Seattle Mariners 2026

Stats, refreshed the morning after every game

Published

August 23, 2026

Last refreshed: August 23, 2026 at 6:26 AM PDT

This site treats a baseball season like a bioinformatics experiment: players are samples, their offensive stats are features, and we look for the hitters and pitchers behaving differently from the rest. The numbers come from FanGraphs and the MLB Stats API and are re-rendered automatically the morning after each Mariners game.

Read the full differential analysis →

Stat glossary

New to the sabermetric alphabet soup? Here’s what every stat on this page means.

wOBA — weighted on-base average: One number that captures a hitter’s total offensive value. Instead of treating every time on base equally (like OBP does), it weights each outcome — walk, single, double, triple, home run — by how much that event is actually worth in run-scoring terms. It’s scaled to look like OBP, so league average is about .320, a good hitter is around .370, and an elite one pushes .400+. Higher is better.
xwOBA — expected weighted on-base average: What a hitter should have produced based on how they hit the ball, not whether it found a glove. Statcast feeds the exit velocity and launch angle of every batted ball (plus sprint speed on grounders) into a model that returns the wOBA a typical hitter earns on contact like that. Because it strips out luck, defense, and ballpark, comparing it to actual wOBA is revealing: wOBA − xwOBA above zero means a hitter is out-producing their contact quality (riding hot, likely to cool off), while below zero flags strong contact that hasn’t paid off yet (an upside regression candidate).
BABIP — batting average on balls in play: Batting average counting only balls put in play — home runs and strikeouts are excluded. League average sits near .300, and sustained extremes are rare, so a sky-high BABIP usually signals good luck or favorable sequencing that regresses, and a very low one signals the opposite.
ERA / xERA — earned run average vs. expected: ERA is the earned runs a pitcher allows per nine innings — the classic bottom-line number. xERA is the contact-quality-based expectation, derived the same way as xwOBA. ERA − xERA separates skill from luck: a negative gap means a pitcher has allowed fewer runs than their stuff predicts (possibly lucky), a positive gap the reverse.
Hard-hit %, exit velocity & launch angle — the batted-ball inputs: Exit velocity is how fast the ball leaves the bat (mph); hard-hit % is the share of batted balls hit at 95 mph or more — a proxy for raw contact quality. Launch angle is the vertical angle off the bat: too low is a grounder, too high is a pop-up, and the productive line-drive window sits in between. Together these are the ingredients xwOBA is built from.
PA / IP — sample size: Plate appearances (hitters) and innings pitched (pitchers). They aren’t performance stats — they tell you how much to trust the rest. Early in the season small samples swing wildly, so a gaudy line on 30 PA means far less than the same line on 300.

Who’s hot, who’s not

Hard contact vs results. Does hitting the ball hard translate into production? The dashed line is the league-wide trend; the grey band is its 95% confidence interval (how sure we are of that average relationship — not a per-player normal range). Players well above the line are out-producing their contact quality (lucky or crafty); well below, under-producing (unlucky or poor launch angles).

Code

chart_hard_hit(batting)

BABIP — luck on balls in play. League average is ~.300; sustained extremes are rare, so red bars tend to cool off and navy bars tend to warm up.

Code

chart_babip(batting)

Pitching: ERA vs the peripherals. Negative bars mean a pitcher has allowed fewer runs than his strikeouts, walks, and contact quality would predict.

Code

chart_pitching_luck(pitching)

The numbers

The biggest gaps between what hitters have actually produced (wOBA) and what their batted-ball quality says they should have produced (xwOBA).

Code

glance <- batting |>
  dplyr::transmute(
    Player = player,
    Pos    = pos,
    PA     = pa,
    wOBA   = round(woba, 3),
    xwOBA  = round(xwoba, 3),
    Diff   = round(woba - xwoba, 3)
  ) |>
  dplyr::arrange(dplyr::desc(Diff))

over  <- dplyr::slice_max(glance, Diff, n = 5, with_ties = FALSE)
under <- dplyr::slice_min(glance, Diff, n = 5, with_ties = FALSE) |>
  dplyr::arrange(Diff)

Code

knitr::kable(over, row.names = FALSE)
knitr::kable(under, row.names = FALSE)

Running hot — over-performing

Player	Pos	PA	wOBA	xwOBA	Diff
Brendan Donovan	3B	159	0.349	0.314	0.034
Miles Mastrobuoni	UNK	30	0.202	0.184	0.018
Randy Arozarena	LF	526	0.366	0.350	0.016
Colt Emerson	UNK	241	0.256	0.248	0.008
Victor Robles	RF	141	0.285	0.281	0.004

Running cold — under-performing

Player	Pos	PA	wOBA	xwOBA	Diff
Patrick Wisdom	UNK	44	0.172	0.265	-0.093
Connor Joe	UNK	45	0.274	0.337	-0.063
Rob Refsnyder	UNK	129	0.207	0.254	-0.047
Julio Rodríguez	CF	517	0.316	0.348	-0.031
Jhonny Pereda	C	79	0.299	0.330	-0.031

