The card on July 21, 2026 is narrow by league count but deep in one place. MLB offers 15 games; MLS and the WNBA are both dark. That concentration matters: with two leagues on break, the model's attention and our published picks route almost entirely through baseball, where starting pitching and travel are the primary levers. Our picks post about 30 minutes before each first pitch and settle afterward on our public, tamper-evident ledger.
MLB
Fifteen games span the slate, and the pitching lines give the model plenty to work with.
The cleanest verified matchup is in Boston. Baltimore visits the Red Sox at a 7:10 PM first pitch, with Kyle Bradish (6-9, 3.61 ERA, 1.35 WHIP, 106 strikeouts) opposing Ranger Suarez (0-0). Boston is listed as a -142 favorite with a total of 8 runs, and the Red Sox carry a five-game home winning streak into the night. A -142 price implies roughly a 59% chance before adjustments; the model's task is to weigh Bradish's steady ERA against a Boston club in form at home, and to decide whether the number already reflects that streak. Suarez's 0-0 line signals limited recent sample, which widens the uncertainty band rather than narrowing it.
Elsewhere, several matchups pair established arms against younger or less-charted opposition. In Philadelphia, Zack Wheeler hosts the Dodgers opposite Justin Wrobleski. In Toronto, Kevin Gausman draws Drew Rasmussen and Tampa Bay. Detroit at the Cubs lists Framber Valdez against David Peterson, and Seattle's Luis Castillo faces Cincinnati's Chase Burns. Each of these is a case where the model weighs the veteran's known distribution of outcomes against the wider variance of the opposing starter, then cross-checks against the total and the park.
A cluster of games features arms we cannot yet profile from verified recent outings — among them Kendry Rojas vs Parker Messick in Cleveland, Bubba Chandler vs Will Warren at Yankee Stadium, Walker Buehler vs Reynaldo Lopez in Atlanta, Noah Schultz vs Kumar Rocker in Texas, and Tyler Phillips vs Tatsuya Imai in Houston. For those, we decline to assert form we cannot confirm. The model leans harder on lineup strength, bullpen rest and run-environment inputs in exactly these spots, and the honest read is that our confidence is lower where the pitcher sample is thin. That is reflected in smaller stated edges, not manufactured ones.
Colorado is its own consideration: Washington's Miles Mikolas faces Michael Lorenzen at altitude, where totals behave differently and any single-number read deserves a wider error bar. St. Louis at the Angels (Matthew Liberatore vs Walbert Urena) and Oakland at Arizona (Jack Perkins vs Kohl Drake) round out the West Coast portion, where late first pitches and travel enter the calculation.
Across all 15, the process is the same: convert the market price to an implied probability, compare it to the model probability, and publish only where the edge clears our threshold. Where the two align, we say so and pass. Every posted pick and its result lands on the ledger.
MLS
No games today. MLS sits between its World Cup restart and its July 29 All-Star Game in Charlotte, with the next full slate expected around July 25. There is nothing to model and nothing to publish here today. When the schedule resumes, our attention returns to rest differentials, congestion from the restart and home-field effects — but forcing a read on an empty card would be the opposite of disciplined, so we hold.
WNBA
Also dark. The league is in its All-Star break, with the AT&T WNBA All-Star Weekend exhibition scheduled for Saturday, July 25 in Chicago. The relevant storyline is availability rather than a matchup: Paige Bueckers and Angel Reese were both hurt on Sunday. Bueckers collided with Nneka Ogwumike in the fourth quarter against Los Angeles and sat Monday against New York, though she cleared concussion protocol. Reese's knee buckled on a jump-stop against Chicago; she rose quickly, went down again and did not return. If either cannot go Saturday, Commissioner Cathy Engelbert will name a replacement. Note plainly that an exhibition is a poor modeling target regardless — effort and rotations are unpredictable — so any future read there would carry an unusually wide uncertainty band, if we engage it at all.
Today, then, the work is baseball. We will post each MLB pick roughly 30 minutes before first pitch and let the ledger record the outcome, wins and losses alike.
Predictions carry risk and can lose. If it stops being fun, stop. 21+; call 1-800-GAMBLER.