Stale Chess (variant idea)

[Ian Thompson]

You have been asserting this for many years. Cam you explain why? Why does the predicted result in terms of wins, draws and losses vary with whether you score 3-1-0 or as at present? If you started to use 3-1-0 in rating calculations, you would break the Elo system, but why otherwise?

If 3-1-0 makes no difference to the pattern, why bother with it? Of course, if people play properly, it will make a difference. Even Bogdan might become more aggressive when white. Anyway, if you don't think much of that, try 4-1-0.

I think most people would agree that 3-1-0 might encourage more attacking play, with the result being more wins and losses than under 1-0.5-0. Why would it change the overall expected performance?

If the stronger player is expected to score 60%, under 1-0.5-0, over 10 games, the results might be 3 wins, 6 draws, 1 loss. Under 3-1-0, the results might be 5 wins, 2 draws, 3 losses. What reason is there to think that the performance will be something other than in this ratio?

[Christopher Kreuzer]

A way to even things up: 1 Nc3 Nc6 2 Rb1 Nb8 3 Ra1 Na6 4 Nb1 Nb8. According to Fritz, the position is now equal. Moreover, an arbiter glancing casually at the position, would think it is the normal one.

Had to read that twice before realising what it was about. You are saying that disallowing White from castling queenside evens the game up. Interesting. What if White is not allowed to castle kingside, or is not allowed to castle either side. What do the silicon monsters say then?

Hmm. The 16 (2^4) permutations are:

(1) Both White and Black can castle both sides.
(2) White can castle both sides, Black can only castle kingside.
(3) White can castle both sides, Black can only castle queenside.
(4) White can castle both sides, Black cannot castle.
(5) White can only castle kingside, Black can castle both sides.
(6) White can only castle queenside, Black can castle both sides.
(7) White cannot castle, Black can castle both sides.
(8) White and Black can both only castle kingside.
(9) White and Black can both only castle queenside.
(10) White can only castle kingside, and Black can only castle queenside.
(11) White can only castle queenside, and Black can only castle kingside.
(12) White can only castle kingside, and Black can't castle.
(13) White can only castle queenside, and Black can't castle.
(14) White can't castle, and Black can only castle kingside.
(15) White can't castle, and Black can only castle queenside.
(16) Neither side can castle.

All from the starting position, with White to play the first move.

I suppose someone has already done extensive evaluation of all 16 permutations.

Or maybe not.

[NickFaulks]

One reason 3-1-0 has never gained popularity is that Geurt Gijssen and I were against it. We feared, if used in Swisses, there would be more cheating by collusion in the last round.

Why just in the last round?

I believe on the games on ChessBase, it is about 51.5% to 48.5%.
Without stalemate I reckon a minimum of 55% to 45%. No proof of course.

But those are human games. We're talking about computers, which (allegedly) play (almost) perfectly.

[Stewart Reuben]

Because players who take greater risks would be rewarded relative to players who take fewer risks.
Just consider the profiles of Bogdan Lalic and Simon Williams.
Players with Black, lower rated than their opponents, would be happy to play the Berlin Wall, in the hope that their opponents would try too hard to win. Since it changes how people approach their chess, why would it not be likely to change the ratio?
Try it another way. Two people play against the same field. A has drawn 8 games and won 1.
The other B has won 5 games and lost 4.
In the current system, both score 5/9. In 3-1-0, A scores 11 and B 15. He wins a bigger prize. They both started with the same rating. Why shouldn't B's rating go up relative to A's?
One of the purposes of a rating system is making predictions. Very useful when betting. The rating should reflect the profile of each player. It would require two numbers for 3-1-0. One to reflect the percentage scored, as currently. One to reflect the percentage of draws.

Christopher K. I researched it solely on my, out of date version, of Fritz 13.
1. Both players can castle both sides: White +0.3
2. White can only castle queens, Black can castle both ways. Level.
White cannot castle either wing. Black can castle both ways. Black +0.3.
I didn't go any further. Be my guest.

Nick Of course there may be collusion in earlier rounds. But it is much easier to see the benefit in the last. Collusion used to be the most common form of cheating internationally. It has been over-shadowed by fear of people using computers. There is also sand-bagging.

[Roger de Coverly]

Why shouldn't B's rating go up relative to A's?

Because according to the mathematics underpinning the Elo system they are the same strength of player.

