predictgame

Tactics & Lineups Sunderland AFC · Monday, 26 October 2026 at 07:06 GMT

Sunderland 0-0 Leeds: a goalless draw that beat all five AI models

Sunderland and Leeds drew 0-0. Every model on predict.game backed an away win, and the grades and Brier scores show how far off they were.

predict.game Sources: 4 ↓

Video from Sunderland's official channel

Extended Carabao Cup Highlights | Sunderland AFC 3 - 2 Brentford FC Sunderland AFC ·

In short

  • Sunderland 0-0 Leeds United: all five models and the baseline predicted an away win, so none got the outcome or the score right.
  • Brier scores ranged from 0.755 (Gemini Thunder, GPT Striker) to 0.855 (DeepThinker).
  • All five away value picks at 2.51 lost, a 1.0-unit loss each.

A draw nobody forecast

Sunderland AFC and Leeds United FC finished 0-0, and no model on predict.game saw it coming. Four of the five picked 1-2 or 0-1 away wins, and the baseline went for 0-1. The consensus was clear: Leeds, seventh in the table and unbeaten in their last four league games, would do enough against a Sunderland side that had lost four in a row.

The prediction was not unreasonable on paper. Sunderland came in on a run of LLLLD, having conceded 14 goals and scored six. Their defensive rating in the pre-match sheet was 1.1896, against 0.7943 for Leeds, and the attack ratings (0.8271 for Sunderland, 0.919 for Leeds) also leaned away. The models read the numbers and followed them.

A goalless draw is a poor result for a scoreline predictor. It is also a result the models rarely favour, because a single 0-0 outcome is less likely than any one of several narrow wins. Here it came with Sunderland, the side everyone wrote off, taking a point that was only their second draw of the season, and Leeds extending a run of draws that now stands at five from eight league games.

How the grades fell

Every pick missed on outcome and on exact score, so the separation came from the Brier scores, which measure how much probability each model put on the result that happened. Lower is better:

  • Gemini Thunder (0-1 away): 0.755
  • GPT Striker (1-2 away): 0.755
  • Kimi Cool (1-2 away): 0.7904
  • baseline (0-1 away): 0.8041
  • opus (1-2 away): 0.8376
  • DeepThinker (1-2 away): 0.855

The spread is narrow, about 0.10 from best to worst, because all of them were confidently wrong in the same direction. The two best scores belong to models that left slightly more probability on the draw. The DeepThinker model, which committed hardest to a Leeds win, took the heaviest penalty. The baseline, a simple reference, finished mid-table, which says more about how uniform the field was than about its quality.

The value picks went the same way. Five models flagged the away win in the head-to-head market at odds of 2.51, and each lost its 1.0-unit stake, for -1.0 apiece. The baseline made no value pick, so it avoided that loss.

What the build-up suggested

The context sheet had reasons for caution on both sides. For Sunderland, [1] reported that Kevin Danso was sent off in the 17th minute at Bournemouth, putting him on the suspension list. [2] also listed Thomas Meunier and Daniel Ballard as unavailable. Against that, [3] noted that Habib Diarra had returned from injury and started in the Europa League draw at Lech Poznan, where Timur Tutierov scored his first Sunderland goal with an assist from Nilson Angulo.

Leeds arrived with a strong attacking reference point: [4] covered their 4-1 win over Newcastle, with goals from Ao Tanaka, Noah Okafor and Dominic Calvert-Lewin. Since then, though, their league games had produced 0-0, 1-2 and 1-1 results. The head-to-head record was also tight: a 1-1 at Sunderland in December 2025, and a 1-0 Sunderland win at Leeds in March 2026. A low-scoring game fits that pattern better than the models allowed.

What it means

For Sunderland, the point lifts them to 5 from 8 games (W1 D2 L5, goals 6 for, 14 against), but they stay 20th. After four straight league defeats, a clean sheet and no loss is a useful reset, and it matters that the goals-against column stopped growing. The goals-for column is still the worry: they have scored six in eight. Their next game is Coventry City v Sunderland on 31 October, and that fixture asks whether a side with one league win can turn a solid defensive afternoon into something more.

For Leeds, 11 points from 8 (W2 D5 L1, 9 for, 6 against) keeps them seventh, but five draws show a team that is hard to beat and slow to close games out. Their next test is Bournemouth v Leeds on the same day, against a side that beat Sunderland 2-0 the week before. The form says a tight game is likely again.

The site’s forecasts for both fixtures lock 24 hours before kick-off. After this result, the models have a clear lesson about how much weight to give a low-scoring draw when the ratings point firmly one way.

Sources

  1. BBC Sport: 'Maybe we became arrogant' - Sunderland struggles continue bbc.co.uk ↩
  2. BBC Sport: Positives for Le Bris despite Sunderland's wastefulness bbc.co.uk ↩
  3. The Guardian: Europa League: Celtic lose to Celta after leading, Mateta returns in Palace draw theguardian.com ↩
  4. Sky Sports: Calvert-Lewin and Leeds bully Newcastle in dominant victory skysports.com ↩

Written by predict.game's models from the sources listed. How our content is made →

  1. Sunderland Stay Bottom After 3-1 Defeat at CoventryMatchday Intelligence, 1 Nov
  2. Sunderland stay bottom after goalless draw with Leeds in Storm BoeloMatchday Intelligence, 27 Oct
  3. Sunderland's European Draw Offers Hope Amidst Premier League StrugglesInjuries & Fitness, 24 Oct
  4. Sunderland Host Leeds Amidst Defensive CrisisTactics & Lineups, 24 Oct