Firscher was right about FischerRandom

[lkaufman]

That game has ended. It's now 11 White wins 1 Black win.

I think you get my point though - where things like patterns are largely removed from the game (not just memory) the 'First Serve' (to use the tennis analogy) gains even more importance. Engine chess might be one thing, but we are talking about human chess here.

I'm not sure which games you are counting, you must be including some rapid games.....

....I will admit that the very variable White advantage is an argument against this format. The ideal format in my thinking would be a double round robin at the slowest time control that permits two rounds per day, probably one hour plus 30" increment. Colors would reverse for the second game with the same start position so it wouldn't matter how much advantage White had, with only a short supervised break between rounds so the players couldn't consult engines (or friends who used engines). Although the time limit would be faster than true classical events, there would be no need for Rapid playoffs (except perhaps to break a tie at the end of the event) so the average time limit would be similar to this event.

Yes, I see now when looking at the chess.com results during a break from work I did not see the full results to that point.
It was actually White: 12 Wins, 4 Losses and 6 draws at that point.
Then of course came the fast blitz and armageddon where Black won all 3.

Listening to the commentary of the non-blitz games, the commentators were often talking about how much of an edge White was getting out of the openings. Even Fabi in the post interview noted how fortunate he was to have White in an important game because black starts off so badly and may be losing. So yes, I think the only rational thing would be alternating colors with the same position.

If there is a 'draw problem' at the highest level, it has next to nothing to do with the nature of the classical game itself.

Apparently you are including Rapid games. After today's games (2 wins and 2 draws), it still remains true that there has not been a single classical time limit game (out of 20) which was drawn without either player reaching a likely won position at some point (per stockfish 16), as both of today's draws involved missed wins (in one case multiple wins by both sides!). With allocating the first missed win to the side who missed it, the score would now be 10 to 10 (White vs Black!). I'm sure this is a fluke, in the long run it should be a clear White plus.

[Modern Times]

DFRC is not a useful game because many positions are not just "advantage" for White but easily won at high level. Of course they could be filtered out. We don't need huge numbers of positions for tournaments or rating lists, 960 is plenty, but we do need ones with evals closer to (but preferably below) 1. Your 324 openings do this, and that is the best we have right now, but it's not the game humans are playing (unfortunately). Maybe some subset of DFRC with a simple rule to define the subset might accomplish what I'm aiming for, I might give that some thought.

Yes I had thoughts a while ago of filtering the list of 900k+ DFRC positions, to produce a subset of openings for example in the range of say +090_+119 but it was too big a job for me. As you say though, neither Chess324 nor DFRC are routinely played by humans so it is just for computer chess.

[lkaufman]

DFRC is not a useful game because many positions are not just "advantage" for White but easily won at high level. Of course they could be filtered out. We don't need huge numbers of positions for tournaments or rating lists, 960 is plenty, but we do need ones with evals closer to (but preferably below) 1. Your 324 openings do this, and that is the best we have right now, but it's not the game humans are playing (unfortunately). Maybe some subset of DFRC with a simple rule to define the subset might accomplish what I'm aiming for, I might give that some thought.

Yes I had thoughts a while ago of filtering the list of 900k+ DFRC positions, to produce a subset of openings for example in the range of say +090_+119 but it was too big a job for me. As you say though, neither Chess324 nor DFRC are routinely played by humans so it is just for computer chess.

I had a new idea today. Suppose we just use the subset of chess960 positions with the kings on their normal squares, e1 and e8. I don't know an easy way to count them, probably there are around 200 such positions (if anyone can easily get the precise number, please post here). These positions will all seem much more "normal" to human players, with FRC castling always meaning a two square king move with the rook hopping (or remaining on) the passed-over square. This should minimize concerns about the weird castling and the lack of familiar patterns, while still having enough variety to make opening memorization of limited practical value, and as a subset of 960 it can be considered a humanly played game. This won't help with the engine draw problem, but perhaps DFRC with this king requirement will minimize the number of positions that are clearly won, though I'm sure there will still be some. I expect that many DRFC positions with kings on e file will have evals close to 1, so it could be a valid game for engines without any books needed if there aren't many +2 or more positions.

