Rook odds match: human vs engines.

[Chessqueen]

Yes, h1 is a larger handicap than a1, though not a not larger, for the obvious reason that castling kingside is generally preferred over castling queenside because the king on c1 needs to move again to b1 to be secure (generally speaking). All historical rook odds games are with a1 removed, never h1. In normal chess, if White can only castle queenside while Black can only castle kingside, current Stockfish rates the position as 50% chance for each side; White has lost his opening advantage. So roughly speaking the difference between h1 and a1 odds is about a half tempo or about 0.2 pawns.

Hi.

Ok, I understand.
But with such a large material advantage, I think that the player getting the handicap should not worry too much about the opponent castling or not castling.
What I have noticed in these games is that, although the rook is a very powerful piece, from the initial position it takes too long to become active, too many moves for its presence to be felt.
So, the priority is to open one file or two and activate the extra rook. Before that, the (weak) human player can lose due to any tactical oversight. After that, it sort of becomes easier... there are still tactical pitfalls, but somehow I feel that they are less disastrous.

Greetings.

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

[lkaufman]

Yes, h1 is a larger handicap than a1, though not a not larger, for the obvious reason that castling kingside is generally preferred over castling queenside because the king on c1 needs to move again to b1 to be secure (generally speaking). All historical rook odds games are with a1 removed, never h1. In normal chess, if White can only castle queenside while Black can only castle kingside, current Stockfish rates the position as 50% chance for each side; White has lost his opening advantage. So roughly speaking the difference between h1 and a1 odds is about a half tempo or about 0.2 pawns.

Hi.

Ok, I understand.
But with such a large material advantage, I think that the player getting the handicap should not worry too much about the opponent castling or not castling.
What I have noticed in these games is that, although the rook is a very powerful piece, from the initial position it takes too long to become active, too many moves for its presence to be felt.
So, the priority is to open one file or two and activate the extra rook. Before that, the (weak) human player can lose due to any tactical oversight. After that, it sort of becomes easier... there are still tactical pitfalls, but somehow I feel that they are less disastrous.

Greetings.

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

We've reached the point where elo differences between top engines are only meaningful when you specify the opening book used. With "normal" openings (i.e. ones seen often in human GM play) Komodo and Stockfish will draw most of the games, so even a 3 or 4 to 1 win ratio won't look like a lot of elo points. But if you force Black to play bad defenses, as both CCC and TCEC do (to avoid excessive draws), then the games start near the boundary between a win and a draw. The stronger engine will generally win as White and draw as Black, since only one suboptimal move per game is likely enough to lose with Black or to allow a draw as White. So even a slightly stronger engine may show up as nearly 192 elo stronger (the elo gap for winning half and drawing half) with such unbalanced openings. In the present case, we have 11 wins for SF to 3 for Dragon, with 15 draws, almost exactly half draws, and it seems Dragon can sometimes hold the bad positions and sometimes win the good ones, but not often. As long as we measure Elo by unbalanced openings (such as S. Pohl's Unbalanced Human Openings 22, which I think is pretty ideal), small improvements will look like huge elo gains, and we may see ratings rise to 5000 or more. But if measured with Classical openings, even reaching 4000 (at Rapid, not blitz) won't be easy. That's one argument for using chess960; there is no need for an opening book, the elo difference is clearly defined.

[Chessqueen]

Yes, h1 is a larger handicap than a1, though not a not larger, for the obvious reason that castling kingside is generally preferred over castling queenside because the king on c1 needs to move again to b1 to be secure (generally speaking). All historical rook odds games are with a1 removed, never h1. In normal chess, if White can only castle queenside while Black can only castle kingside, current Stockfish rates the position as 50% chance for each side; White has lost his opening advantage. So roughly speaking the difference between h1 and a1 odds is about a half tempo or about 0.2 pawns.

Hi.

Ok, I understand.
But with such a large material advantage, I think that the player getting the handicap should not worry too much about the opponent castling or not castling.
What I have noticed in these games is that, although the rook is a very powerful piece, from the initial position it takes too long to become active, too many moves for its presence to be felt.
So, the priority is to open one file or two and activate the extra rook. Before that, the (weak) human player can lose due to any tactical oversight. After that, it sort of becomes easier... there are still tactical pitfalls, but somehow I feel that they are less disastrous.

