Do the new World Cup tiebreaker rules make the final Group Stage games worse?
A comparison of head-to-head and goal difference tiebreakers, featuring chess tournaments
For the first time, FIFA World Cup groups will be ranked primarily on the head-to-head performance of teams, marking a change from the use of goal difference. In this way, if teams are even on points, they will be ranked based on their results in the matches between the even teams. This brings the World Cup in line with the Euros, Asian Cup and AFCON - which all importantly also award knockout positions to some 3rd place teams, which we will discuss.
Here are the complete tiebreaker rules:
Let’s explore some implications of the new rules that are already beginning to show.
The World Cup dead rubber
As an example of what this means, let’s have a look at how Group D looks right now:
The US have beaten Australia and Paraguay, who have each beaten Turkey. It is entirely possible that, for example, Australia beats Paraguay and catches up to the US on 6 points and that Turkey beats the US and catches up to Paraguay on 3 points. However, as we are considering head-to-head results, the US would remain top, and Turkey would remain bottom. In this way, we have two teams whose positions cannot change. So the USMNT-Turkey game is going to be a complete dead rubber that will not impact the outcome of the group.
This is strictly impossible under a goal difference system. As a simple proof, if a team is untouchable on 6 points, being the most points possible after 2 games, then there must be no other team on 3 or more points. The bottom team would therefore be able to catch another team that is no more than 3 points ahead of them (a useful situation to consider is the group looking like 6-2-1-1 points after 2 rounds). In this way there can only be one team that has guaranteed their position under a goal difference system.
The old system still allowed group stage dead rubbers to occur, as both teams may have already been eliminated if the group was 6-6-0-0 points after 2 rounds. However in this case at least both sides were in the same position of playing for pride, as opposed to the 1st v 4th dead rubber that we will see with US-Turkey and may see in another group.
And of course, I must mention that some would consider the 3rd place playoff a dead rubber as well. So the new system has not invented the World Cup dead rubber, but it has however introduced a situation where multiple teams will have their positions locked after the first two rounds.
It’s worth mentioning that the FIFA World Rankings will still give massive importance to the US-Turkey dead rubber, weighting it as much as any other World Cup game. The World Rankings are important for World Cup qualification draws, so this feels incorrect. That being said, it is probably impractical to hardcode a change to the World Rankings formula for a specific game.
What about being better against better teams?
You may have noticed that in the example where Turkey beats the US to finish even on 3 points with, say, Paraguay, Turkey would have beaten the top team in the group (imagine it wasn’t a dead rubber), and Paraguay would not have. If we’re not considering goal difference, there are two competing arguments here, which we’ll call:
Paraguay’s case: Paraguay are better than Turkey and should go through, because Paraguay beat Turkey; and
Turkey’s case: Turkey are better than Paraguay and should go through, because Turkey beat the top team in the group, and Paraguay lost to the top team in the group.
The old goal difference system is agnostic as to these options.
For a direct comparison of the two teams, Paraguay’s case is pretty persuasive. However, sport is obviously non-transitive - that is to say results don’t form a single linear ranking where better teams always beat worse teams. There are rock-paper-scissors scenarios where just because Team A beat Team B and Team B beat Team C, it doesn’t mean Team A will beat Team C. This is partially because upsets happen and partially because some play styles match up with certain play styles better than others.
You might hear these rock-paper-scissors scenarios called ‘circles of parity’ or ‘Condorcet cycles’ (technically, this refers to an outcome concerning elections in social choice theory, but I think it’s an accurate descriptor and so does Smead (2019)). This is a long way of saying a single game is often not determinative - which is why we have leagues and group stages.
In this way, if you asked a chess tournament organiser they would probably agree with Turkey’s case. Most round-robin chess tournaments use the Sonneborn-Berger Score (SB Score) tiebreak method, defined as:
the full conventional score of each defeated opponent and half the conventional score of each drawn opponent
Basically your tiebreaker score is the sum of the scores of the people you beat. The motivation is simple - beating better people must be more impressive. If you’ve ever played a chess tournament on chess.com, the SB Score was probably used.
