from google.cloud import bigquery
import pandas as pd
client = bigquery.Client()
batting_query = """
SELECT
league_name, player_name, team_name,
homeRuns, CAST(ops AS FLOAT64) AS ops, CAST(avg AS FLOAT64) AS avg,
plateAppearances,
RANK() OVER (PARTITION BY league_name ORDER BY CAST(ops AS FLOAT64) DESC) AS ops_rank
FROM `maydaystats.mlb_season_stats.batting_latest`
WHERE plateAppearances >= 150
ORDER BY league_name, ops_rank
"""
# Every qualified hitter, ranked; used below for specific-player lookups
# (some storyline players, like Guerrero, rank well outside the top 15).
batting_all = client.query(batting_query).to_dataframe()
batting = batting_all[batting_all["ops_rank"] <= 15].copy()
pitching_query = """
SELECT
league_name, player_name, team_name,
CAST(era AS FLOAT64) AS era, strikeOuts, CAST(whip AS FLOAT64) AS whip,
wins, CAST(inningsPitched AS FLOAT64) AS ip
FROM `maydaystats.mlb_season_stats.pitching_latest`
WHERE CAST(inningsPitched AS FLOAT64) >= 60
QUALIFY RANK() OVER (PARTITION BY league_name ORDER BY era ASC) <= 15
ORDER BY league_name, era ASC
"""
pitching = client.query(pitching_query).to_dataframe()
# The full announced rosters (starters, pitchers, and reserves), plus how
# each player got there. Source: MLB.com's official roster announcement
# and the subsequent replacement announcements for Vladimir Guerrero Jr.
all_stars = {
# American League
"Shea Langeliers": "Elected starter", "Vladimir Guerrero Jr.": "Elected starter (withdrew, injury)",
"Ernie Clement": "Elected starter", "Junior Caminero": "Elected starter",
"Bobby Witt Jr.": "Elected starter", "Mike Trout": "Elected starter",
"Byron Buxton": "Elected starter", "Aaron Judge": "Elected starter",
"Yordan Alvarez": "Elected starter",
"Bryan Baker": "MLB's choice", "Dylan Cease": "Player-elected",
"Aroldis Chapman": "Player-elected", "Jacob Latz": "MLB's choice",
"Parker Messick": "Player-elected", "Drew Rasmussen": "Reserve",
"Joe Ryan": "Player-elected", "Cam Schlittler": "Player-elected",
"Cade Smith": "Player-elected", "Ranger Suarez": "MLB's choice",
"Louis Varland": "Player-elected", "Michael Wacha": "MLB's choice",
"Dillon Dingler": "Player-elected", "Adley Rutschman": "MLB's choice",
"Travis Bazzana": "Player-elected", "Nick Kurtz": "Reserve -> now starting (Guerrero replacement)",
"Kevin McGonigle": "Player-elected", "Ben Rice": "MLB's choice",
"Miguel Vargas": "Player-elected", "Randy Arozarena": "Player-elected",
"Cody Bellinger": "Player-elected", "Riley Greene": "Player-elected",
"Yandy Díaz": "Player-elected", "Justin Verlander": "Legend Pick",
"Willson Contreras": "Roster replacement (for Guerrero)",
# National League
"Drake Baldwin": "Elected starter", "Freddie Freeman": "Elected starter",
"Ozzie Albies": "Elected starter", "Max Muncy": "Elected starter",
"CJ Abrams": "Elected starter", "Brandon Marsh": "Elected starter",
"Juan Soto": "Elected starter", "Andy Pages": "Elected starter",
"Shohei Ohtani": "Elected starter",
"Chase Burns": "Player-elected", "Jhoan Duran": "Player-elected",
"Raisel Iglesias": "Player-elected", "Max Meyer": "MLB's choice",
"Mason Miller": "Player-elected", "Jacob Misiorowski": "Player-elected",
"Eduardo Rodriguez": "MLB's choice", "Chris Sale": "Player-elected",
"Cristopher Sánchez": "Player-elected", "Paul Skenes": "Player-elected",
"Logan Webb": "MLB's choice", "Yoshinobu Yamamoto": "MLB's choice",
"William Contreras": "Player-elected", "Hunter Goodman": "MLB's choice",
"Luis Arraez": "Player-elected", "Bryce Harper": "Legend Pick",
"Otto Lopez": "Player-elected", "Matt Olson": "Player-elected",
"Sal Stewart": "Player-elected", "Corbin Carroll": "Player-elected",
"Pete Crow-Armstrong": "Player-elected", "Jordan Walker": "MLB's choice",
"James Wood": "Player-elected", "Kyle Schwarber": "Player-elected",
}
batting_all["selection"] = batting_all["player_name"].map(all_stars).fillna("Not selected")
batting["selection"] = batting["player_name"].map(all_stars).fillna("Not selected")
pitching["selection"] = pitching["player_name"].map(all_stars).fillna("Not selected")Fan Vote vs. the Numbers: Grading the 2026 All-Star Rosters
Baseball
Featured
All-Star Game
The starters are elected by fans, and that’s fine. Here’s who the stat lines say actually earned it.