Greetings.

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

We've reached the point where elo differences between top engines are only meaningful when you specify the opening book used. With "normal" openings (i.e. ones seen often in human GM play) Komodo and Stockfish will draw most of the games, so even a 3 or 4 to 1 win ratio won't look like a lot of elo points. But if you force Black to play bad defenses, as both CCC and TCEC do (to avoid excessive draws), then the games start near the boundary between a win and a draw. The stronger engine will generally win as White and draw as Black, since only one suboptimal move per game is likely enough to lose with Black or to allow a draw as White. So even a slightly stronger engine may show up as nearly 192 elo stronger (the elo gap for winning half and drawing half) with such unbalanced openings. In the present case, we have 11 wins for SF to 3 for Dragon, with 15 draws, almost exactly half draws, and it seems Dragon can sometimes hold the bad positions and sometimes win the good ones, but not often. As long as we measure Elo by unbalanced openings (such as S. Pohl's Unbalanced Human Openings 22, which I think is pretty ideal), small improvements will look like huge elo gains, and we may see ratings rise to 5000 or more. But if measured with Classical openings, even reaching 4000 (at Rapid, not blitz) won't be easy. That's one argument for using chess960; there is no need for an opening book, the elo difference is clearly defined.

So why CCC and TCEC do NOT let the programmers to use their preferred Balanced Opening Books which is included or should be included when you purchase Komodo Dragon, To me an engine is an entire package which should always be used with its own Opening Book, and if I were you, I would NOT send Komodo Dragon to CCC or TCEC unless they accept to use your opening Book.

[lkaufman]

Yes, h1 is a larger handicap than a1, though not a not larger, for the obvious reason that castling kingside is generally preferred over castling queenside because the king on c1 needs to move again to b1 to be secure (generally speaking). All historical rook odds games are with a1 removed, never h1. In normal chess, if White can only castle queenside while Black can only castle kingside, current Stockfish rates the position as 50% chance for each side; White has lost his opening advantage. So roughly speaking the difference between h1 and a1 odds is about a half tempo or about 0.2 pawns.

Hi.

Ok, I understand.
But with such a large material advantage, I think that the player getting the handicap should not worry too much about the opponent castling or not castling.
What I have noticed in these games is that, although the rook is a very powerful piece, from the initial position it takes too long to become active, too many moves for its presence to be felt.
So, the priority is to open one file or two and activate the extra rook. Before that, the (weak) human player can lose due to any tactical oversight. After that, it sort of becomes easier... there are still tactical pitfalls, but somehow I feel that they are less disastrous.

Greetings.

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

We've reached the point where elo differences between top engines are only meaningful when you specify the opening book used. With "normal" openings (i.e. ones seen often in human GM play) Komodo and Stockfish will draw most of the games, so even a 3 or 4 to 1 win ratio won't look like a lot of elo points. But if you force Black to play bad defenses, as both CCC and TCEC do (to avoid excessive draws), then the games start near the boundary between a win and a draw. The stronger engine will generally win as White and draw as Black, since only one suboptimal move per game is likely enough to lose with Black or to allow a draw as White. So even a slightly stronger engine may show up as nearly 192 elo stronger (the elo gap for winning half and drawing half) with such unbalanced openings. In the present case, we have 11 wins for SF to 3 for Dragon, with 15 draws, almost exactly half draws, and it seems Dragon can sometimes hold the bad positions and sometimes win the good ones, but not often. As long as we measure Elo by unbalanced openings (such as S. Pohl's Unbalanced Human Openings 22, which I think is pretty ideal), small improvements will look like huge elo gains, and we may see ratings rise to 5000 or more. But if measured with Classical openings, even reaching 4000 (at Rapid, not blitz) won't be easy. That's one argument for using chess960; there is no need for an opening book, the elo difference is clearly defined.

So why CCC and TCEC do NOT let the programmers to use their preferred Balanced Opening Books which is included or should be included when you purchase Komodo Dragon, To me an engine is an entire package which should always be used with its own Opening Book, and if I were you, I would NOT send Komodo Dragon to CCC or TCEC unless they accept to use your opening Book.

