Chelsea vs Tottenham Preview: High-Pressure London Derby

Stamford Bridge hosts a high‑pressure London derby as Chelsea and Tottenham meet in the Premier League with just two rounds left. Chelsea sit 10th on 49 points (13‑10‑13, 55‑49), while Tottenham are 17th on 38 points (9‑11‑16, 46‑55) and still looking over their shoulder at the relegation battle. Despite the table and home advantage pointing slightly towards Chelsea, the underlying prediction model and market prices both suggest a much more balanced contest.

Form is a major red flag for Chelsea. Their standings “form” string is “DLLLL”, and the prediction feed rates their last‑five performance at only 7%, with 2 goals scored and 11 conceded (0.4 for, 2.2 against per game). That is relegation‑level output over the recent block. Across the league campaign they average 1.5 goals for and 1.4 against per match, but those season‑long numbers mask a sharp downturn: defensively they have become porous and the attack has stalled. The comparison section quantifies this: Chelsea trail badly on form (11% vs 89%), attack (25% vs 75%) and defence (31% vs 69%).

Tottenham, by contrast, arrive in much better short‑term shape. Their last‑five metrics show 53% form, with 6 goals scored and 5 conceded (1.2 for, 1.0 against per game). Over the full league campaign they are slightly less potent than Chelsea in attack (1.3 goals per match vs 1.5) but similar defensively (1.5 conceded vs Chelsea’s 1.4). Crucially, their away profile is solid: 7 wins, 5 draws and 6 losses on the road, with 25 scored and 24 conceded. That is a mid‑table away record, significantly stronger than their fragile home form. Given Chelsea’s very ordinary home record (6‑5‑7, 24‑24), the venue edge is not overwhelming.

Head-to-Head Data

Head‑to‑head data in the prediction feed covers a rich recent history. All of the following are competitive fixtures, with dates and competitions verified:

• 2025‑11‑01 (Premier League, at Tottenham Hotspur Stadium): Tottenham 0‑1 Chelsea.
• 2025‑04‑03 (Premier League, at Stamford Bridge): Chelsea 1‑0 Tottenham.
• 2024‑12‑08 (Premier League, at Tottenham Hotspur Stadium): Tottenham 3‑4 Chelsea.
• 2024‑05‑02 (Premier League, at Stamford Bridge): Chelsea 2‑0 Tottenham.
• 2023‑11‑06 (Premier League, at Tottenham Hotspur Stadium): Tottenham 1‑4 Chelsea.
• 2023‑02‑26 (Premier League, at Tottenham Hotspur Stadium): Tottenham 2‑0 Chelsea.
• 2022‑08‑14 (Premier League, at Stamford Bridge): Chelsea 2‑2 Tottenham.
• 2022‑01‑23 (Premier League, at Stamford Bridge): Chelsea 2‑0 Tottenham.
• 2022‑01‑12 (League Cup, at Tottenham Hotspur Stadium): Tottenham 0‑1 Chelsea.
• 2022‑01‑05 (League Cup, at Stamford Bridge): Chelsea 2‑0 Tottenham.

In Premier League meetings specifically, Chelsea have repeatedly found ways to win both home and away, including the most recent league clash at Tottenham Hotspur Stadium on 2025‑11‑01 (1‑0 to Chelsea) and the 1‑0 home victory at Stamford Bridge on 2025‑04‑03. However, that historical dominance is at odds with the current form and the model’s forward‑looking assessment.

The prediction engine explicitly flags Tottenham as the side to be with on a “win or draw” basis, with an official advice of “Double chance: draw or Tottenham”. Implied probabilities from the model are 10% home, 45% draw, 45% away, which is much more bearish on Chelsea than the bookmakers. Market odds cluster around 2.05–2.13 for Chelsea, 3.60–3.84 for the draw and 3.10–3.50 for Tottenham. Converting roughly, the books have Chelsea near 46–48% implied, the draw around 26–27%, and Tottenham around 28–31%. That means the model sees Chelsea as heavily overpriced and the away side significantly undervalued.

Given Chelsea’s collapsing recent form, Tottenham’s stronger short‑term performance and robust away record, plus the prediction model’s clear tilt towards the visitors on double chance, the most data‑aligned play is to follow that advice rather than the raw H2H trend.

Betting verdict: the recommended angle is Double Chance – Draw or Tottenham, opposing Chelsea at short home odds and aligning with the 45%/45% draw‑away probability split provided by the prediction data.