The Mariners Still Look Like an All-Time Anomaly
The Mariners lost two of three to the Dodgers over the weekend. It wouldn’t be fair to say the series was an accurate representation of the Mariners’ season, but it works as a convenient caricature. On Friday, the Mariners lost to the Dodgers by ten. On Sunday, the Mariners lost to the Dodgers by eleven. On Saturday, the Mariners beat the Dodgers, by one, in the tenth inning, on a walk-off balk. The Mariners avoided a sweep, and, indeed, the Mariners actually still have a better record than the Dodgers do. Over the three games, though, the Mariners were outscored by a margin of 27-7. Sunday was the Mariners’ worst loss of the year.
It’s hardly new information that the Mariners’ winning percentage and their run differential don’t exactly match up. This has been true for a matter of months, and it partly helps to explain why the AL West is as close as it is. But before we all just collectively get used to something, we should take a step back so we can reexamine precisely what’s been going on. Although the Mariners have slipped out of playoff position, they’re still within striking distance of both the A’s and the Astros. The Mariners are 3.5 back in the wild-card hunt, despite a run differential of -42. The Rays are 7.5 back of the Mariners, with a run differential of +10. The Angels are 8.5 back of the Mariners, with a run differential of +39. The Twins are 11 back of the Mariners, with a run differential of -22. Every year, there are run differential overperformers and underperformers. Yet this is far more extreme than is typical.
The last time I took a look at the Mariners was July 3. Through play on July 3, the Mariners had baseball’s fourth-best record. However, they had baseball’s 11th-best run differential. It was easy, at that point, to say the Mariners were due for a bit of regression. In a sense, they have regressed. And yet the numbers still look bizarre. Since play on July 4, the Mariners have baseball’s seventh-worst record. They also have baseball’s very worst run differential, worse even than the terrible Marlins (and, by definition, everyone else). A team’s Pythagorean winning percentage is simply its winning-percentage estimate based on runs scored and runs allowed. Through July 3, the Mariners were beating their Pythagorean estimate by 109 points. Since July 4, as the Mariners have slumped, they’ve still beat their Pythagorean estimate by 83 points. You could say even the bad Mariners have overachieved.
I live to put baseball data in context, so I’m going to return to an image I embedded in that linked post above. This is updated through the present day. In this plot, you’re going to see every team-season since 1900, expressed in terms of overall winning percentage, and run differential per game. The 2018 Mariners are the dot in yellow. The data sample numbers 2,520 team-seasons.
This plot illustrates a fundamentally obvious baseball principle: Good teams tend to outscore their opponents. Even though it’s easy to understand how, in theory, a team’s winning percentage and run differential might diverge, in truth the former has an established tendency to follow the latter. I don’t think I need to cover that ground. You already get it! The linear relationship in that image is apparent. The Mariners stick out. They stick out in part because I put them in yellow, sure, but the 2018 Mariners have been legitimately weird. How weird? Here’s a table of the top ten Pythagorean overachievers since the turn of the previous century:
| Team | Year | Win% | Pythag% | Difference |
|---|---|---|---|---|
| Mariners | 2018 | 0.568 | 0.466 | 0.102 |
| Athletics | 1955 | 0.409 | 0.319 | 0.090 |
| Reds | 1981 | 0.611 | 0.523 | 0.088 |
| Cubs | 1907 | 0.709 | 0.623 | 0.086 |
| Rangers | 2016 | 0.586 | 0.505 | 0.081 |
| Tigers | 1905 | 0.516 | 0.435 | 0.081 |
| Diamondbacks | 2005 | 0.475 | 0.399 | 0.076 |
| Angels | 2008 | 0.617 | 0.543 | 0.074 |
| Dodgers | 1954 | 0.597 | 0.525 | 0.072 |
| Padres | 2017 | 0.438 | 0.366 | 0.072 |
The Mariners are still holding on to that first-place ranking. For something similar, here’s another top-ten table. For this one, I’ve taken the absolute values of the differences between winning percentage and Pythagorean estimated winning percentage. This way, underachievers and overachievers are grouped together.
