Case 3: Team A has one draw and one loss.
In these situations, there is a result of the last group game (B vs. C) that eliminates A.
Let’s imagine that the United States is Team A and that Canada and Mexico are teams B and C. (This would not be a realistic World Cup group, but we’re using these three as examples because on Wednesday they’re expected to win the competition to jointly host the 2026 World Cup, which will feature the new format.)
If the United States (A) won by 1-0 against Canada and lost by 2-0 against Mexico (Case 1), Mexico could agree to lose by 1-0 against Canada in the last group game and would still win the group, with Canada as the runner-up.
If the United States has two draws of 0-0 and 1-1 (Case 2), Mexico and Canada could arrange a 2-2 draw to eliminate the U.S. because of a higher number of goals scored. In Case 3, any draw between Mexico and Canada eliminates the U.S.
In all other cases, Team A will already have advanced or been eliminated before the last group game. To minimize the risk of match fixing, Team A should be the strongest team in the group, so that it has higher chances of having already advanced. Or it should be the weakest team, if very weak, so that there are higher chances that it is already eliminated with two losses. But arbitrarily deciding which team will play the first two games in a group seems unfair because it’s the only team that may be the victim of collusion.
My computations show that, for a realistically unbalanced group, the risk of collusion is about 15 percent in any given group if Team A is the strongest in the group. But it climbs to around 50 percent otherwise. This means that it will be almost certain — around 90 percent in the first case (if Team A is the strongest in the group) and more than 99 percent otherwise — that at least one of the 16 groups will face suspicion of match fixing.
Aware of this danger, FIFA, world soccer’s governing body, has considered forbidding draws during the group stage. Games ending in a draw would be decided by penalty shootouts. This would indeed rule out Cases 2 and 3.