[Michael Farthing]

Well exactly - which is why the ELO system would need amendment in such circumstances It would no longer be fit for purpose.

[NickFaulks]

Because according to the mathematics underpinning the Elo system they are the same strength of player.

But what does "strength" mean when the target is asymmetrical? That's why we don't rate Armageddon games ( aside from the time difference ) and why I wouldn't want anything to do with 3-1-0 either.

[Stewart Reuben]

Nick >
But what does "strength" mean when the target is asymmetrical? That's why we don't rate Armageddon games ( aside from the time difference ) <

The two numbers that would be needed for 3/1/0 would help you establish a probale result between two players. I presume the term 'stronger' player would mean one likely to make a bigger score in a tournament.
We don't rate Armageddon games because they are not played according to the FIDE Laws of Chess. Armageddon is a chess variant.
Should we rate 3/1/0 games in the Elo system? No, but there are so few of them we don't care.
Should 3/1/0 be considered to be chess and in the FIDE Laws? Probably not, but I never thought about that when Mike Basman asked me to get the system recognised by FIDE.
CHESS 960 Rules are in the FIDE Laws because it is a convenient place to keep them. They shouldn't be there.
From 2021 only games with increments will be FIDE Rated. That is in order to eliminate quickplay finishes.

[NickFaulks]

From 2021 only games with increments will be FIDE Rated. That is in order to eliminate quickplay finishes.

You keep saying this. Is anyone else privy to this information?

[Roger de Coverly]

I presume the term 'stronger' player would mean one likely to make a bigger score in a tournament.

I would have thought it meant the one most likely to score a higher percentage, ignoring win/draw ratios against the same field. With that definition 3-1-0 scoring is irrelevant. That's how the Elo mathematics are derived.

If you want to predict the expected score using 3-1-0, you need a new model of rating the ability of chess players, one that has win/draw ratios as something that's measured.

[Stewart Reuben]

Nick >You keep saying this. Is anyone else privy to this information?<

How about Makro? Of course, it may have vanished from proposed statutes, or it may be buried somewhere in the FIDE Minutes.or in the minutes of the QC when I was secretary.

Roger >
If you want to predict the expected score using 3-1-0, you need a new model of rating the ability of chess players, one that has win/draw ratios as something that's measured.<

Precisely. You agree with me. The current FIDE Rating system would be inadequate. It would be more complex. What is the point of devisining such a system when very few tournaments are played that way? The only substantial event I know of that uses 3/1/0 is the Football League. When they started usining it, there was no appreciable increase in the number of positive results. I don't knw about now,

The current system used is NOT the ELO Rating System. His was designed for round robins with a mnimum rating of 2200 and tied in with the title norm system. Let us take a 9 game RR. Take the average of your opponents' ratings, Determine We. W -We gave the rating change. But it was illogical as the rating to expectancy is not a straight line. With the help of Vishy Anand, I got them to change to rating each game separately. Makro was flattered that a strong player had shown an interest.

[NickFaulks]

Of course, it may have vanished from proposed statutes, or it may be buried somewhere in the FIDE Minutes.or in the minutes of the QC when I was secretary.

Wherever it may have been or not been, it certainly isn't there now.

[Roger de Coverly]

When they started usining it, there was no appreciable increase in the number of positive results. I don't knw about now,

The supposed theory is that the potential loss of giving your rivals 3 points offsets your own loss of only scoring 1 point.

The Basman system more often than not awards the winner's prize to the player with the most losses. Thus played 6, won 5 lost 1, wins against played 6, won 4, drawn 2.

[Stewart Reuben]

Roger >The supposed theory is that the potential loss of giving your rivals 3 points offsets your own loss of only scoring 1 point.
The Basman system more often than not awards the winner's prize to the player with the most losses. Thus played 6, won 5 lost 1, wins against played 6, won 4, drawn 2.<

The first sentence is very puzzling.e
The second paragraph relates to a tiebreak system, not to 3/1/0. The late Mike O'Hara was first to point out the defect of that particular tiebreak system.

[IM Jack Rudd]

For sufficiently short tournaments, 3/1/0 is equivalent to the normal scoring system with "most wins" as the first tiebreak. (It made a difference at Euro 2016, for example, in that Wales thereby finished above England, but it didn't make any difference* at the previous World Cup, or at Euro 2012.)

*As in: the same results would have produced the same finishing orders of the teams. It may, of course, have affected how teams approached the matches.