If they used good (balanced) opening books, nearly every game between the top 3 engines would end in a draw, and hardly anyone would watch. Chess with free choice of opening is virtually analyzed to a draw with engines playing at the level of these 3 at Rapid time controls or longer. It makes little difference whether they use a specific opening book, it's simply that if the book only contains lines that end with just a small White plus, the engines can hold the draw nearly every time. Maybe this is the correct way to test engines for rating, that is a matter of opinion, but for tournaments designed to attract an audience it just doesn't work anymore. For computer chess to remain interesting, the draw percentage can't be too large. In my view, 50% is ideal, up to 60% is acceptable, but beyond that people start to lose interest, whether it's human chess or engine chess. At 98% draws (roughly what I'd expect with good books, 8 threads, and Rapid tc between the top 3 engines), it's just a joke. It's time for the chess world to face up to this problem. So far the solution for engines is to force Black to play bad openings, and for humans it is to emphasize fast chess. Neither of these is ideal, but they are better than ignoring the problem. I hope that both the human and engine chess worlds find better solutions.

[pepechuy]

Hi.

Now, the second game against Crafty, removing the a1 rook.
A win for me, ¡but a very tough one!
I misplayed the opening, and lost a knight. Which means, I was only an exchange up.
Fortunately, Crafty had to exchange queens, there was less risk for me.
I do not understand why Crafty played 15. fe4 (instead of 15. Re4), to my eyes it just loses a pawn. ¿Perhaps that pawn was doomed anyway?
Then I made another mistake with 23... b6, and lost the exchange. Again, I was lucky and got a second pawn.
Against a human opponent around my strength, I would feel confident about winning a rook ending two healthy pawns up.
But against an engine, I was disheartened and not very optimistic. I wanted to offer a draw, but I did not find any button in Banksia to do that (I do not know if I am simply missing something very obvious, or if there is no such a button).
I also thought about "adjudicating" the game as a draw (from a practical point of view, there is almost no way I would lose that with a ten second increment), but refrained from doing so, and played on. I admit I was afraid of spending two hours just shuffling pieces.
But Crafty did not defend well, at least from a practical point of view. With 38. Rd2 (allowing me to swap rooks, reaching a pawn ending two pawns up), it basically gave up.
And even before that, I did not feel it was offering much resistance. I guess that the position is more lost for Crafty than what I felt during the game, and that there was simply no "good" way of defending it.

Oh, I set resign = 327.65, and Crafty did resign, I think three moves before mate. I do not know how to prevent Crafty from resigning.

[pgn]
[Event "banksia game"]
[Site "?"]
[Date "2022.04.10"]
[Round "?"]
[White "Crafty-25.3"]
[Black "hm::Garcia Ruvalcaba, Jose de Jesus"]
[Result "0-1"]
[WhiteElo "2600"]
[Time "19:50:47"]
[TimeControl "900+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "resign"]

