There is a to-and-fro middle game which ends up in this ending:

click for larger view

114.Kf5!▢

Soon aferwards we reach this well known drawn position,

click for larger view

So the improvement, as played, is really by the AlphaZero team, who determined opening play before unleashing the engines.

Here are the two games from the match (each engine get one chance at handling both sides of the board):

http://www.chessgames.com/perl/ches...

Here's the bookend game:

AlphaZero vs Stockfish, 2018

Of course <offramp>'s comment is interesting in a historical sense - but for exploring the opening play from here:

click for larger view

This should be obvious - and the AlphaZero team really should have released other games, if they indeed even had any, from the same position.

(OK, I have to reread their paper - but <CG> seems to have all the games A0 released)

But one has to question exactly why they explored this given position in the first place.

I'll have a look at the database stats in a future post (see last comment of this post, below).

But in the other game I commented negatively about the relevancy of the move 5...Nxd5. It turns out that Heisman (of the Dan H fame), actually published an entire book devoted to this move:

https://www.amazon.com/Fried-Liver-...

Whereupon we find two one-star comments, one of which I'll quote here:

<
C. Amari
1.0 out of 5 stars?!
August 25, 2008

This is an analysis of the Fried Liver that covers only one black option on move 5. There is no coverage of 5...Na5 or 5...Nb4, for instance. I'm not sure that the sole move addressed, 5...Nxd5 is even the most common. I suppose one can, by way of defining "Fried Liver Attack" narrowly, limit the scope of this exercise, but those who expect the analysis to begin at 4.Ng5 will be disappointed. Indeed, since there is no relevant comparative analysis at all, there is no real support for the view that 5...Nxd5 is even best. Hence anyone seeking a resource for Black to plan a response to 4.Ng5 (I suspect the motivation of many potential purchasers) will find that all the paths charted begin with a single step - and that may be a misstep.

Also, the interface with Chessbase Reader has some bugs in it.

>

In point of fact, my <MillBase> db shows only 2% of the Two Knight Fried Liver games have Black playing 5...Nxd5.

Also, the interface with Chessbase Reader has some bugs in it.>

<zanzibar>

I'm pretty sure that 5....Nxd5 6.Nxf7 <is> the Fried Liver. Even 5....Nxd5 6.d4 doesn't qualify, strictly speaking. The FL is just an (obsolete) variation of the Two Knights Defense. Readers can certainly question Heisman's judgment in writing a whole book about one (lousy) variation of the Two Knights, but having made that choice, he was pretty much obligated to leave 5....Nd4 and 5....Na5 (and 4....Bc5) out.

<Albion 1959: Why did they not pull the plug at move 118 ? Save everyone another eighty pointless moves !>

My mistake, and as you point out, 6.Nxf7 is the correct <Fried Liver> sequence.

The 6.d4 variation is the <Lolli>, I believe.

Still, the points I was trying to make remain true - Black shouldn't allow White to choose between the two (see the AlphaZero paper- this TCEC (2016) opening is one of the most lopsided).

I wonder what Heisman's rationale for his book was? I thought he was a pretty straight shooter for beginners - and this stuff should be avoided, unless you want a lopsided game.

(Which, btw- was part of the rationale for the change of openings in TCEC (2016). These openings were chosen, somewhat ad-hoc, by venator (aka Jeroen Noomen)

http://rybkaforum.net/cgi-bin/rybka...

<

Now it is time to tell you more about my opening selection for this final. Here is my summary:

"My first intention for the TCEC 9 super final was to use short, balanced opening lines. While making the set during the TCEC 8 super final (played in 2015), I changed my mind, though. In this final it became clear that balanced lines were leading to relatively level and quiet games and the number of wins turned out to be very limited: only 11 wins, the remaining 89 games being drawn.

During the TCEC 8 final many people were asking for more imbalanced positions, to get more exciting chess and more decisive results. I decided to listen to the suggestions in the chat and use them in the opening set for the TCEC 9 super final. Here is a small selection of the suggestions that I used for the set:

• Play sharp main lines
• Play gambits
• Play shorter lines, but also longer lines if this is necessary to get an interesting game
• Play positions with opposite castling to create an asymmetrical position
• Play positions with other imbalances, like f.e. weak pawns, pair of bishops, more space, non-symmetrical pawn structures
• Play positions that are positionally more difficult for engines
• Play the most topical opening variations to get as much variety as possible

A position from the final set had to met several criteria:

• No 0.00 score after leaving book
• No easy way to make a draw
• No easy way to exchange a lot of pieces quickly
• Most pieces should still be on the board

In the end I think I succeeded to assemble an interesting set, that can lead to a lot of exciting games. More exciting chess has been my main goal. The other one being (hopefully) more decisive results. Of course it takes two to tango, the engines must be willing to make a great fight of it! >

http://rybkaforum.net/cgi-bin/rybka...