| Team | Year | Win% | Pythag% | |Difference| |
|---|---|---|---|---|
| Mariners | 2018 | 0.568 | 0.466 | 0.102 |
| Yankees | 1931 | 0.614 | 0.705 | 0.091 |
| Mets | 1993 | 0.364 | 0.455 | 0.090 |
| Athletics | 1955 | 0.409 | 0.319 | 0.090 |
| Reds | 1981 | 0.611 | 0.523 | 0.088 |
| Cubs | 1907 | 0.709 | 0.623 | 0.086 |
| Browns | 1905 | 0.353 | 0.435 | 0.082 |
| Pirates | 1917 | 0.331 | 0.413 | 0.082 |
| Athletics | 1946 | 0.318 | 0.400 | 0.082 |
| Pirates | 1986 | 0.395 | 0.477 | 0.082 |
Since 1900, the biggest difference between a team’s winning percentage and its Pythagorean winning percentage has been 91 points. The 2018 Mariners are hanging around at 102 points, more than three-quarters of the way through the year. Based on our current projections, the Mariners are going to finish with a final difference of 78 points — still historically significant, if no longer the most extreme anomaly. But our projections don’t really attempt to figure out run-differential overachievers and underachievers. The Mariners don’t have to be *that* weird the rest of the way to look like one of the strangest teams that anyone’s ever seen.
In the first table, you might’ve spotted some teams from recent seasons. The 2017 Padres were a big Pythagorean overachiever, and so were the 2016 Rangers. As such, you might wonder if we’re seeing more divergence of late. The long and short of it: no. There’s no evidence to suggest that teams have now figured out ways to beat their run differentials. For every individual team-season, I found the absolute value of the difference between its winning percentage and its estimated winning percentage. Here are the average and median differences, broken down by decade:
Nothing there to be seen, I don’t think. It’s all held pretty steady. Sure, teams now operate in ways different from how they did before, but the goal has always been to get good, and getting good involves outscoring the other team more often. I don’t think anyone has unlocked the key to being an overachiever. I mean, yeah, the key is being better in the clutch, but that’s never been proven to be a sustainable skill. Perceived advantages can disappear in the blink of an eye.
We’re getting away from the Mariners, now, so let’s bring it back. Here’s what their season looks like, in terms of rolling winning percentage and rolling Pythagorean winning percentage:
After the second game of the season, the Mariners had a winning percentage of .500, and a Pythagorean winning percentage of .500. Since then, they’ve been an overachiever every single day, with little in the way of evident regression. In this last plot, I’m just going to show you the rolling difference between those two lines. I’ve also included a flat line at .091 — remember that the 1931 Yankees finished with a winning percentage that was 91 points removed from their Pythagorean winning percentage. That’s the present record for the biggest all-time gap.
Those are the Mariners. Their winning percentage has never once dropped below their Pythagorean winning percentage. The gap has been enormous since the third game of the year. The Mariners might not finish with the biggest gap of all time, and, in fact, they probably won’t finish with the biggest gap of all time, based on the projections, but they’re going to be close. Just a couple weeks ago, the Mariners had a gap of 88 points. It’s now up to that previously-noted 102, thanks in part to what happened against the Dodgers. Another series or two like that and the Mariners will be on a collision course for history.
We can already say that, historically, the 2018 Mariners appear unusual. The only thing we don’t really know is whether they’ll come out as the most unusual. It’s another thing to monitor over the season’s final several weeks. Because of how well the Mariners have hit in the clutch, and because of how good Edwin Diaz has been at closing games, the Mariners remain alive in the playoff hunt. Catching up to the next-best team is, of course, the highest priority. But even if the Mariners do wind up on the outside looking in, they could also end up as the most anomalous team in modern baseball history. Depending on how you look at it, that could be either a good thing or a bad thing. There’s no debate that it would be an interesting thing.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
As promising as the season began for the mariners, based on how well the a’s are playing and how good everyone else in the AL is, it looks like they’ll be on the outside looking in again. Makes me feel sad, knowing that this was probably the last real shot at the playoffs for this group, and it was looking so good at the beginning of the year.
Then again, they have Cano, Cruz and Gordon who are all key contributors and have all tested positive before. Don’t know how to feel about that.
Agree with the point in general, even though I’d disagree with Dee Gordon being a major contributor. He has been downright awful, and is starting to fall out favor and losing playing time.
And Colome to complete the PED foursome. My own feeling is their acquisitions were more due to DiPoto exploiting market conditions for undervalued performers. That they were all undervalued due to PED history, or rumored usage, was coincidental in itself. A business decision in every case. On one hand, it’s giving people a second chance. On the other,
it can perhaps be seen as pretty cold and calculated.