The 2026 All-Star Game is July 14 in Philadelphia, and the rosters are set. Starters are elected by fans in a multi-phase vote, most reserves are added by a mix of player ballot and each league’s coaching/front office staff, and the Commissioner gets one “Legend Pick” per league on top of all that. That process is built around name recognition and storylines as much as this season’s stats, and there’s nothing wrong with that: the All-Star Game is partly a popularity contest by design, and always has been.
But it also means “elected starter” and “best player at the position this year” aren’t the same claim. With our own season stats pipeline now tracking every hitter and pitcher’s numbers daily (see the pipeline behind this analysis for the batting/pitching side of things), it’s worth actually checking: who’s on the roster because the numbers say so, and who’s there for other reasons?
Who actually leads the league
Two lists: OPS leaders (minimum 150 plate appearances) and ERA leaders (minimum 60 innings), split by league, with each name tagged by whether (and how) they made an All-Star roster.
import matplotlib.pyplot as plt
from matplotlib.patches import Patch
al_bat = batting[batting["league_name"] == "AL"].head(12).iloc[::-1]
colors = ["#2c3e50" if s != "Not selected" else "#c0392b" for s in al_bat["selection"]]
legend_handles = [
Patch(color="#2c3e50", label="On an All-Star roster"),
Patch(color="#c0392b", label="Not selected"),
]
fig, ax = plt.subplots(figsize=(8, 6.5))
ax.barh(al_bat["player_name"], al_bat["ops"], color=colors)
ax.set_xlabel("OPS")
ax.set_title("AL OPS Leaders (min. 150 PA)")
ax.spines[["top", "right"]].set_visible(False)
ax.legend(
handles=legend_handles, loc="upper center", bbox_to_anchor=(0.5, -0.1),
ncol=2, frameon=False,
)
plt.tight_layout()
plt.show()
nl_bat = batting[batting["league_name"] == "NL"].head(12).iloc[::-1]
colors = ["#2c3e50" if s != "Not selected" else "#c0392b" for s in nl_bat["selection"]]
fig, ax = plt.subplots(figsize=(8, 6.5))
ax.barh(nl_bat["player_name"], nl_bat["ops"], color=colors)
ax.set_xlabel("OPS")
ax.set_title("NL OPS Leaders (min. 150 PA)")
ax.spines[["top", "right"]].set_visible(False)
ax.legend(
handles=legend_handles, loc="upper center", bbox_to_anchor=(0.5, -0.1),
ncol=2, frameon=False,
)
plt.tight_layout()
plt.show()
Red bars are the players leading their league in OPS who aren’t on either All-Star roster in any capacity, elected or otherwise. Most of the list is dark blue, which is the actual headline: the system mostly works. But “mostly” is doing some work in that sentence, and it’s worth looking at where it didn’t.
The starters who back it up
elected = batting[batting["selection"] == "Elected starter"].sort_values("ops", ascending=False)
top_elected = elected.head(5)Start with the good news. 9 of the 18 elected position-player starters show up in the top 15 of their league in OPS, a solid hit rate for a system built on votes rather than end-of-July stat lines. The clearest examples that fans got exactly right:
top_starters_tbl = top_elected[["player_name", "team_name", "league_name", "homeRuns", "ops", "avg"]].reset_index(drop=True)
top_starters_tbl.index += 1
top_starters_tbl| player_name | team_name | league_name | homeRuns | ops | avg | |
|---|---|---|---|---|---|---|
| 1 | Yordan Alvarez | Houston Astros | AL | 29 | 1.030 | 0.310 |
| 2 | Juan Soto | New York Mets | NL | 21 | 0.993 | 0.297 |
| 3 | Shohei Ohtani | Los Angeles Dodgers | NL | 20 | 0.939 | 0.290 |
| 4 | Junior Caminero | Tampa Bay Rays | AL | 27 | 0.913 | 0.277 |
| 5 | Aaron Judge | New York Yankees | AL | 17 | 0.908 | 0.248 |
Yordan Alvarez leads the entire AL in OPS, and Juan Soto leads the entire NL, both elected starters, both correct. Shohei Ohtani, elected as the NL’s starting DH, is right behind Soto. This is what the selection process is supposed to produce, and for most of both rosters, it does.
