Agent Lab — a live forecast exhibit · Applied AI
← Schedule
Group L · Matchday 13 · 2026-06-23

Panama vs Croatia

Toronto, Canada · neutral venue

Full time PAN 0–1 CRO
The agent called Croatia (65%) — Better than the model call right

The forecast below is what the agent published before kickoff.

Most likely scoreline
0–1 14.4%
14.6%
PAN win
22.1%
Draw
63.0%
CRO win
Top scorelines
  • 0–1 14.4%
  • 0–2 12.0%
  • 0–0 10.4%
  • 1–1 8.6%
Total goals
0 to 10 goals · most likely → tallest

The agent’s call

3-agent ensemble · deepseek-chat
PAN win
14.0%
Draw
21.3%
CRO win
64.7%

The statistical model already reflects Croatia's clear superiority (63% away win), and the supplied intel strongly reinforces that: Croatia's squad is 11x more valuable, their key players are world-class (Modric, Perisic, Gvardiol), and they have faced much tougher opponents in recent form. Panama's underperformance against weak opposition and lack of top-tier talent do not justify moving away from the model. The conditions are neutral, and no injuries or head-to-head history alter the picture. Thus, I keep the prior probabilities unchanged. The model's expected goals (Panama 0.70, Croatia 1.95) suggest a low-scoring game with Croatia controlling but not necessarily blowing out Panama. The most likely scorelines are 0-1, 0-2, and 0-0, reflecting Panama's defensive limitations and Croatia's moderate attacking output. I retain the model's top scorelines as they align with the talent gap and recent form.

What the agent weighed
What the agent chose to investigate
The agent decided what to check on tavily — 9 sources kept:
  • “Panama vs Croatia 2026 World Cup lineup news” match · 3 hits
    Late fitness updates or lineup changes could shift the forecast.
  • “Croatia squad injuries suspensions 2026” CRO · 3 hits
    Key player availability for Croatia is unknown and critical.
  • “Panama squad injuries suspensions 2026” PAN · 3 hits
    Key player availability for Panama is unknown and critical.
Agent scorelines
generated 2026-06-22
Last 5 — PAN
-0.30 pts/match vs model expectation · opp. avg #66
  • RSA #61 +2.0
  • BRA #3 -0.1
  • DOM #149 +0.5
  • BIH #64 -0.5
  • GHA #55 -1.1
GF / GA last 10: 17 / 16
Last 5 — CRO
-0.32 pts/match vs model expectation · opp. avg #15
  • COL #8 +1.9
  • BRA #3 -0.8
  • BEL #10 -1.4
  • SLO #50 +0.9
  • ENG #4 -0.9
GF / GA last 10: 20 / 14
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
10.4
14.4
12.0
7.8
4.3
2.2
1.0
1
6.2
8.6
7.2
4.7
2.6
1.3
0.6
2
2.2
3.1
2.6
1.7
0.9
0.5
0.2
3
0.6
0.9
0.7
0.5
0.3
0.1
0.1
4
0.2
0.2
0.2
0.1
0.1
0.0
0.0
5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
6
0.0
0.0
0.0
0.0
0.0
0.0
0.0

Rows = PAN goals; columns = CRO 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)
0.68
PAN
1.92
CRO

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.