Opening Variety is dead

[Ovyron]

Just one year ago there were exciting times for opening theory. Alpha Zero and Leela were proving that many variations were playable, others were uncovered as traps that A/B engines would fall into. It was a renaissance of exploration and discovery, old books would die on the spot when facing new ones, and with the huge leaps of Stockfish keeping up with the exponential nature of chess, people were playing mostly anything and even claiming some mainline openings were busted.

So how's everything in 2020? Well, most of that was abandoned. Mostly everything became the same. What did it became?

From the last 26 games I played at Playchess, 50% of them were the D43 ECO! And 73% of the total was in the D43-D46 ECO!

This is the variation that ate up chess:

1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Nf3 c6

[d]rnbqkb1r/pp3ppp/2p1pn2/3p4/2PP4/2N2N2/PP2PPPP/R1BQKB1R w KQkq -

Basically, you can build a book that has only this exclusively, and be able to perform well, because two thirds of your opponents wouldn't even notice you didn't cover anything else!

Nobody was able to improve over this for either side, so this became chess' ultimate mainline.

From all the alternative moves that were had, White can't do nothing better than bringing its knights out without blocking the C one, and Black can't do nothing better than overprotecting the d5 pawn with two pawns and bringing the F Knight up.

Against 1.d4, 1...d5 finally won the battle against 1...Nf6, but most shockingly of all:

1.d4 won the battle against 1.e4. With what's out there currently even 1.c4 is winning the battle against 1.e4.

The D43-D46 ECO lines have become increasingly popular with no sign of stopping, being adopted as the mainlines of more and more books, I'd not be surprised if one month from now they reach 80% of the games, that's 4 out of 5 games being like this!

I wonder if this is the beginning of the death of chess, and the best 4 chess moves from both sides have already been solved and are on this post. Because if they weren't, surely someone would have found a way to defeat these moves (since they're the most popular you work on them and you become a winner) and with the anti-Semi-Slav lines being known people would stop using them and other lines would get played.

Instead, this is like a black-hole that is sucking all the opening variety of chess. You can get lazy and adopt these moves and just Copy-Paste what others play against you on these lines. Who's gonna stay for that?

[the_real_greco]

Remember when poker was just people playing super tight? And then it got flooded with people betting super crazy all the time because they decided that was the best way to win?

That's what's going to happen here.

Long Live Opening Variety!

[Ovyron]

Yes! Stop playing D43-46 people! Think of the future of chess!

[Colin-G]

Out of curiosity, I compared the position against the 38,500 engine-engine match games in my Scid database that have been played on my computers over the last 20 years.
The position occurs 1881 times which is 4.9% of the games.
While Stockfish prefers the most common next move of e3, both Bg5 and Qb3 have been more successful.
See the Scid Tree window below

[Ovyron]

While Stockfish prefers the most common next move of e3, both Bg5 and Qb3 have been more successful.

Thanks for the report. Apparently Bg5 and Qb3 were "busted" in 2019, so today's books are full of black variations that equalize against them without problems. Only e3 remains with winning chances, though from what I've seen, it's more like e3 is the only one remaining with traps to catch unsuspecting opponents.

I have Bg5 and Qb3 at 0.00.

Oh, just remembered ChessDB, I wonder what does it say about this?



It also claims that Bg5 is better than e3. I wonder if Bg5 can fall below e3 if I feed the DB the variations from the 2020 opening books.

I also thought it'd be interesting to check its disagreements with current theory:



Does not agree that Nf3 is best. Actually cxd5 is very interesting, engines like it as well and give it high score even from the root. I don't think it has been explored fully yet.



Here it disagrees that Nf6 is the best defense, and thinks c6 first would be a better move order.



Finally, against 1.d4 ChessDB disagrees that 1...d5 has won the battle against 1...Nf6.

I'll be feeding variations from current theory into ChessDB and see what happens. I'd suck if I can just refute its attacks and make D43 the queen of chess. But maybe I get it refuted and become the winner I was talking about

[Ovyron]

Oh well, here's the black moves that refute ChessDB's Bg5:



The DB doesn't see any better for white than this 3-fold repetition. Yet it scores it as 0.10, so no luck making it show it as 0 to make Bg5 score better.

So never mind.

[Uri Blass]

Just one year ago there were exciting times for opening theory. Alpha Zero and Leela were proving that many variations were playable, others were uncovered as traps that A/B engines would fall into. It was a renaissance of exploration and discovery, old books would die on the spot when facing new ones, and with the huge leaps of Stockfish keeping up with the exponential nature of chess, people were playing mostly anything and even claiming some mainline openings were busted.