Or on the other hand it is pretty cold and calculated to not want to give people a second chance.
Really? Because Cruz and Cano were signed by Jack Z, not DiPoto, so there goes half of your foursome. And Cano’s PED suspension was years after being signed by the M’s, so clearly it did not affect his market value at time of signing (nor did anyone consider his price tag discounted in any way). Colome had a suspension in 2014 and later developed into a big name reliever – do you really think his trade price in 2018 was notably discounted explicitly because of the remote history of PED suspension?
I seriously fail to understand why so many people seem to think that this is Seattle’s last chance for the foreseeable future. Most of the key players are under control for at least two more years, with Cruz being the exception. I would argue that they’ve made progress getting to where they want to be, regardless of what happens this season.
The A’s and Astros aren’t going away, I get that, (the Angels will attempt to rebound as well) and Seattle has a ton of committed salary for 2018 (without Cruz). None of which makes things easy.
But It seems to me they simply need to reach a little deeper this coming offseason and find the dough for the last couple of pieces (and Cruz’s return). They’ll need an outfielder, preferably a center fielder, but perhaps they could cut corners a bit and target the international market with players such as Yusei Kikuchi (whom they’ve been scouting) and Tetsuto Yamada, who would allow them to use Gordon in more of a utility role and slide Cano over to first base. Getting rid of these sub .300 OBP bats from the regular lineup is as important as strengthening the rotation.
They’ll get some salary relief in 2020 with Felix off the books, would it really kill them to pay a little more for a year? Maybe they can get creative to shed Nicasio’s salary (among others) next season, but with so much control over the majority of the team, it would seem pointless to blow it up.
That said, I would hope they don’t shell out a ton of salary to keep the exact same cast and expect different results.
Edwin Diaz and Alex Colome are third and fourth in +WPA. Correctly utilizing a top-heavy bullpen can increase win probability dramatically when you need it, at the expense of your run differential when you don’t need it.
Top-heaviness of the bullpen shows up as part of the overall pitching Clutch stat for a team. The Mariners pitching staff is +6 clutch, 2nd only to the A’s +8.
(And the M’s have the highest team offensive clutch, also at +6)
I wonder if it is possible for us to take the decade analysis for the first table a little bit further (and by “us” I mean “you” as I am not smart enough to figure it out)? The table covers, basically, 12 decades but only 5 decades are represented in the data: the 1900s (1905, 1907), 1950s (1954, 1955), 1980s (1981), 2000s (2005, 2008) and 2010s (2016, 2017, 2018). Now, maybe this outcome is within reasonable expectations (I don’t know, that’s why I am asking) but maybe there was something unique going on in each of these decades to *cause* the outcome? As Tony Wisconsin alludes to above, we might be able to explain the recent divergences to concentration of high quality bullpen arms on a select few teams, but was there anything going on in those other decades that might have lead to a similar result?
They have had a very lopsided pitching staff in general which probably explains most of this. They have had some really good deep pitching shut down starters, and some utter dogs that have been blown out a lot, with a few fluky pitchers that sometimes pitch well and sometimes get blown out. They also had a very Jekyll and Hyde bullpen who had a few really good relievers to close games where they were leading, but a fairly mediocre mid inning/losing big pitchers who let blowouts get worse. I would think the combination of these two factors as well as a little timely hitting really spawned this.
Isn’t the easiest way to over perform Pythagoras to use position players as often as possible whenever your expected win percentage fell below 5% or something? You’d go from losing by 4-5 runs to maybe 8-15 runs! Mariners should jettison their non-Felix long relievers and replace them with Ryon Healy clones.
Jeff, what runs data and version of Pythag gives the 1955 A’s a .319 pythag? B-Ref data gives them a .343 pythag — based on 638 runs scored, 911 runs allowed, and a 1.83 exponent. That puts their excess W% at +.066 — still quite anomalous, but just the 19th biggest positive in modern history.
I’ve been running these numbers every few days ever since Seattle got off to such an anomalous start, so it really surprised me to see those ’55 A’s atop your list of past teams.
Some thoughts on this interesting piece:
— I find a better frame of reference when this stuff is expressed in wins per 162 games, rather than W%. Just an easier way to think of team-seasons, IMO.
— Seattle’s pace of almost +17 wins above pythag would be an all-time high (plus or minus). They’re on pace for 92-70, when pythag pegs them at 75-87.