[Uri Blass]

DFRC is not a useful game because many positions are not just "advantage" for White but easily won at high level. Of course they could be filtered out. We don't need huge numbers of positions for tournaments or rating lists, 960 is plenty, but we do need ones with evals closer to (but preferably below) 1. Your 324 openings do this, and that is the best we have right now, but it's not the game humans are playing (unfortunately). Maybe some subset of DFRC with a simple rule to define the subset might accomplish what I'm aiming for, I might give that some thought.

Yes I had thoughts a while ago of filtering the list of 900k+ DFRC positions, to produce a subset of openings for example in the range of say +090_+119 but it was too big a job for me. As you say though, neither Chess324 nor DFRC are routinely played by humans so it is just for computer chess.

I had a new idea today. Suppose we just use the subset of chess960 positions with the kings on their normal squares, e1 and e8. I don't know an easy way to count them, probably there are around 200 such positions (if anyone can easily get the precise number, please post here). These positions will all seem much more "normal" to human players, with FRC castling always meaning a two square king move with the rook hopping (or remaining on) the passed-over square. This should minimize concerns about the weird castling and the lack of familiar patterns, while still having enough variety to make opening memorization of limited practical value, and as a subset of 960 it can be considered a humanly played game. This won't help with the engine draw problem, but perhaps DFRC with this king requirement will minimize the number of positions that are clearly won, though I'm sure there will still be some. I expect that many DRFC positions with kings on e file will have evals close to 1, so it could be a valid game for engines without any books needed if there aren't many +2 or more positions.

king is at e1.
Now there are some cases for the rooks:

case 1:rooks at black squares(2 options a1,g1 and c1,g1)
In this case one bishop has to be at a1 or c1(the only remaining black square) and we have 4 remaining white squares for second bishop and 3 remaining white squares for the queen(total number of 2*4*3=24
case 2:rooks at white squares or different color of squares.
total number of options:4*3-2=10
4*3 is because rook at a1-d1 is 4 options and rook at f1-h1 is 3 options but I do not count case 1 so I reduced 2.

In this case bishops have 3*2 options and queen has 3 options.

I get 10*3*2*3=180

180+24=204

[Uri Blass]

Based on the 12 Classical time limit games played so far in the oddly-named "Freestyle Goat Challenge" event of Chess960 aka FischerRandom, it does appear that for human play 960 solves both problems of playing by memorized computer analysis and excessive draws at Classical time limits. Officially, 9 of the 12 games were decisive, but one of the 3 draws was agreed in a winning position just to clinch the match, so really 10 out of 12 decisive, with all the participants at the very top of World chess! Many games reach a winning position for one side by move ten or so, not due mainly to White's initial advantage but just to the difficulty of figuring out the way to play in the opening without assistance of engines or memorized theory. Basically I think we are seeing chess as it was in the 1800s again, when little was known about how to play the openings properly. I would expect that this event will lead to increased popularity of 960 (the new name freestyle is already used in chess for engine-assisted play, so it is confusing).

Unfortunately it is still too drawish for top engines with many cores at non-blitz time limits. We do have the chess 324 variant that is substantially less drawish. Perhaps we can find other ways to make 960 work just as well for engines as for humans. The problem is that White's advantage is not even halfway to the win/draw dividing line in most or all of the positions. Perhaps some limitation on Black's castling rights or on allowed opening moves (for example no copycat first move or no double pawn first move or first two moves) would be enough to get closer to the win/draw line which would make 960 fully viable (without need for opening books) for engine tournaments and rating lists. Many ideas to consider.

I think that one of the reasons for not many draws in FRC between humans is that humans did not invest most of their time in FRC but in normal chess.

If top players decide to invest most of their time in FRC then I guess that FRC at classical time limits will have more draws.

[chesskobra]

I think that one of the reasons for not many draws in FRC between humans is that humans did not invest most of their time in FRC but in normal chess.

If top players decide to invest most of their time in FRC then I guess that FRC at classical time limits will have more draws.

I agree with this. Once FRC becomes a serious business with money, top players will get back to their preparation routine, and game the system so to speak. The problem of opening preparation in human chess must be solved once and for all. FRC will only solve it temporarily.