So how's everything in 2020? Well, most of that was abandoned. Mostly everything became the same. What did it became?

From the last 26 games I played at Playchess, 50% of them were the D43 ECO! And 73% of the total was in the D43-D46 ECO!

This is the variation that ate up chess:

1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Nf3 c6

[d]rnbqkb1r/pp3ppp/2p1pn2/3p4/2PP4/2N2N2/PP2PPPP/R1BQKB1R w KQkq -

Basically, you can build a book that has only this exclusively, and be able to perform well, because two thirds of your opponents wouldn't even notice you didn't cover anything else!

Nobody was able to improve over this for either side, so this became chess' ultimate mainline.

From all the alternative moves that were had, White can't do nothing better than bringing its knights out without blocking the C one, and Black can't do nothing better than overprotecting the d5 pawn with two pawns and bringing the F Knight up.

Against 1.d4, 1...d5 finally won the battle against 1...Nf6, but most shockingly of all:

1.d4 won the battle against 1.e4. With what's out there currently even 1.c4 is winning the battle against 1.e4.

The D43-D46 ECO lines have become increasingly popular with no sign of stopping, being adopted as the mainlines of more and more books, I'd not be surprised if one month from now they reach 80% of the games, that's 4 out of 5 games being like this!

I wonder if this is the beginning of the death of chess, and the best 4 chess moves from both sides have already been solved and are on this post. Because if they weren't, surely someone would have found a way to defeat these moves (since they're the most popular you work on them and you become a winner) and with the anti-Semi-Slav lines being known people would stop using them and other lines would get played.

Instead, this is like a black-hole that is sucking all the opening variety of chess. You can get lazy and adopt these moves and just Copy-Paste what others play against you on these lines. Who's gonna stay for that?

There are many way to get a draw so I see no best moves that most people are going to play

If I look at the 21 olympiad that started in 2020

https://www.iccf.com/event?id=83284

I see at board 1 the following openings so far for finished games:

1.d4 d5 2.c4 c6 3.Nf3 Nf6 4.Nc3 e6(draw after move 39 of white,draw after move 31 of black)
1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3 c6(same position but different order of moves draw after move 28 of black)

one position appeared 3 time.
rest of the positions after 4 moves appeared only one time in finished game so far:

1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.f3 0-0(draw afte move 21 of black)
1.d4 d5 2.c4 c6 3.Nf3 Nf6 4,e3 e6(draw after move 20 of white twice)
1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6(draw after move 23 of black)
1.e4 c5 2.Nc3 Nc6 3,Nf3 g6 4.d4 cxd4(draw after move 33 of white)
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Nf3 0-0(draw after move 30 of white)
1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 Nf6(draw after move 27 of black)
1.Nf3 Nf6 2,c4 e6 3.Nc3 d5 4.d4 Bb4(draw after move 18 of white)
1.d4 Nf6 2.c4 g6 3.Nc3 d5 4.cxd5 Nxd5(draw after move 29 of white)
1.e4 e5 2.Nf3 Nf6 3.Nxe5 d6 4.Nf3 Nxe4(draw after move 24 of black)
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Qc2 0-0(draw after move 14 of black)

[Ajedrecista]

Hello:

It is not new at all. These lines are quite logical and I play them online (human chess, not engine vs engine) most of the time, so the first guilty is here!

I think there was a time where PlayChess games were Sicilian Najdorf (ECO B90) most of the time, probably ten years ago. Am I wrong?

It also happened at OTB chess: the 1927 WCC between Capablanca and Alekhine featured 32 QGD out of 34 games:

Capablanca vs Alekhine 1927

------------

Moreover, it is not exclusive in chess. It also happened in English draughts, specifically in a match of 1863 between James Wyllie and Robert Martins, two of the greatest draughts players of that time:

The Perfect Game?

The opening that was played 21 times out of 40 games was played exactly move by move the 21 times according with the records. It was a Glasgow opening (draughts openings are also named) and the variant that was repeated over and over became so famous that it was nicknamed "Martins' Rest":

https://www.fmjd.org/docs/article.pdf

Martins' Rest. The game below was played 21 times out in a 40-game match between James Wyllie and Robert Martins in 1863. It led to the introduction of the Two-Move Restriction is Scotland in 1890, followed by the Three-Move Restriction in the USA in 1934. These measures were designed to prevent excessive repetition of games by forcing players to play unfamiliar openings.

So Martins' Rest had consequences after all.