If we have a look at this example tournament provided by chess.com, we can see the top player got a slightly higher tiebreak score because they beat 3rd place twice and 4th place once, as opposed to 2nd place beating 4th place twice and 3rd place once.
These methods don’t always produce different results - for example, in a 6-6-3-3 points group, the team that won the game between the two top teams would finish top both on a head-to-head tiebreaker method and on an SB score method. They would have 9 SB score (beating teams with 6 points and 3 points) as opposed to 2nd place’s 6 SB score (3 points and 3 points). However, as seen above, they do usually produce different results for teams lower down.
I’ll leave it up to you for the moment to decide if you like Paraguay’s case or Turkey’s case more. I should note that this SB score method is of course different to the old goal difference method - the SB score method is instead a different way to the head-to-head method of relying on results before considering margins.
3 Team and 4 Team splits
It’s also worth mentioning how this would work if 3 teams were all on the same points. There are a few ways this could happen, being 6-6-6-0, 5-5-5-0, 4-4-4-3, 9-3-3-3 and 9-2-2-2.
In all of these cases, each even team’s performance against either the top or bottom team would be disregarded. In each of these cases, the even teams must have an even points record against each other (as they either all won, all lost or all drew with the other team). The ranking would therefore be decided on goal difference in the matches between the even teams.
But in the 5-5-5-0 and 9-2-2-2 cases, the even teams all drew with each other! So then the ranking would come down to goals scored.
If all 4 teams are on the same points, then we clearly jump straight to considering goal difference and then goals for.
All 4 teams finishing on the same points has happened before at a World Cup:
Could a 4th Place team have a better record than a 3rd Place team that goes through?
Norway were unlucky to be eliminated as a team with 4 points. In this World Cup, the top 8 of 12 third place teams will go through to the Round of 32. I (and others) have previously calculated that you have a very good shot to advance with 4 points, and that 3 points with a decent goal difference (-2 or better) will probably be enough.
Indeed, Norway had a better record than some 3rd placed teams in their tournament. However, they would not have advanced to the next stage:
In this way it is very possible that there will be a 4th place team at this World Cup that has a better record than a 3rd place team that makes the Round of 32.
However, it actually goes further than this. It is entirely possible that there will be a 4th place team at this World Cup that has a better record than a 3rd place team IN THEIR OWN GROUP that makes the Round of 32.
Let’s suppose, again, that Turkey win their last game by a goal and Paraguay lose by a goal. Group D will then look like this:
Paraguay could go through to the next round as one of the 8 best 3rd placed teams. But, as far as the 3rd placed teams tiebreaker is concerned, Turkey would be ranked ahead of them (they would of course not be considered for the next round). It’s very possible that something like this will happen in one of the groups.
Should we go back to Goal Difference?
Overall, there are a few quirks that emerge from the new head-to-head system. But in comparing it to the alternative - which is the goal difference method, not the SB Score - I think it comes out slightly ahead.
Group stages purport to rank teams, and I think it is easiest to rank teams when we have a direct comparison between the two of them. It is very hard, in a small group, to complain that you should go through instead of a team that beat you only because they lost to someone else.
I also noted earlier that this change coincided with the introduction of the best 3rd place teams qualifying for the knockouts. In this way, the teams mainly affected by this change are 4th placed teams, who again should probably have just beaten the 3rd place team in the group. I’d be interested to see if FIFA gave any reasons for these changes coinciding - I wasn’t able to find anything myself.
There’s also an extent to which the head-to-head system incentivises aggressive play by a trailing team late in the game, because the reward of getting a draw is now increased in comparison to conceding an extra goal. I’d be surprised if teams are intentionally playing as if they’re aware of this, and I’m not sure this is completely good, but it probably does make for more exciting games.
Let me know if you have any thoughts or areas to explore in the comments below.
I’m surprised more people haven’t been writing about this. GD is a much fairer tiebreaker because it incorporates all group games, not one, and makes the third game more meaningful.
The new rules are demonstrably worse as will shortly be seen in the final round where those who have already "won" their group can play weakened sides so robbing this round of integrity. It's as if nobody at FIFA remembers the disgrace of Gijon in 1982...