[Nordlandia]

Is chess18 or chess324 with randomized castling rights something to think about, it might be something

-
k
q
kq
K
Kk
Kq
Kkq
Q
Qk
Qq
Qkq
KQ
KQk
KQq
KQkq

[towforce]

To be more precise, the problem is not that it is drawn with perfect play, but that the drawing margin is too wide. GO with the proper integer Komi (6 or 7 maybe depending on rules) is also theoretically drawn, but the margin is so slim that it is fully playable even with today's super engines. Chess could also be like this with some changes, either to rules or to initial position (and Armageddon rule).

Very well put!

There are many differences between chess and go - but regarding the draw margin, the key difference is that the branching factor is much higher in go than it is in chess: the lower the branching factor, the easier it's going to be to find the optimal solution, especially for computers - but also for humans as well.

In the end, if you want computer chess to have fewer draws, you'll need bigger boards. This would solve the problem, but would make the games too long for most people to be willing to play the game - or even watch it. Cricket has an event called "5 day test match", but the game has mostly been moving to shorter formats in recent decades. Maybe leisure time will start increasing, and events will start lengthening again - but there's no sign of that right now.

[lkaufman]

To be more precise, the problem is not that it is drawn with perfect play, but that the drawing margin is too wide. GO with the proper integer Komi (6 or 7 maybe depending on rules) is also theoretically drawn, but the margin is so slim that it is fully playable even with today's super engines. Chess could also be like this with some changes, either to rules or to initial position (and Armageddon rule).

Very well put!

There are many differences between chess and go - but regarding the draw margin, the key difference is that the branching factor is much higher in go than it is in chess: the lower the branching factor, the easier it's going to be to find the optimal solution, especially for computers - but also for humans as well.

In the end, if you want computer chess to have fewer draws, you'll need bigger boards. This would solve the problem, but would make the games too long for most people to be willing to play the game - or even watch it. Cricket has an event called "5 day test match", but the game has mostly been moving to shorter formats in recent decades. Maybe leisure time will start increasing, and events will start lengthening again - but there's no sign of that right now.

Bigger boards might solve the problem, but it is far from the only solution. The branching factor isn't the main reason for the wide draw margin in chess, it is the draw rules. In GO repetition is forbidden, in chess it is allowed, it could be prohibited, as it is in both Chinese and Japanese chess. Stalemate is logically a win for the superior side, it was changed to a draw apparently to INCREASE the draw margin, which now makes no sense. Bare king should also be a loss, as it was before the queen gained super-power. With those three changes, the draw margin would shrink dramatically and engine chess might be playable even from the initial position. If that's not enough, some tie-break rule for fifty move rule draws could be added.

[lkaufman]

I think that one of the reasons for not many draws in FRC between humans is that humans did not invest most of their time in FRC but in normal chess.

If top players decide to invest most of their time in FRC then I guess that FRC at classical time limits will have more draws.

I agree with this. Once FRC becomes a serious business with money, top players will get back to their preparation routine, and game the system so to speak. The problem of opening preparation in human chess must be solved once and for all. FRC will only solve it temporarily.

While I don't doubt that the draw percentage will rise as the players take it as seriously as they do standard chess (which now seems inevitable as the organizer is talking about millions of dollars in future prize money!!), it will never come close to what it is in standard chess. Each player will only know about 0.1% as much about each position as he does about the one standard position, given equal time spent. Sure, there will be new opening principles worked out, but such generalizations only go so far. I disagree about the "temporary" comment (unless you meant for engines); a thousand to one increase in the possibilities is not like just one new position to analyze, it is beyond any human capability to remember more than one or two main lines for each position. It's like in standard chess, how much of an advantage would you have if you knew the main line of the Berlin and no other openings vs an opponent who knew no openings at all? Maybe 2 or 3 elo advantage at most I would guess.

The tournament has ended, and after 24 games at classical time limits, not even one game remained within drawn bounds from start to finish (per Stockfish 16), every game should have been one by someone! To me that is astonishing. We in the chess engine community just have to find the best way to participate in 960.