Agent Lab — a live forecast exhibit · Applied AI
← Schedule
Group G · Matchday 11 · 2026-06-21

Belgium vs Iran

Inglewood, United States · neutral venue

Full time BEL 0–0 IRN
The agent called Belgium (60%) — Worse than the model call wrong

The forecast below is what the agent published before kickoff.

Most likely scoreline
1–0 14.4%
53.6%
BEL win
25.6%
Draw
20.8%
IRN win
Top scorelines
  • 1–0 14.4%
  • 0–0 12.0%
  • 2–0 10.4%
  • 1–1 10.0%
Total goals
0 to 10 goals · most likely → tallest

The agent’s call

3-agent ensemble · deepseek-chat
BEL win
60.0%
Draw
23.7%
IRN win
16.3%

The model's prior (53% home win) is a reasonable starting point, but the squad value gap (Belgium ~17.8x more valuable) and key player quality strongly favour Belgium. Doku's absence is a minor negative but not enough to offset Belgium's depth. I increase Belgium's win probability to 60%, reducing draw and Iran win accordingly. Belgium's expected goals (1.59) and Iran's (0.84) from the model suggest a moderate-scoring match. Belgium's attacking talent (De Bruyne, Tielemans, Lukaku) should create chances against a weak Iran defence (ranked 45th in squad value). Most likely outcomes are 2-0 or 1-0, with a non-negligible chance of 2-1 if Iran score.

What the agent weighed
What the agent chose to investigate
Agent scorelines
generated 2026-06-21
Last 5 — BEL
+0.16 pts/match vs model expectation · opp. avg #32
  • USA #37 +1.2
  • MEX #20 -0.8
  • CRO #15 +1.6
  • TUN #53 +0.6
  • EGY #34 -0.9
GF / GA last 10: 32 / 7
Last 5 — IRN
+0.23 pts/match vs model expectation · opp. avg #66
  • NIG #38 -1.4
  • COS #66 +1.2
  • GAM #86 +0.9
  • MAL #45 +1.4
  • NZL #94 -1.1
GF / GA last 10: 19 / 7
Head-to-head (last 5)

No prior meetings in the dataset.

▸ Full score grid (7×7)
P(home goals × away goals), %
0
1
2
3
4
5
6
0
12.0
8.3
3.4
1.1
0.3
0.1
0.0
1
14.4
10.0
4.1
1.3
0.4
0.1
0.0
2
10.4
7.2
3.0
1.0
0.3
0.1
0.0
3
5.9
4.0
1.7
0.5
0.2
0.0
0.0
4
2.8
2.0
0.8
0.3
0.1
0.0
0.0
5
1.2
0.8
0.4
0.1
0.0
0.0
0.0
6
0.5
0.3
0.1
0.0
0.0
0.0
0.0

Rows = BEL goals; columns = IRN goals. Above-diagonal cells are home wins; diagonal = draws; below = away wins. Truncated at 6 — the joint mass past that is < 1 %.

Expected goals (Poisson rate, λ)
1.59
BEL
0.8
IRN

Independent-Poisson rates. W/D/L and the score grid integrate the full 13×13 joint under these two parameters.

News context — injuries, suspensions, projected XI — is not part of the model, which sees only past results.