giants||18||1471.4||68.9||860|
In a plot twist, the Indians’ streak now rises to the top — a function of being accomplished in an era of (theoretically) more parity and a higher chance for some scrub pitcher to mess the streak up thanks to a bigger rotation than older teams had.
So how does this stack up against notable streaks from another sport like, say, basketball? Using the same Elo-based method,6 I calculated the odds of a generic contending NBA team (with a 1660 Elo7) pulling off some of the longest streaks in pro basketball history. And even the most impressive streaks on the hardwood can’t compare with baseball’s hottest runs.
The longest winning streak in NBA history, the 1972 L.A. Lakers’ 33-game winning streak, would have a 1 in 720 chance of being accomplished by our generic contender. The Golden State Warriors’ 24-game streak to start the season a couple years ago raises the bar a bit, with a 1 in 1,879 chance of being achieved by a generic contender, since the Warriors faced a much more difficult slate of opponents. But even a streak as memorable as the Houston Rockets’ 22-gamer from 2008 seems weak (1 in 247 odds) when compared with the baseball streaks we looked at above.
Streaks are nice, but the Indians surely have another accolade in mind: the World Series trophy. As of now, we give them a 1 in 4 chance. Given what they just pulled off, doesn’t seem so hard, does it?