The Twins starting the season with a nine-game losing streak was not good. Good teams don't lose a lot of games and even more so don't lose them in long streaks. That got me wondering about how teams with losing streaks fared overall.
It was pointed out here that the Twins streak was more noteworthy because it started the season so it looks even worse. Baseball-Reference only allows searching from the beginning of the season if you want to include all teams over all years so that's the best quick investigations can do. In order to search throughout the season, some work needed to be done.
I processed every season in the modern era through last year and grouped all games together by their streaks. A stretch that went WLLWWW would be a one-game winning streak, a two-game losing streak, and a three-game winning streak. Some brief spot checking of last year's data showed the streaks were properly classified. I merged that with the season results for those seasons in order to associate how teams did given a streak of a certain length.
To start, I decided to look at the range of results for teams with at least a losing streak of every length. The full dataset contains every season since 1900. Here are the number of seasons that contained a losing streak of this length from 1900 through 2015.
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 23 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2117 | 2116 | 2102 | 1896 | 1469 | 1014 | 605 | 377 | 210 | 130 | 73 | 43 | 27 | 19 | 9 | 2 | 4 | 3 | 4 | 4 | 1 | 1 |
First thing that popped out to me was the 2117 seasons with a one-game losing streak but 2116 seasons with a two-game losing streak. I checked the data to find the season, looked it up and determined it was right. If you want to know which team it is, the answer is at the bottom.
I did not want to have to normalize the seasons to match 162-game seasons nor have to deal with seasons shortened by a strike so I processed only the 162-game seasons. Converting that to a box plot for a pretty picture gives us this.
The widths of the box plots represent the number of seasons. The total number of seasons in the data set are 1234. In tabular form it looks like the following.
| Losing streak length | Seasons with that streak | 95th percentile | 75th percentile | Median | 25th percentile | 5th percentile |
|---|---|---|---|---|---|---|
| 1 | 1234 | 109 | 89 | 81.0 | 73.0 | 50 |
| 2 | 1234 | 109 | 89 | 81.0 | 73.0 | 50 |
| 3 | 1232 | 109 | 89 | 81.0 | 72.5 | 50 |
| 4 | 1121 | 109 | 88 | 80.0 | 72.0 | 50 |
| 5 | 882 | 109 | 86 | 78.0 | 70.0 | 50 |
| 6 | 605 | 108 | 84 | 76.0 | 68.0 | 50 |
| 7 | 361 | 102 | 82 | 75.0 | 67.0 | 51 |
| 8 | 229 | 103 | 82 | 73.0 | 66.0 | 43 |
| 9 | 121 | 95 | 77 | 71.0 | 64.0 | 51 |
| 10 | 66 | 92 | 76 | 68.0 | 64.0 | 50 |
| 11 | 37 | 89 | 76 | 68.0 | 62.0 | 43 |
| 12 | 23 | 91 | 73 | 67.0 | 59.0 | 53 |
| 13 | 12 | 77 | 70 | 67.0 | 63.0 | 60 |
| 14 | 7 | 70 | 66 | 63.0 | 60.5 | 60 |
| 15 | 4 | 65 | 60 | 53.0 | 51.0 | 51 |
| 17 | 3 | 67 | 64 | 61.0 | 50.5 | 40 |
| 19 | 2 | 57 | 57 | 56.5 | 56.0 | 56 |
| 20 | 1 | 52 | 52 | 52.0 | 52.0 | 52 |
| 21 | 1 | 54 | 54 | 54.0 | 54.0 | 54 |
Nothing knowing except the Twins having a nine-game losing streak, we would expect them to finish with 71 wins. We do know more than that and the preseason projections were not as kind to project the Twins for 81 wins. Many of them were closer to the high 70s so somewhere between the median and 25th percentile marks seems a more likely scenario.
I did another check on what happens if the losing streak happens in the first quarter of the season but the results are very similar. The sample size also becomes a problem as you get into the longer streaks. So I am skipping that and moving on to postseason probabilities.