— It’s also one of the most USEFUL overachievements, making them a contender despite lacking the usual criteria. Most big overachievements go for naught: Of the prior top 25 in wins above pythag, just 9 wound in the postseason (I’m including the ’81 Reds in that count).
— It’s worth noting that this year’s Astros are on the other side of historic-anomaly scoresheet. Their pace of 11.3 wins *below* pythag would be the 12th-biggest shortfall in modern history. Houston leads Seattle by 4.5 games, when pythag says the gap should be 26.5.
It would be a pretty Mariners thing to be the most overachieving team in history and still miss the playoffs. Last time smoke and mirrors conspired like this to make Seattle appear to be a contender, Bill Bavasi traded Adam Jones and Chris Tillman for Erik Bedard to try to get the team over what he assumed was a modestly-sized hump. Not sure the franchise ever recovered.
What ever happened to the BaseRuns standings? Simple pythag seems a bit crude these days. Wonder who the most over-/under-achieving in terms of team wins vs. team WAR are…
I went ahead and did the math for my query in the last post about biggest difference in team WAR vs team wins. I blended rWAR and fWAR then adjusted everyone to 162 games. I found setting the cutoff at 1900 meant a bunch of deadball and Federal League teams at the top of the list. If we start instead with the liveball era in 1920, then the Mariners are on pace to be the 26th most overachieving team since that date. Here are the top 25:
Jeff, I assume you didn’t use Base Runs because they didn’t tell as interesting of a story, even though they are closer to representing a team’s true talent level. Are they an all-time anomaly using BR to tell the story?
Ah, the poor 1981 Reds…best record in the NL, didn’t make the playoffs.
Just for fun, with Diaz logging yet another save Monday, I broke down Seattle’s record by margins of 3 or less vs. 4 or more:
— 3 or less: 59-26, .694 — MLB-high wins and W% , AL-high games.
— 4 or more: 13-28, .317 — #28 wins, #27 W%.
— Net in the “close” games: +.377
That last figure would be the biggest ever, by quite a bit.
Their .694 W% in “close” games would be 7th.
And their .317 W% in the bigger margins would be worst ever by a team with overall W% over .470.
81.9% of their wins are margins of 3 or less — 13th overall, and the highest by a “winning” team.
seems to not have lasted.
More on team W% splits by margins of 3 or less vs. 4 or more (subtracting the latter from the former to get “net close W%”):
— Houston has a very high negative “net close W%” of -.255 (38-38 in 3-or-less margins, 37-12 otherwise). That would be 20th-lowest ever. But another 2018 team is lower…
— The Dodgers are -.283 in “net close W%”: 38-48 in 3-or-less margins, 29-11 otherwise. That would be 6th-lowest ever.
Most very good teams are at least mildly negative in this freak stat; it’s a natural outflow of the pythagorean relationship. Last year’s Dodgers and Astros were both below -.100, as were 3 other playoff teams; 8 of 10 playoff teams were negative.
But last year’s Dodgers exceeded pythag, and were very good in close games — their 59-39 in margins of 3 or less was the #1 W% in MLB. This year’s team is now 38-48 in “close games,” the #22 W% and worst by far among contenders. And they’re 2nd-worst in pythag shortfall.
Takes a special breed to downvote facts and research, where no opinion is expressed.
While better than most on the internet, the comment section here isn’t quite on the High Heat Stats level of reasonability.
[tips cap] “Cap’n!”
It isn’t so soecial. There are a lot of people reading fangraphs who do that. They don’t like intelligent people who havr independent opinions.
They have certainly been an interesting team this year, at least, so that’s good.
Edwin Diaz has saved exactly 2/3 of Seattle’s wins, 48 of 72.
That 67% share would be #2 among 40-save seasons.
#1 is Bryan Harvey, saving 70% of team wins for the ’93 Marlins (45 of 64).
Then comes Gagne, 65% for the 2003 Dodgers (55/85 in his “perfect” season).
There’s a rather dire trend among those high on this list, though it may just show the general inconstancy of top closers:
— Harvey got hurt the next year and was done.
— Gagne had one more great year, then got hurt and was never much after.
— Joakim Soria (43/67 for the 2010 Royals) was never again very good.
— Mike Williams (46/72, 2002 Pirates) came out of nowhere and was done the next year.
— Addison Reed (40/63, 2013 White Sox) lost his closer job after one more year, though he’s had some good years setting up.
