better swiss system software question

[Uri Blass]

I wonder if there is a program that support better swiss system.

The problem with the swiss system is that the ranking at the middle of the table is sometimes not fair and 2.5 points against strong opponents out of 6 may be worth more than 3 points against weak opponents.

The question is if there is a software that rank the players not based on number of points and also later do the pairing not based on number of points when the idea is that players are going to play against players with similiar ranking so for example the leader with 7 out of 8 in the last round may play against strong player with 4.5 out of 8 and not against some weaker player with 5 out of 8 who played against significantly weaker opponents.

Of course players are not ranked at the end of the tournament based on number of points so the player with 4.5 who got the leader in the last round is not so unlucky because even a draw may give him better ranking than the ranking that he could get by beating some weaker player with 4.5 out of 8.

[hgm]

It would not be 'Swiss'.

But apart from that, it is easy enough: just run a rating calculator like BayesElo or Ordo on the tournament results, and apply the usual pairing algorithm to the ranking like the rating is the umber of points they scored.

[Ovyron]

The problem with the swiss system is that the ranking at the middle of the table is sometimes not fair and 2.5 points against strong opponents out of 6 may be worth more than 3 points against weak opponents.

But engines assume they're going to get 1.0 points for a win and 0.5 points for a draw, and they're tuned so that they increase the points they get regardless of the strength of the opponents. That is, they expect that they'll get more points against weaker opposition than against stronger opposition, if you limit the opponents that they face or award more points against stronger opponents it'd be unfair for them.

If you propose a different system then engines should be tuned to perform better in that system, and it may be different from the current system (i.e. if they never face weak opposition.)

[Branko Radovanovic]

The problem with the swiss system is that the ranking at the middle of the table is sometimes not fair and 2.5 points against strong opponents out of 6 may be worth more than 3 points against weak opponents.

How about the accelerated Swiss? It tends to initially pair strong players with other strong players, and weak players with other weak players, which should make the situation you describe unlikely.

In fact, the accelerated Swiss works precisely in the way you describe, i.e. by (initially) ranking the players not on the number of points.

[hgm]

The problem with the swiss system is that the ranking at the middle of the table is sometimes not fair and 2.5 points against strong opponents out of 6 may be worth more than 3 points against weak opponents.

But engines assume they're going to get 1.0 points for a win and 0.5 points for a draw, and they're tuned so that they increase the points they get regardless of the strength of the opponents. That is, they expect that they'll get more points against weaker opposition than against stronger opposition, if you limit the opponents that they face or award more points against stronger opponents it'd be unfair for them.

If you propose a different system then engines should be tuned to perform better in that system, and it may be different from the current system (i.e. if they never face weak opposition.)

This is not true if the result is based on Elo, and Elo is calculated in the usual way, where a draw counts for half a a win.