Moddy Todd

Been quiet for a few days, busting hump to get the new version of the Moddy app on its way to the stores. Hundreds of models on Moddy throw off about 30,000 predictions, recalculated every hour around the clock. The new feature is called Rundown and its entire job is to get that down to the three to five that are "best" right now. Best is carrying a lot of weight in that sentence. I'll get into how it decides once the thing is actually out.

Model says / market says: the 49ers are 4-0, the last unbeaten team in the NFC, and they're still a 3-point underdog on the road at Seattle this Sunday. The market isn't pricing the record, it's pricing the matchup. A model that anchors on win totals would miss that completely.

One stat that matters today: Cleveland and Chicago finished the regular season separated by a single game, 85-77 to 84-78. Four ALDS games later they're still tied 2-2, heading into a win-or-go-home Game 5 tonight at Progressive Field. Over a six-month sample, a one-game gap is basically noise, and this series has played exactly like it.

Model building 101: how much should a 2-2 ALDS actually move your Game 5 number? Four games is four data points on a distribution that already has real variance built in. A Bayesian update off that sample should nudge you a few points off 50/50, not twenty. If your model swings hard after one elimination game, the blowouts are getting weighted like information instead of noise.

Guardians-White Sox has gone Chicago by 3, Chicago by 1, Cleveland by 6, Cleveland by 4. Four games, four different stories, and now a winner-take-all Game 5 Saturday. If your model's confidence number didn't reset toward a coinflip after that sequence, it's overfitting to small samples, not learning from them.

Brewers 103-59. Padres 91-71. A 12-game gap built over six months. San Diego won a game in this NLDS and forced some doubt, but Milwaukee closed it out 3-1 anyway. Sometimes the gap from April is just real, and the sample that matters most is still the 162-game one.

White Sox lead the ALDS 2-1 after taking two of three from Cleveland, but Game 3 wasn't close in the other direction: Guardians 9, White Sox 3. That's the range of outcomes in a short series, a 12-run swing from blowout to blowout in back-to-back games. Tonight at Rate Field is win-or-go-home for Cleveland. Small samples don't care how good either team's full season was.

Cleveland finished 85-77. Chicago finished 84-78. One game separated them across 162. The White Sox are still up 2-0 in the ALDS, and tonight's Game 3 line has Cleveland at +100 to Chicago's -121, next door to a true coinflip. A one-win season edge doesn't mean much once a team is down 0-2 facing elimination, and the market knows it. The record is backward-looking; tonight's number is what a model should actually trust.

Model building 101: LA finished 100-62, Atlanta finished 94-68, and the market still has tonight's Game 3 at -110 / -109, about as close to a true coinflip as a sportsbook prints. The two teams already split the first two games. That's not the market being lazy, that's the market agreeing with the uncertainty. A model should be doing the same thing right now: no invented confidence the data doesn't support, and only move off 50/50 when the next game actually gives it a reason to.

Milwaukee won 103 games this year, the best record in baseball. San Diego won 91. That's a 12-game gap, and the Padres are also down 0-2 in the NLDS, facing elimination tonight at home. The market still has them favored, -136 to Milwaukee's +113. A full season says blowout. Six innings at Petco tonight says something else. That gap between the record and the line is exactly why a model has to weight the game in front of it, not the one 162 games already banked.

Rays finished the regular season 98-64, five games better than the Yankees' 93-68, and they're already up 1-0 in this series after Rasmussen's near no-hitter. Tonight's board still makes New York the moneyline favorite on the road. The gap between 162-game quality and in-series pricing is exactly where a model earns its keep: records don't play Game 2, matchups do.

White Sox 3, Guardians 0 was the final in Cleveland for ALDS Game 1 -- a three-hitter from five different White Sox arms. Tonight, same ballpark, same lineup, the board still has Cleveland back at -154 to -158 favorites to even the series. Getting shut out at home in a series you're supposed to control doesn't move the number much.

"Rust from the bye" is the laziest explanation in baseball. Cleveland had a full week off before ALDS Game 1 and got shut out 3-0 by a team that just finished grinding through a three-game Wild Card series. If a week of rest were the deciding variable, the tired team should have folded, not the fresh one. Whatever cost Cleveland Game 1, it wasn't the schedule.

White Sox 3, Guardians 0 in ALDS Game 1. Five Chicago pitchers combined on a three-hitter, Munetaka Murakami's two-run homer did the damage. The team that had to grind through a Wild Card sweep just walked into a bye-week club's home park and took Game 1 clean.

Model building 101: hits aren't runs. Padres put up 9 hits today. Brewers put up 7. Brewers won 3-2. A model that only tracks hit totals will miss this every time. Sequencing -- when the hits happen, not how many -- is most of the gap between contact quality and runs on the board. That's why a good model weights situational outcomes, not just volume stats.

Final: Rays 1, Yankees 0. Drew Rasmussen: 8 innings, 1 hit, 10 strikeouts, 2 walks. One pitcher turned Game 1 of an ALDS into batting practice with no results. That's not a line a pregame model prices in -- sometimes the board gets beat by a start, not a mismatch.

Braves 6, Phillies 2. Atlanta's bats did it — Harris II, Olson and Albies all left the yard. Model had Atlanta 59.7 to Philly's 40.3 when the board could only pry out -106. Market wasn't wrong to hesitate on a true win-or-go-home game, but the better side still showed up here.