As promised here is the first 10 game match between the Millennium ChessGenius Pro2024 and Chessnut Evo’s revised implementation of the Maia9 bot. More information about the Maia project can be found here https://maiachess.com/

In essence, the project has created nine sets of data each one targeted to a particular ELO rating and were created by training the neural networks with 12 million games for each set. The games come from human play at the chosen ELO rating. The idea behind the Maia project is zero is look ahead, and just using the data to pick the “human move” not necessarily the best move at the respective ELO level. The current highest ELO used for training the network is 1900 but as many of you are aware, the Maia does not reach this level of play and perhaps after making some reasonable moves will frequently blunder pieces and lose the game like a complete beginner and not the expected play from a human player at this level. The expectation might be that a neural network trained with 12 million human games between players rated at 1900 that the neural network would match this level in play, but trained with such data is one thing, playing at the same level is something else. 12 million games sound a large number, but dealing with the game such as chess with such a vast number of permutations and combinations then this figure is not so impressive and obviously leaves large gaps of knowledge in positions. For it to identify and avoid blundering pieces without looking ahead would surely require a magnitude higher level of data? Therefore, why not just implement a selectable ply search that takes just a few seconds? This is what Chessnut have implemented after receiving feedback from its customers such as myself. This feature as many will be aware has been implemented as a slider control allowing ply depths from 0 to 6. So we can have it play as it did before, blundering pieces or set a maximum depth of search. Setting a few ply search still results in near instant response but now Chessnut Evo’s Maia no longer blunders, in fact it plays rather well. It could be argued that it’s no longer following the concept of the Maia project with searching enabled, but I’m happy for the option to be added and have suggested to Chessnut that we have a greater range of ply settings for experimentation.

I decided to use the MCGP2024’s classic book for this match. I’m not sure what book was used for the ELO rating on the Wiki ELO page? I also opted for the Active time control on the MCGP2024. The Maia bot on the Evo does not use the clock for its thinking and just searches to the specified depth, thus for most moves the response is almost immediate or at most just a few seconds. I chose the classic book because with this the ChessGenius machines seems to perform better against other machines. However, here we are dealing with a neuronal engine so it may be a little different especially as Maia plays a little more human-like. The Maia does not have a book opening option, it just makes its moves from the data it was trained with, so in essence, knows the book openings from this and now aided by some searching to avoid pitfalls. I’ve mentioned this in a previous post but the Maia bot will always play 1. P-K4 when playing white, which is disappointing if we want more variety. That being said, it’s doing this because of the 12 million human games it was trained with, so not that unusual.

After that preamble onto the games:

Game 1/10



Although ending in a draw, Maia played very well in this game holding a better position for most of it. For example in the following position black had just moved 16…Qa5:

Black’s Queen has been pushed back to the side and Maia now moved 17.b4! Enclosing Black’s Queen. The position is very good for White with the Bishop on b2 commanding a good diagonal especially after the move d5 that comes later. The Maia bot (Lc0 engine with Maia weights file) is playing very well with good piece positioning and attacking style. However, despite having a Queen and 3 pawns against Rook and 5 pawns in the endgame, it only managed 3-fold repetition. Maybe with more depth of search it would have found a way to win?

Ray

Game 2/10

Here is the second game of my match between the MCGP2024 and Chessnut Evo Maia1900 with 6 ply look ahead.



Here the Bishop’s Opening, Berlin Defence was played until the MCGP2024 went out of book on move 6.Ng5

From here the MCGP2024 slowly but surely outplayed the Maia bot. There were no obvious bad moves from the Maia bot, just more accurate play by the MCGP2024.

Current score 1 ½ - ½ to theMCGPro2024

Game 3/10

The Maia bot shows what it can do in the shortest game of the match!



Here is the position after Black had just moved 6…h5

Here white replied 7.g5 which takes the MCGP2024 out of book. A better move may have been 7.Nh3 with the option of moving it to g5 or 7.Ne2 developing a piece.
In the played position neither side has developed pieces except for Black’s Bishop on g6. However, White’s advanced pawns especially the one on e5 needs some attention such as opposing it with 7…e6 Alas Black decided to develop his Knight with 7…Nd7 and Maia seized the opportunity by advancing the King’s Pawn 8.e6! From this point onwards Maia dominates the game.