Where fan vote and merit actually diverged
guerrero = batting_all[batting_all["player_name"] == "Vladimir Guerrero Jr."].iloc[0]
kurtz = batting_all[batting_all["player_name"] == "Nick Kurtz"].iloc[0]
contreras = batting_all[batting_all["player_name"] == "Willson Contreras"].iloc[0]The clearest case this year is Toronto’s Vladimir Guerrero Jr., elected as the AL’s starting first baseman on name recognition and a strong track record, but this season he’s hitting 0.262 with 5 home runs and a 0.694 OPS, ranking 87th in the AL, nowhere near the top 15. He withdrew from the game with a lower back injury, and what happened next made the point better than any stat table could: the AL didn’t just plug in whoever was next alphabetically. Nick Kurtz, already on the roster as a reserve, was elevated to start at first base after receiving the most votes among AL first basemen on the players’ ballot, backed by a 0.902 OPS and 20 home runs of his own. And Willson Contreras, who had a 0.921 OPS and 20 home runs largely outside the national conversation, was added to fill the vacated roster spot.
Guerrero’s original selection wasn’t wrong, exactly: he’s a six-time All-Star with a real track record, and fan voting is allowed to weigh that. But when an injury forced the question, the mechanisms built to handle exactly this (player ballots, front-office and player-elected reserve picks) pulled in two players with genuinely stronger seasons. That’s the system working as a whole, even when the first step (the fan vote) didn’t land on the best hitter at the position this year.
The two Legend Picks, and why the exception exists
harper = batting_all[batting_all["player_name"] == "Bryce Harper"].iloc[0]Commissioner Rob Manfred’s “Legend Pick” is a one-per-league honor for a notable veteran, granted independent of this year’s stats rather than as a reward for them. The two picks this year show why that exception is worth having, for two very different reasons.
Philadelphia’s Bryce Harper got the NL nod, and it’s worth noting he barely needed it: at 0.870 OPS with 20 home runs, he ranks 16th in the NL among hitters with at least 150 plate appearances, one spot outside the group that gets in purely on merit, and squarely in the same conversation as the players who did. The exception mechanism and the stat line landed in almost the same place this time.
Detroit’s Justin Verlander, the AL’s Legend Pick, is the opposite case, and the clearer illustration of what the exception is actually for: one start all season, a 0-1 record, and a 12.27 ERA over 3.2 innings, having not taken the mound since March 30 because of hip and hamstring injuries. Nobody’s arguing he earned a roster spot on this year’s numbers, and the selection isn’t pretending otherwise. Verlander announced this week that 2026 will be his final season, and this is his tenth career All-Star selection: a stats-independent sendoff for a three-time Cy Young winner, exactly the case the rule was written for.
Who the numbers left off entirely
batting_snubs_tbl = batting[batting["selection"] == "Not selected"][
["player_name", "team_name", "league_name", "homeRuns", "ops", "avg"]
].reset_index(drop=True)
batting_snubs_tbl.index += 1
batting_snubs_tbl| player_name | team_name | league_name | homeRuns | ops | avg | |
|---|---|---|---|---|---|---|
| 1 | Munetaka Murakami | Chicago White Sox | AL | 20 | 0.938 | 0.240 |
| 2 | Dominic Canzone | Seattle Mariners | AL | 15 | 0.884 | 0.269 |
| 3 | Jonathan Aranda | Tampa Bay Rays | AL | 13 | 0.835 | 0.287 |
| 4 | Mickey Moniak | Colorado Rockies | NL | 15 | 0.924 | 0.279 |
| 5 | Luis García Jr. | Washington Nationals | NL | 20 | 0.894 | 0.291 |
| 6 | Andrew Vaughn | Milwaukee Brewers | NL | 2 | 0.893 | 0.319 |
| 7 | Jake Bauers | Milwaukee Brewers | NL | 17 | 0.873 | 0.268 |
| 8 | Bryan Reynolds | Pittsburgh Pirates | NL | 14 | 0.872 | 0.281 |
pitching_snubs_tbl = pitching[pitching["selection"] == "Not selected"][
["player_name", "team_name", "league_name", "era", "strikeOuts", "whip", "wins"]
].reset_index(drop=True)
pitching_snubs_tbl.index += 1
pitching_snubs_tbl| player_name | team_name | league_name | era | strikeOuts | whip | wins | |
|---|---|---|---|---|---|---|---|
| 1 | Sonny Gray | Boston Red Sox | AL | 2.61 | 82 | 1.10 | 10 |
