Sunday's Premier League fixtures bring four matches and a model that is clear on three of them. Four weeks into the season, xG profiles are stabilizing and situational factors are starting to compound in ways that matter. Two picks land on the under. Two land on the over. The reasoning behind each is different and I will go through all of it. One pick is explicitly a lean and I will label it as such from the start. The edge is either there or it is not, and I am not in the business of pretending otherwise.

Liverpool vs AFC Bournemouth | Under 2.5 Goals (+180)

Liverpool press at a PPDA of 7.1. That is the most aggressive pressing figure in this dataset and it becomes most effective when the opponent is fatigued or fielding a rotated lineup. AFC Bournemouth came off a Europa League fixture on Thursday evening. Marco Rose is expected to rotate his backline heading into Sunday. Liverpool, by contrast, recalled Van Dijk, Wirtz, and Alisson after Tuesday's Carabao Cup rotation and enter with their first-choice defensive spine fully restored.

The xG profiles support a tight, controlled match. Liverpool sit at 1.74 xG per game with a 1.50 xGA. Bournemouth are at 1.79 xG per game with a 1.43 xGA. Blend those numbers together and the combined expected goals for this fixture lands at approximately 2.5. That is the under threshold. The model is not projecting a high-scoring game. It is projecting precisely the kind of game the under requires.

Andoni Iraola managed Bournemouth before joining Liverpool this summer and addressed the emotional weight of returning to Vitality Stadium directly. As he put it: "From a personal side, I'm really looking forward to see them, players that I love and staff I enjoyed working with. I wish them all the best, but we need to go there and beat them and be as ruthless as possible." Ruthless and controlled is exactly the outcome the under needs. Liverpool are not traveling to put on a show. The under is priced at +180 against -192 on the over. That pricing gap is significant given what the blended projection actually says. Confidence: medium.

Crystal Palace vs Leeds United | Over 2.5 Goals (-135)

Crystal Palace carry a xGA of 2.23 per game. Their PPDA of 16.4 tells you precisely how they defend: passively, without pressure, inviting opposition build-up in their own half. They arrive at Elland Road missing two starting centre-backs and with their goalkeeper situation further complicated. This is Crystal Palace at their most structurally exposed point of the season.

Leeds United are unbeaten through their Premier League matches and their most dangerous striker has scored in five consecutive home league appearances. Leeds average 2.0 goals per home game. Palace have conceded two or more goals in ten consecutive Premier League matches. When you combine Palace's xGA of 2.23 against Leeds's home attacking output and the near-certain participation of both teams in the scoring column, the case for clearing 2.5 is straightforward. Palace's PPDA of 16.4 means they will not press Leeds off the ball. Leeds will have time and space to build the way they want.

Daniel Farke addressed his team's position without ambiguity: "It's not a coincidence. It's quality. The table is currently not important. Eight points are important." That is a focused, structured team with statistical backing. Palace arrive in their worst defensive shape of the season and facing a striker operating at full confidence. The over at -135 reflects the market's awareness of this setup. The data supports the price. Confidence: medium.

Sunderland vs Manchester City | Over 2.5 Goals (-161)

Manchester City average 2.09 expected goals per game at home. That is the highest attacking xG figure in this dataset. Their xGA sits at 1.02, reflecting a defensive organization that eliminates meaningful counter-attack risk. Their PPDA of 8.2 confirms they will press hard against a Sunderland side coming in fatigued.

Sunderland arrive at the Etihad for their fourth match in twelve days. Reinildo is suspended. Alderete carries a muscle injury from Wednesday's Europa League fixture and is a major doubt. Habib Diarra is unavailable. Their left-back situation may require a player to make his Premier League debut against City's wide attackers. Sunderland have scored three goals in four league matches this season. Their xG of 1.74 per game suggests they are generating chances at a reasonable rate. Converting them is a different problem entirely and it is one they have not solved yet.

There is a documented pattern of City's primary striker not having scored against Sunderland across prior appearances. I will acknowledge it and then set it aside. One player's head-to-head record does not override a team's 2.09 xG per game output. City generate enough expected output through multiple sources to breach 2.5 regardless of who finishes the chances. Any Sunderland contribution to the total only makes the outcome cleaner. The over at -161 is the price for a structurally favorable situation in a heavily mismatched fixture. Confidence: medium.

Manchester United vs Fulham | Under 2.5 Goals (+160)

LOW confidence means what it says. The model projects the true probability at approximately 45 percent against 38.5 percent implied by the +160 odds. That is positive expected value in the technical sense. It is not a thick edge and I would not fault anyone for passing on this game entirely. Forcing a play on every match on a slate is how bankrolls erode. Patience is the most underrated skill in this business.

Manchester United carry the highest xG figure in this dataset at 2.12 per game. They lead the league in touches in the opposition box and rank second in total expected goals output. The process metrics are elite. The results across four matches have not followed, which is a sharp-money puzzle that the model is still processing. Sesko is not traveling with the squad, removing an established finishing option from a system that is already underconverting relative to its shot quality. Fernandes and Mbeumo remain capable weapons. But the absence of a primary striker at the margin matters when the blended total is sitting at the line.

Fulham sit bottom of the table but their attacking profile at home is not what the table reflects. Their deep completions per game at home are the highest figure in this entire dataset, indicating genuine quality in the final third on their own pitch. Their xGA of 2.08 confirms the defensive leakiness is real and cuts against the under. The blended total for this match lands at 2.5 exactly. That is the least comfortable number to be positioned on either side of. The model says the under at +160 offers a calculable positive edge. I am taking it as a lean and sizing it accordingly. The edge is not thick enough to treat it as anything more than that.