— Randy Myers (53/84, ’93 Cubs) struggled the next year, had just one more big year.
— Roberto Hernandez (43/69, ’99 Rays) had a 3.02 ERA through that year, 3.97 after.
It probably doesn’t mean much, beyond “pitching at all is not healthy, pitching a lot is worse.”
Diaz certainly is pitching a lot — on pace for 79 IP, 79 games and 70 games finished. All would set or match 10-year highs among 30-save men. And his pace isn’t slowing; he’s been in 9 of their last 15 games, 7 of their last 9. He’s had a big workload in the last 3 years, including 92 relief innings in his rookie year (majors+minors). And he’s just 24. It would be a shame if he got hurt due to Seattle’s against-all-odds pursuit.
Wins and run differential are zero-sum games (i.e. MLB as a whole has an equal number of runs scored and wins for and against), which means that if the Mariners exceed their projection, the rest of the league must net fall short of their projections.
A good way to compare seasons is to take the average gap for all teams between Win% and Proj%. I bet the result would closely match the all-time seasons for obvious reasons.
It’s the 2012 Orioles (29-9 in 1-run games) all over again.
After 135 games in 2012, the Orioles were beating their Pythag by 11 games, or .081. But they were closing the gap. By the time the season ended, they were “only” 10 games ahead of their Pythag. Smart, lucky teams leverage their luck by improving down the stretch. We should see the Mariners close the gap in September and finish at the bottom of – or out of – the top 10.
Of the 12 teams with the best records this year, the Mariners have the fewest runs scored AND the most runs allowed.
Hanging out with the 1907 Cubs and the 1931 Yankees should make the Mariners feel pretty good about themselves, I would think.
If the M’s finish 13.5 games behind and miss the playoffs, like those ’31 Yankees, I doubt they’ll feel warm and fuzzy.
I would guess that the most straightforward way to have a much better winning percentage than run differential would be to have a great closer, like diaz or treinen.
Eh, if it were that easy to figure it would be super easy to exploit. Should be easy enough to check though – check out Mo Rivera’s best years (pythag vs actual) or Kimbrel or Wagner or Gagne or Nen or Jansen, etc.
“Great bullpen” may correlate with exceeding pythag. “Great closer” by itself does not, since most modern closers are used inefficiently.
If you were specifically trying to exceed pythag, you would spend your resources most aggressively in the highest-leverage spots, including late-inning TIES. Even trailing by 1 can be more hi-lev than leading by 3, or even more.
Just one example — The 2008 Phillies (Lidge 41 for 41) fell 1 win short of Pythag.
My initial comment was unclear, so I will try again.
It does not make sense to compare 3/4 seasons to full seasons, for two reasons. Firstly, it is harder to sustain luck over a whole season the same way that it is harder to sustain luck for two games than it is for one.
More importantly, teams that are lucky for the first 2/3 or 3/4 of the season will usually try to upgrade from “lucky” to “good.”
Take a look at the 2012 Orioles. They were extremely lucky for the first 2/3 of the season, but then an interesting thing happened to them that will likely play out with the M’s. Namely, the Orioles made some deadline trades to upgrade around the edges (Scott Feldman, Nate McLouth) and called up Manny Machado (to replace a lousy 3B, I don’t even remember who). For the last two months of the season they were actually a good team. They even finished the season with a positive run differential. They still beat their Pythag by 10 wins over the course of the season, but it’s worth noting that they were beating their Pythag by 11 wins after 135 games (a Pythag-real Win% of over 8%, which would put them in the top half of this list).
The moral of the story is this: A lucky team might beat their Pythag Win% by 10% plus once in a century. A smart lucky team will bank 2/3 season worth of luck and upgrade in July and August for a postseason run. Of course, they may continue to be lucky AND good, but they are, by definition, relying less on smoke and mirrors.
It would be interesting to see a ranking of luckiest teams on July 31 or August 31, and then see how many of those fell off of the all-time list – even if they kept winning.
Interesting idea, but extremely hard to research. Too many moving targets.
For one thing, the trade deadline dates don’t correspond to an exact stage of the season for all teams even within the same year, much less across multiple years. Within the past 10 years, team games through July 31 range from 101 to 111.
And then, trades that may have been aimed at “upgrading from lucky to good” don’t all happen right around the deadline. Manny Machado was dealt on July 18 this year. CC Sabathia was dealt on July 7, 2008; Cliff Lee on July 9, 2010; etc.