Other source of Martins' Rest:

American Checkers Federation Bulletin #322, August 2006

[...] Interesting was their copy of the 1863 Martins - Wyllie WCM, the infamous Martins' Rest match which shot down GAYP for a long time. Their copy is handwritten by R. Martins, in a small notebook. The players played interesting games for six days, and 17 games. Then both started repeating two earlier Glasgows, with the last 20 games (four straight days) being identical Martins' Rests, played by BOTH players. The final score: 49 draws, with the 50th game not played, Martins retaining his Title. It was very strange.

As you can see, the number of games played and repeated depends on the source. I think that I read somewhere that one of the players had other business elsewhere and it was a quick way to finish the match, but the spectators (and their bets) were disappointed or even angry. I am not able to find that story now, it was quite interesting.

------------

So I expect opening repetitions from time to time.

************************

Going a little off-topic, I will write a little about draw percentage in certain openings. Much has been written about Petrov's Defence, but what about Slav Exchange (ECO D14)?

[pgn]1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. cxd5 cxd5 5. Nc3 Nc6 6. Bf4 Bf5 7. e3 e6 8. Bd3 Bxd3 9. Qxd3 Bd6 10. Bxd6 Qxd6[/pgn]

1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. cxd5 cxd5 5. Nc3 Nc6 6. Bf4 Bf5 7. e3 e6 8. Bd3 Bxd3 9. Qxd3 Bd6 10. Bxd6 Qxd6

Looking right now at ChessGames Opening Explorer:

Code: Select all

Chess Opening Explorer. Chessgames (4th June 2020):

             Games     WW(%)     D(%)     BW(%)
 1. d4      317,857     37.7     35.7      26.6
... d5       84,892     40.0     36.3      23.6
 2. c4       63,505     40.0     38.0      22.0
... c6       31,194     38.4     40.1      21.5
 3. Nf3      20,691     39.2     39.7      21.1
... Nf6      26,366     38.4     40.2      21.5
 4. cxd5      1,214     21.6     59.8      18.6
... cxd5      1,670     21.9     58.1      20.0
 5. Nc3       2,329     22.2     58.2      19.6
... Nc6       2,434     21.0     59.7      19.4
 6. Bf4       2,421     21.4     60.6      18.0
... Bf5       1,050     17.0     69.3      13.7
 7. e3        1,096     15.6     70.1      14.3
... e6        1,022     13.8     72.2      14.0
 8. Bd3         456      2.2     93.6       4.2
... Bxd3        433      1.8     94.2       3.9
 9. Qxd3        450      2.2     93.6       4.2
... Bd6         360      0.8     98.6       0.6
10. Bxd6        335              99.1
... Qxd6        332              99.1          

I hope no typos. Over 99% draws with more than 300 games played! Similar stats can be extracted from other databases, for example from Shredder web online opening database:

Code: Select all

Shredder web online opening database. Huge Book. (4th June 2020):

             Games    WW     D     BW
[...]
 8. Bd3       201      9    151    17
... Bxd3      170      8    146    16
 9. Qxd3      170      8    146    16
... Bd6       131      1    128     2
10. Bxd6      120      0    119     1
... Qxd6      121      0    120     1
11. O-O       123      0    122     1
... O-O       123      0    122     1

Please take a look at these stats. This is what we are approaching with top engines (at least at slow TC and big hardware), just like English draughts programmes, mainly KingsRow and Cake. Don't you think so?

Regards from Spain.

Ajedrecista.

[Terje]

Oh well, here's the black moves that refute ChessDB's Bg5:



The DB doesn't see any better for white than this 3-fold repetition. Yet it scores it as 0.10, so no luck making it show it as 0 to make Bg5 score better.

So never mind.

The problem with repetitions in ChessDB is a bit annoying, yeah. Where would one find these opening books, are they (or some subset) free? I like to explore lines in CDB to improve it from time to time, but as I know very little theory I doubt I'm really helping much - would be cool to have some better stuff to feed it

[Ovyron]

It is not new at all.

What is new to me is *the volume*. I don't recall seeing a percentage of samey things in the history of computer chess theory. And the worst part is that if you ignore those D43-46 ECO games, then most of the rest is eaten by D35 (and D36-D38).

I think there was a time where PlayChess games were Sicilian Najdorf (ECO B90) most of the time, probably ten years ago. Am I wrong?

Was the volume of the Sicilian Najdorf of 2010 this high? If so this is a fad and we'll soon see other openings rise.

What I'm seeing in the current level of competition (for 5 minute games, anyway) is that engines and NNs have become so strong that any little mistake is punished. Any suboptimal move can lead to a loss, so people can't afford to play any opening that has a worse performance than other ones, so they mark them red, and the variety shrinks. If my perception is true then I don't see how we'll get out of the hole, people losing their games with other variations will make them disappear.

(today I lost 3 games against D46 and 1 in D43 so my opponents keep showing why they plan them...)