Pitching staff — ERA − xERA

Code

pitching |>
  dplyr::transmute(
    Player = player,
    Role   = role,
    IP     = ip,
    ERA    = round(era, 2),
    xERA   = round(xera, 2),
    `ERA − xERA` = round(era_minus_xera, 2)
  ) |>
  dplyr::arrange(`ERA − xERA`) |>
  knitr::kable(row.names = FALSE)

Player	Role	IP	ERA	xERA	ERA − xERA
Matt Brash	RP	16.2	0.54	4.19	-3.65
Cole Wilcox	SP	27.0	4.00	5.44	-1.44
Gabe Speier	SP	34.0	2.65	4.01	-1.37
Eduard Bazardo	SP	58.0	3.10	4.42	-1.32
Nick Davila	SP	26.1	4.10	4.65	-0.54
Emerson Hancock	SP	133.2	3.37	3.87	-0.50
Logan Gilbert	SP	150.1	3.35	3.69	-0.33
José A. Ferrer	SP	52.2	3.59	3.56	0.03
Cooper Criswell	SP	30.2	3.52	3.38	0.14
George Kirby	SP	140.0	4.18	4.00	0.18
Bryan Woo	SP	138.1	4.10	3.56	0.54
Bryce Miller	SP	89.2	3.71	3.04	0.68
Michael Rucker	SP	29.1	4.30	3.24	1.05
Andrés Muñoz	SP	47.1	4.37	3.28	1.09

Player pages on Baseball Savant

Want the pitch-by-pitch Statcast detail behind these numbers? Each player links to their Baseball Savant page.

Code

batting |>
  dplyr::arrange(dplyr::desc(pa)) |>
  savant_bullets("hitting")
pitching |>
  dplyr::arrange(dplyr::desc(ip)) |>
  savant_bullets("pitching")

How this updates: a scheduled GitHub Actions workflow checks each morning whether the Mariners finished a game the night before. If so, it re-fetches the stats, re-renders this site, and deploys it to Vercel. Off-days are skipped, and the synthetic fallback is never published.