Here is the position after White had just moved 12.Qxd3

Both sides have undeveloped pieces but at least White is able to develop them. Black’s King-side Knight, Bishop are stuck and even the Rook has little mobility. White’s black bishop and Knight on e2 are poised ready to swing into action and the Queen on d3 is going to cause some grief sooner rather than later. Black really needs to get the King to safety by castling Queen-side but alas moves 12…e5

Play continue 13.Qg6+ kd8 the only option 14.pxp Nxe5 15.Qf5 attacking the Knight and Bishop 15…Nd7 needed to protect the Bishop 16.Nf4 another piece joins the onslaught 16…e5 17.Ne6+ and a double attack 17…Ke7 Black’s position is not looking good at all!

18.Be3 threatening to move to c5 so Black should move 18…b4 18…Qb5 19.b4 preparing for the Bishop on c5 19…g6 The MCGP2024 cannot see a way out. A better move would have been 19…d4 and play might have continued with 20.Na3 threatening the Queen 20…Qa6 21.Bxd4 g6 complicating things for White and giving Black some chances.

From here Maia stormed into the enemy camp and the MCGP2024 was helpless so resigned.

An impressive game perhaps showing how the knowledge embedded within a set of numbers in the weights file coupled with some look ahead can produce good play.

Current score 1 ½ - 1 ½

Slight delay in writing up the next game but it's coming!

Game 4/10



The King’s Indian, Fianchetto, Classical main line was played and the MCGP2024 was out of book on move 7.

The game was all level at move 16 with the following position and White to move:

Here the MCGP2024 needed to protect the attacked knight on c3 and played 17.Qb3. Maybe the idea is to advance the c4 pawn opening up the diagonal to the enemy King but it does hamper the Queen. A better developing move might have been 17.Bb2.

It is from this point that Black made less than ideal moves until the following position after the MCGP2024 had just moved 48. Kc3

Here the Maia bot was a pawn ahead and in a better position but made the poor move 48…Kb5?
This gave White the option of drawing by 3-fold repetition which it wisely did. A better move for Black would have been 48…b5 with chances of advancing his queen-side pawns. The Maia bot is obviously lacking some knowledge here and missed its chance.

Overall an interesting game and we must remember that the Maia bot is limited to 6 ply search unlike the MCGPro2024 that was often looking much further ahead. That being said, Maia does have all the knowledge from 12 million games, although these were from 1900 rated players.

Current score all level at 2 – 2

Game 5/10



The Caro-Kann, Advance variation was played until move 7 when Maia played a dubious move. Here is the position after Black moved 6…h5

Here White really should have developed some pieces such as 7.Ne2 but instead moved 7. g5?!
We now have a repeat of game 3 and Black played the same move as before 7…Nd7 but needed to oppose White’s King Pawn to prevent it moving to e6.

Maia did indeed move 8.e6! Now in the same way as game 3 crushed the MCGPro2024 which resigned. It will be interesting to see what may happen in the next match when I set the MCGPro2024 to the modern (HIARCS) book.

Current score 3 – 2 to the Maia bot.

Game 6/10

The MCGPro2024 gets revenge. I will tell you now the match does not end in a draw, but will keep you wondering as to which machine won!



I have seen before how the Maia bot does not always follow usual opening moves, occasionally throwing in little surprises. In this game, the Queen’s Bishop Attack (Levitsky Attack) was played but on the second move Maia played 2…c5 instead of the preferred h6. But we also know that the MCGP2024’s classic book also has unusual lines. Perhaps this is Maia’s attempt to gain an early advantage by throwing the MCGP off its tracks? Well, obviously this tactic did not work in this game as we can see from the result.

Play followed 3.dxc5 and then Maia played 3…Qa5+?!

I think the Maia bot has an attacking style of play, but in this opening the played moved is a little premature. Better maybe would have been 3…h6 to ward off White’s aggressive Bishop.

In the following position Black had just moved 9…O-O giving the following position:

Despite the Maia making a questionable opening move with its Queen earlier, there is a possible Queen-side attack forming with the two Bishops, Knight and Queen; I’ve noticed that Maia can be very good in forming attacks, although as we shall see in this game it does not always work out.