| 2 | Nick Martinez | Tampa Bay Rays | AL | 2.61 | 61 | 1.13 | 7 |
| 3 | Casey Mize | Detroit Tigers | AL | 2.64 | 72 | 0.98 | 4 |
| 4 | Shane McClanahan | Tampa Bay Rays | AL | 2.83 | 82 | 1.13 | 8 |
| 5 | Walbert Ureña | Los Angeles Angels | AL | 2.88 | 78 | 1.32 | 5 |
| 6 | Tarik Skubal | Detroit Tigers | AL | 3.06 | 84 | 0.95 | 5 |
| 7 | J.T. Ginn | Athletics | AL | 3.10 | 86 | 1.22 | 7 |
| 8 | Payton Tolle | Boston Red Sox | AL | 3.14 | 80 | 1.07 | 5 |
| 9 | Keider Montero | Detroit Tigers | AL | 3.15 | 60 | 0.99 | 5 |
| 10 | Logan Gilbert | Seattle Mariners | AL | 3.19 | 114 | 0.95 | 7 |
| 11 | Ben Brown | Chicago Cubs | NL | 1.85 | 65 | 0.94 | 4 |
| 12 | Zack Wheeler | Philadelphia Phillies | NL | 2.28 | 98 | 0.91 | 9 |
| 13 | Justin Wrobleski | Los Angeles Dodgers | NL | 2.69 | 73 | 1.02 | 10 |
| 14 | Foster Griffin | Washington Nationals | NL | 2.77 | 109 | 1.02 | 10 |
| 15 | Shane Drohan | Milwaukee Brewers | NL | 2.97 | 61 | 1.24 | 4 |
| 16 | Kyle Harrison | Milwaukee Brewers | NL | 3.01 | 101 | 1.08 | 8 |
| 17 | Michael McGreevy | St. Louis Cardinals | NL | 3.01 | 66 | 1.10 | 4 |
A caveat before reading anything into these tables: our pipeline tracks performance, not health. It has no way to know a player was hurt, and several of the names above were, which changes what their absence actually means. Checked individually:
Chicago’s Munetaka Murakami ranks third in the entire AL in OPS, but he’s been out since May 29 with a grade 2 hamstring strain and only returned to the active roster today, with his first game back not until tomorrow, so he simply wasn’t available when the roster was set. The NL’s Ben Brown, third in ERA, is a cleaner version of the same story: he’s been out with a neck stress reaction since late June and isn’t expected back before late July at the earliest.
Two others are more of an in-between case: healthy right now, but with a big enough gap in their season to explain why the roster process looked past them. Colorado’s Mickey Moniak missed a month on the IL with an ankle injury (late May to late June) and has been excellent since returning, but that’s exactly the stretch when All-Star cases get made, and he simply wasn’t on the field for most of it. Detroit’s Tarik Skubal is similar: an elbow procedure cost him six weeks, and even though he’s been pitching like an ace since his early-June return, that missed time is a real chunk of a season’s worth of starts he doesn’t get back. Their current form is real, but so is the hole in their season that made them easy to set aside.
None of these four are evidence the process missed anything: their absences track health, not stats.
The players who were genuinely healthy and still left off make a cleaner case. Philadelphia’s Zack Wheeler, who publicly called it “kind of BS”, is fully healthy. His exclusion comes down to a scheduling rule, unrelated to how he’s pitched. He’s scheduled to start Philadelphia’s last game before the break, and by rule, a starter who pitches that game isn’t eligible to also pitch in the All-Star Game, regardless of how good his season has been. Boston’s Sonny Gray is the more straightforward case: healthy, 10-1, a 2.61 ERA, and left off anyway on a crowded pitching staff, though a Red Sox teammate’s injury (Ranger Suarez, to the IL) has already put him in the conversation for a roster spot before the game is played. Tampa Bay’s Jonathan Aranda, similarly, appears to be simply squeezed out on a Rays roster that already sent Junior Caminero and Yandy Díaz.
None of this is a scandal. Roster spots are limited, some of the best staffs in baseball only get to send two or three of their own, injuries take real players out of contention regardless of how they’re hitting, and a system built partly on fan enthusiasm and storylines was never going to perfectly match a system built purely on this year’s stat line. The interesting part is being able to point at exactly where, why, and by how much the two systems disagree.
Note
This post uses Quarto’s frozen execution (freeze: auto): the numbers above reflect mlb_season_stats’s snapshot as of whenever this was last rendered locally, not a live query on every page load. The underlying tables append a new snapshot daily rather than overwriting, so the same query run again next week will reflect that day’s standings instead.