It was already known at the beginning of the season that reaching the postseason would be unlikely. Well, losing nine straight hurts that a lot.
Note the "0" does not mean did not lose at all but instead is a placeholder to mean the average for all seasons.
Good thing they avoided losing ten games instead of nine, unless they managed to stretch it to eleven games. Not much to say here other than don't lose a lot of games if you want to make the postseason.
In conclusion, don't expect good things this year. The team started out with marginal chances for reaching 81 wins and now they're trying to do it in 153 games.
Great stuff, Sean.
The 1985 Twins season had a Jeckle-and-Hyde 10-game winning streak and losing streak in it. As a matter of fact, the 10-game winning streak immediately followed a 9-game losing streak.
Thanks for all your work, sean. It is appreciated.
Seriously. This is great.
So ... you're saying there's a chance!
Three teams have done it before!
1. Does this also remove all of the seasons in the 162-game schedule years where one or two games were cancelled due to weather and not made up due to them not counting in the final standings?
2. How could there be 2117 seasons with one losing streak of one game? I don't believe there have ever been an odd number of teams in the majors. Shouldn't that be an even number, or was there one (or 3 or 5 or ...? ) teams that never had a one-game losing streak?
1. Not to my knowledge. The minimum number of games played in a 162-game schedule is 159. The Orioles played that number some years ago. It's a useful quick filter. I am pretty sure that is the right number but I did not exhaustively check.
2. A team season could have been filtered out. I am pretty sure the overall number of seasons is 2117 so that rules out a season that had zero one-game losing streaks.
Aside to you or anyone else: If you want the dataset (it's in CSV), let me know.
That bit on the only team to never have a two game losing streak is amazing trivia.
There's one stretch from July 2 to July 4 that looks like they only lost two games in a row (they won the night game on July 4). But July 3 was also a doubleheader and they lost both games by the same score. The night game was a walk-off loss too.
The problem with the postseason probability is it's not the same for all seasons for any team in general. The AL started out with 8 teams, meaning there was a 1 in 8 chance of winning the AL for each team. With the initial expansion, that became 1 in 10. Then it went to 1 in 6 when they added divisional play, then 1 in 7 in '77, then with wildcard and uneven divisions, chances weren't the same, since teams in AL West had a 1 in 4 chance of winning the division, teams in NL Central were 1 in 6 and all other teams were 1 in 5. Plus AL teams had a 1 in 11 chance of winning wildcard. With expanded wildcard, all teams are equal with 1 in 5 chance of winning division and 2 in 12 (or 1 in 6) chance of the wildcard. So the probabilities are all over the place plus most years were extremely low to begin with and now the probability of making the postseason for any team has never really been higher.
The 12.5% was never a problem because pre-1961 seasons were ignored. The 1961-68 seasons would bring down the probability but that's only eight seasons.
I noticed the average for the low end of losing streaks (1, 2, etc.) is 81 wins, but that seems high considering how long the teams played 154 game schedules. Shouldn't the mean season wins be around 79?
Including all seasons, which I did originally, gives a median of 79 games.
Awesome post, sean. Wish I'd seen it on Wednesday;; it must've been buried pretty quickly at the bottom of the current post tiles. Any way to sticky something like this for a day or two?
Thanks. There are the two featured spots but they're currently occupied with recent material that I am loathe to bump. The middle four posts are the four non-Cup posts with LTEs so it can stick around there for a little bit. Having an off day on Thursday helped so there's no game log with a lot of activity.
Geez, we just have too much content around here!
(I think it was you who voiced the desire for more baseball analysis – like the old place – posted around here? If so, way to get us off the schneid.)
Yes. I was in the process of running the numbers when I said that. I have some more ideas but it's going to be a while until it's all done.
maybe we need a designated tile spot for baseball analysis pieces?
D(esignated) H(analysis)?