White’s Knight on c3 is pinned but the MCGP2024 played 10.O-O-O moving its King to the Queen side which I felt was a little dangerous. Maybe better would have been 10.Ne2 supporting the pinned Knight on c3 and allowing for King-side castling. On the plus side, White now has a Rook on the open d file.
Play continued 10…Bb4 Deep HIARCS thinks 10.Ne5 would have been better but it was looking 20 ply ahead! 11. Nge2 Rfe8 maybe 11…f5 would have made more of a statement. 12.a3 and now the MCGP2024 is fighting off the attack and gaining an advantage. 12…Bxa3?! The Maia is aiming for some form of sacrificial glory:


Play continued 13. Bxc6 bxc6 14.bxa3 Qxa3+ and now Black is down a useful Bishop has double pawns on the f file and White can easily move away from danger.

The MCGP2024 continued to outplay Maia until the final checkmate, well done to the MCGPro2024!

Current score 3 – 3
A close match so far, but who will win in the end?

Game 7/10

The MCGP2024 shows Maia what it can do!



The Ruy Lopez, Classical Defence deferred was played until move 5 when the MCGP2024 played out of book. The Maia bot of course does not have a book opening as such, all its moves come from the neuronal weights file.

All was level at move 19 albeit with some previous inaccuracies, ending up in this position:

It’s Maia to play with the White pieces and here it played 20. e5? At first glance this may look ok until Black replies 20…Qe8. Now if White’s e5 pawn is moved, Black takes control of the e file with plenty of attacking options with its double Bishops in the opposite corner to the enemy King. A case maybe of the Maia looking for short-term gains and not seeing the dangers ahead!
Play continued 20…Qe8 of course. 21. Bxf6? Another bad move from Maia. After 21…Rxf6 e5xf6? Maia is lost! The e file is now open and the MCGP2024’s Queen is coming!
From here the Maia continued to make bad moves and the MCGP2024 punished it as a result. Well done MCGP2024.

Current score 4 – 3 to the MCGP2024

Game 8/10

The MCGP2024 outsmarts the Maia with the Ruy Lopez!



The Maia bot may not blunder pieces in an obvious way now that it has some look-ahead ability but sometimes still makes bad moves. Take the following position from the game:

The MCGP2024 was now out of book and had just taken the f5 pawn with 6.exf5. Black now needs to support the King Pawn that is currently only protected by the Knight on c6 that is itself under attack by White’s Bishop. I don’t know if Maia’s next move can be considered a more human move, I think not at this level of play! Here it somehow decides that castling would be good 6…O-O? It doesn’t require much look ahead to see that Black will lose its Bishop on c5. 7. Qc4+ The MCGP2024 seizes the gift! 7…d5 8. Qxc5 Bxf5 and now Maia is a Bishop down. So soon into the game and Maia is in a bad way.

Play continued 9. Bxc6 bxc6 10. Qxc6 and now Black has lost a Pawn.

A few moves later in the following position:

White had just moved 12. Ne2 and now Black’s Pawn on e5 can be captured with the Knight without worrying about it being pinned by Black's Rook. 12…Be4? Is Maia playing suicide chess!? Better would have been 12…Qd6. Play continued 13. Nxe5 Bxg2 14. Rg1 and White is in a good position with a strong King-side attack mounting:


From here the MCGP2024 cruised to victory in the second shortest game of the match.

The Maia bot has no clue how to play the Ruy Lopez, maybe this indicates the games it was trained with were not good examples or possibly more likely a magnitude higher number of games is required to handle complicated openings. That being said, considering its limited depth of search compared to the MCGPro that was always looking at least 1 extra ply ahead, and 4 or 5 extra in the end games, the Maia has not done too badly and as we have seen, has managed to outplay the MCGP in the least number of moves in this match.

Current score 5 – 3 to the MCGPro2024

Game 9/10

The Maia tests the MCGP2024 with the Ruy Lopez. What will the result be this time?



It seems the MCGP2024 manages much better than the Maia playing as black against the Ruy Lopez. Instead of losing badly manages a draw.

In the following position Maia had just moved 11. d5 and the position is level:

Here, the MCGP2024 played 11…Nd4 better would have been 11…bxa4 12. dxc6 Ne7 and White would have been a pawn down. Also, White having two bishops may prove more useful. However, play continued with White having a small advantage but all was level again at move 30…Nf5+

After a few more moves the game ended with 3-fold repetition.

Current score 5 ½ – 3 ½ to the MCGP2024
Whatever the result of the next game Maia has lost the match!