The result and the grades
Tottenham Hotspur beat Crystal Palace 2-0 at home on 31 October 2026, and every forecaster on the site called the outcome correctly. The difference came in the detail. Claude Keeper predicted 2-0 and was the only one to land the exact score, which also gave it the best Brier score of the group at 0.2168. Gemini Thunder followed at 0.2418, then GPT Striker at 0.2678.
The baseline model, which went for a 1-0 home win, scored 0.2946. That put it ahead of DeepThinker (0.305) and Kimi Cool (0.3464), even though those two predicted 2-1 and a more attacking game. Brier scoring punishes confidence, so a model can pick the right winner and still grade poorly if it put too much probability on the wrong margin or on goals at both ends. Kimi Cool is the clearest example: right outcome, wrong score, and the weakest probability calibration of the six.
Four of the five models went for 2-1, expecting Palace to score once. They did not. Palace failed to score for the third match running, which was the one piece of the picture those models underweighted.
Why the models leaned Tottenham
The pre-match case for the home side was reasonable, if not overwhelming. Roberto De Zerbi’s team arrived on a W-W-D run, having won 2-0 at Chelsea on 24 October and 3-1 against Coventry City five days earlier. Their last home league game before this one had been a 2-3 defeat to Aston Villa, so there was some home fragility to weigh.
Tottenham also had selection problems. [1] reported that Xavi Simons was out with a knee problem until 1 January, and [2] noted that Rodrigo Bentancur was out after a red card. Against that, the club had added Mykhailo Mudryk, Tosin Adarabioyo and Sávio in the last week of the window, according to [3], and Mudryk had already returned to competitive action after a 645-day absence, as [4] described.
Palace’s profile pointed the other way. They had lost 0-5 at Brighton on 18 October and 0-2 at home to Newcastle United on 25 October. Their only recent win was 3-1 over Nottingham Forest. The head-to-head record was mixed: Palace won 3-1 at Spurs in March 2026, while Tottenham took a 1-0 win at Selhurst Park in December 2025. That history gave the models a reason for caution, but recent form outweighed it.
Value picks and what they returned
All five models also submitted a value pick. Four of them, Kimi Cool, Gemini Thunder, GPT Striker and DeepThinker, backed the home side on the head-to-head market at 1.57. All four hit, each returning +0.57 units. That is a safe, short-priced pick that rewarded the consensus view without separating anyone.
Claude Keeper took a different route, choosing the under in the totals market at 2.285. With the game finishing 2-0, that landed and returned +1.285 units, more than double the other models’ profit. It fits the model’s score call: if you expect a 2-0 game, the under is the natural expression of that view, and the longer price paid out accordingly. The baseline submitted no value pick.
One result is a small sample, and a single winning under should not be read as proof of a lasting edge. It does show the exact-score pick and the value pick pointing the same way, which is how a consistent forecast should behave.
What it means
For Tottenham, the win lifts them to 8th in the Premier League with 12 points from nine games (W3 D3 L3, goals 10 for and 10 against). Three straight league wins, over Coventry, Chelsea and now Palace, is their best run of the season, and the goal difference has reached level. The next test is away at Leeds United on 7 November, a trip that asks whether they can win on the road for a second time in three weeks after the Chelsea result. The site’s forecast for that match locks 24 hours before kick-off.
For Crystal Palace, the picture is bleaker. They are 18th with 7 points from nine games (W2 D1 L6), with nine goals scored and 21 conceded. The last three results read 0-2, 0-5 and 0-2, a run of no goals and nine conceded. Their next fixture is at home to Liverpool on 8 November, which will demand a much tighter defensive display than they have managed recently. The site’s forecast for that game also locks 24 hours before kick-off.
The grading lesson is simpler. When a team has stopped scoring, models should be wary of assigning Palace a goal by default, and Claude Keeper’s 2-0 was the cleanest read of that trend.
Sources
- The Guardian: Premier League team news: predicted lineups for the weekend action theguardian.com ↩
- BBC Sport: Response after red card 'builds mentality' at Spurs bbc.co.uk ↩
- BBC Sport: Tottenham to fly to Marbella for team bonding bbc.co.uk ↩
- The Guardian: Glasner claims VAR was ‘guessing’ after Forest are denied winner against Spurs theguardian.com ↩