--- title: "Seattle Mariners 2026" subtitle: "Stats, refreshed the morning after every game" date: today page-layout: full code-fold: true --- ```{r setup, include=FALSE} source(here::here("R", "00_setup.R")) source(here::here("R", "charts.R")) # In CI the data is fetched once before the render. When this page is rendered # standalone the .rds files may not exist yet — fetch then. if (!file.exists(here("data", "batting_2026.rds"))) { source(here::here("R", "01_fetch_data.R")) } batting <- readRDS(here("data", "batting_2026.rds")) pitching <- readRDS(here("data", "pitching_2026.rds")) refreshed <- format(Sys.time(), tz = "America/Los_Angeles", format = "%B %e, %Y at %l:%M %p %Z") # Note when the underlying numbers came from the synthetic fallback rather than # the live feed, so the page is never silently wrong. src <- tryCatch(readLines(here("data", "data_source.txt"), n = 1), error = function(e) "unknown") ``` ::: {.refreshed-badge} Last refreshed: `r refreshed` ::: ```{r live-warning, echo=FALSE, results='asis'} if (!identical(src, "live")) { cat("> ⚠️ **Heads up:** these numbers are from the synthetic fallback — the live", "FanGraphs / MLB Stats API feed was unavailable at render time.\n") } ``` This site treats a baseball season like a bioinformatics experiment: players are *samples*, their offensive stats are *features*, and we look for the hitters and pitchers behaving differently from the rest. The numbers come from FanGraphs and the MLB Stats API and are re-rendered automatically the morning after each Mariners game. [**Read the full differential analysis →**](/analysis) ## Stat glossary New to the sabermetric alphabet soup? Here's what every stat on this page means. ::: {.stat-glossary} <dl> <dt>wOBA — weighted on-base average</dt> <dd>One number that captures a hitter's total offensive value. Instead of treating every time on base equally (like OBP does), it weights each outcome — walk, single, double, triple, home run — by how much that event is actually worth in run-scoring terms. It's scaled to look like OBP, so league average is about **.320**, a good hitter is around **.370**, and an elite one pushes **.400+**. Higher is better.</dd> <dt>xwOBA — expected weighted on-base average</dt> <dd>What a hitter should have produced based on how they hit the ball, not whether it found a glove. Statcast feeds the exit velocity and launch angle of every batted ball (plus sprint speed on grounders) into a model that returns the wOBA a typical hitter earns on contact like that. Because it strips out luck, defense, and ballpark, comparing it to actual wOBA is revealing: wOBA − xwOBA above zero means a hitter is out-producing their contact quality (riding hot, likely to cool off), while below zero flags strong contact that hasn't paid off yet (an upside regression candidate).</dd> <dt>BABIP — batting average on balls in play</dt> <dd>Batting average counting only balls put in play — home runs and strikeouts are excluded. League average sits near .300, and sustained extremes are rare, so a sky-high BABIP usually signals good luck or favorable sequencing that regresses, and a very low one signals the opposite.</dd> <dt>ERA / xERA — earned run average vs. expected</dt> <dd>ERA is the earned runs a pitcher allows per nine innings — the classic bottom-line number. xERA is the contact-quality-based expectation, derived the same way as xwOBA. ERA − xERA separates skill from luck: a negative gap means a pitcher has allowed fewer runs than their stuff predicts (possibly lucky), a positive gap the reverse.</dd> <dt>Hard-hit %, exit velocity & launch angle — the batted-ball inputs</dt> <dd>Exit velocity is how fast the ball leaves the bat (mph); hard-hit % is the share of batted balls hit at 95 mph or more — a proxy for raw contact quality. Launch angle is the vertical angle off the bat: too low is a grounder, too high is a pop-up, and the productive line-drive window sits in between. Together these are the ingredients xwOBA is built from.</dd> <dt>PA / IP — sample size</dt> <dd>Plate appearances (hitters) and innings pitched (pitchers). They aren't performance stats — they tell you how much to trust the rest. Early in the season small samples swing wildly, so a gaudy line on 30 PA means far less than the same line on 300.</dd> </dl> ::: ## Who's hot, who's not **Hard contact vs results.** Does hitting the ball hard translate into production? The dashed line is the league-wide trend; the grey band is its 95% confidence interval (how sure we are of that *average* relationship — not a per-player normal range). Players well *above* the line are out-producing their contact quality (lucky or crafty); well *below*, under-producing (unlucky or poor launch angles). ```{r hard-hit, fig.width = 9, fig.height = 6} chart_hard_hit(batting) ``` **BABIP — luck on balls in play.** League average is ~.300; sustained extremes are rare, so red bars tend to cool off and navy bars tend to warm up. ```{r babip, fig.width = 9, fig.height = 5} chart_babip(batting) ``` **Pitching: ERA vs the peripherals.** Negative bars mean a pitcher has allowed fewer runs than his strikeouts, walks, and contact quality would predict. ```{r pitching-luck, fig.width = 9, fig.height = 5} chart_pitching_luck(pitching) ``` ## The numbers The biggest gaps between what hitters have actually produced (**wOBA**) and what their batted-ball quality says they *should* have produced (**xwOBA**). ```{r glance} glance <- batting |> dplyr::transmute( Player = player, Pos = pos, PA = pa, wOBA = round(woba, 3), xwOBA = round(xwoba, 3), Diff = round(woba - xwoba, 3) ) |> dplyr::arrange(dplyr::desc(Diff)) over <- dplyr::slice_max(glance, Diff, n = 5, with_ties = FALSE) under <- dplyr::slice_min(glance, Diff, n = 5, with_ties = FALSE) |> dplyr::arrange(Diff) ``` ::: {layout-ncol="2"} ::: {} **Running hot — over-performing** ```{r over} knitr::kable(over, row.names = FALSE) ``` ::: ::: {} **Running cold — under-performing** ```{r under} knitr::kable(under, row.names = FALSE) ``` ::: ::: **Pitching staff — ERA − xERA** ```{r pitching-glance} pitching |> dplyr::transmute( Player = player, Role = role, IP = ip, ERA = round(era, 2), xERA = round(xera, 2), `ERA − xERA` = round(era_minus_xera, 2) ) |> dplyr::arrange(`ERA − xERA`) |> knitr::kable(row.names = FALSE) ``` ## Player pages on Baseball Savant Want the pitch-by-pitch Statcast detail behind these numbers? Each player links to their [Baseball Savant](https://baseballsavant.mlb.com) page. ```{r savant-helper, include=FALSE} # Emit a markdown bullet list of Savant links for a data frame that has # `player` + `mlbam_id`. Players without an id are skipped. savant_bullets <- function(df, kind) { if (!"mlbam_id" %in% names(df)) return(invisible()) urls <- savant_url(df$player, df$mlbam_id, kind) ok <- !is.na(urls) cat(paste0("- [", df$player[ok], "](", urls[ok], ")"), sep = "\n") cat("\n") } ``` ::: {layout-ncol="2"} ::: {.savant-list} ### Hitters ```{r savant-hitters, results='asis'} batting |> dplyr::arrange(dplyr::desc(pa)) |> savant_bullets("hitting") ``` ::: ::: {.savant-list} ### Pitchers ```{r savant-pitchers, results='asis'} pitching |> dplyr::arrange(dplyr::desc(ip)) |> savant_bullets("pitching") ``` ::: ::: --- *How this updates: a scheduled GitHub Actions workflow checks each morning whether the Mariners finished a game the night before. If so, it re-fetches the stats, re-renders this site, and deploys it to Vercel. Off-days are skipped, and the synthetic fallback is never published.*