1. e4 {-5.53/25 42259 127903574} 1... e5 2. Nf3 {-5.61/24 38521 71662151} 2...
Nf6 3. Nxe5 {-5.59/24 34252 106462846} 3... d6 4. Nf3 {-5.58/24 30735 84594931}
4... Nxe4 5. d4 {-5.50/25 47994 163161395} 5... d5 6. Bd3 {-5.52/23 29723
97274340} 6... Be7 7. O-O {-5.55/24 28325 70631808} 7... Nd7 8. Re1 {-5.14/24
37487 98715086} 8... Ndf6 9. c4 {-5.32/23 40103 121133608} 9... Bg4 10. cxd5
{-4.59/24 30511 102823038} 10... Qxd5 11. Qe2 {-4.39/23 27206 82105145} 11...
O-O-O 12. Bxe4 {-3.42/26 27677 110280973} 12... Nxe4 13. Qxe4 {-3.24/27 24468
75039693} 13... Bxf3 14. gxf3 {-3.39/26 23867 85999392} 14... Qxe4 15. fxe4
{-3.32/26 33963 94895258} 15... Rxd4 16. Kg2 {-3.32/24 22803 6249813} 16... Rhd8
17. Re2 {-3.31/25 22406 72130389} 17... Rd1 18. Rc2 {-3.18/24 22107 65955740}
18... c6 19. Nc3 {-3.17/25 26959 83442444} 19... R1d3 20. Be3 {-3.13/25 21059
21963948} 20... Kb8 21. Kf1 {-3.03/25 46626 161237784} 21... Bf6 22. Ke2
{-2.91/26 19840 58613501} 22... Bxc3 23. bxc3 {-2.89/26 21628 73065227} 23... b6
24. Bd4 {-3.01/26 18977 55335051} 24... R8xd4 25. cxd4 {-2.42/27 18799 48716189}
25... Rxd4 26. Rxc6 {-2.54/26 42920 119087498} 26... Kb7 27. Rc3 {-2.50/25 24424
84186843} 27... Rxe4+ 28. Kd2 {-2.46/26 19963 65189044} 28... Re7 29. Rh3
{-2.41/23 88011 307700520} 29... h6 30. Rb3 {-2.41/23 29346 89079307} 30... Kc8
31. Ra3 {-2.35/23 87312 258620344} 31... Kd8 32. Kd3 {-2.38/23 38180 92714735}
32... Ke8 33. Ra4 {-2.29/21 35483 110904361} 33... Kf8 34. Rh4 {-2.41/23 39138
131072864} 34... Re6 35. Ra4 {-2.28/23 22510 78119472} 35... a5 36. Rc4
{-2.44/23 35589 111629920} 36... g5 37. Rc2 {-2.52/23 25507 80400655} 37... Kg7
38. Rd2 {-2.77/23 21738 56918526} 38... Rd6+ 39. Kc2 {-4.67/25 10567 28999757}
39... Rxd2+ 40. Kxd2 {-8.68/27 13565 55653036} 40... Kf6 41. a4 {-7.52/28 12240
51463154} 41... Ke5 42. Kd3 {-8.25/29 12856 41210596} 42... Kd5 43. Ke3
{-9.68/28 10404 41672903} 43... f5 44. Ke2 {-10.26/27 11144 34643190} 44... Kc4
45. Ke3 {-13.93/25 10579 29564044} 45... h5 46. h3 {-11.53/22 10284 37198035}
46... b5 47. axb5 {-14.80/25 9090 34572274} 47... Kxb5 48. Kd4 {-16.13/25 10567
47358400} 48... a4 49. Ke5 {-14.95/26 10311 39671048} 49... g4 50. hxg4
{-327.31/26 10189 51030099} 50... fxg4 51. Kf5 {-327.37/24 7947 44050346} 51...
a3 52. Kg5 {-327.41/26 6774 28030678} 52... a2 53. Kxh5 {-327.45/27 6968
25590553} 53... a1=Q 54. Kxg4 {-327.49/28 7104 19101959} 54... Qf6 55. f4
{-327.51/30 7129 24689825} 55... Kc4 56. f5 {-327.53/36 7144 20190377} 56... Kd4
57. Kf4 {-327.55/50 7137 20339980} 57... Kd5 58. Kf3 {-327.55/35 7247 25392366}
58... Qxf5+ 59. Kg2 {-327.57/70 7445 24371581} 59... Ke4 60. Kg3 {-327.59/124
1371 4537710} 60... Qf3+ 0-1
[/pgn]

Next game, against Komodo, removing the h1 rook, when I get the chance.

Greetings.

[pepechuy]

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

You are right in most positions.
But something I have learned: ¡Do not exchange pieces just because you are up in material!
There are also bad exchanges when you have more material.

The point I wanted to make: an extra active rook is a very large advantage, even if there are many more pieces on the board.
An inactive extra rook is very difficult for us to exploit against a much stronger opponent (an engine).

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

As GM Kaufmann already explained, that is because they are using openings with a large advantage for White.
Otherwise, there would be almost only draws.

Greetings.

[Chessqueen]

With the h1 Rook odds, which is a huge handicap, if you can manage to trade as many pieces as Dragon allows you to trade, your advantage will grow exponentially there is NOT much that Dragon can do to avoid trading, since at some point it is unavoidable, Now if you do NOT trade when you have a chance it will become harder to beat Dragon, since it is a monster tactically and the only way to make it wreaker is as you trade down, and at the same time Dragon has less pieces on the board to make the game more complicated for the human.

You are right in most positions.
But something I have learned: ¡Do not exchange pieces just because you are up in material!
There are also bad exchanges when you have more material.

The point I wanted to make: an extra active rook is a very large advantage, even if there are many more pieces on the board.
An inactive extra rook is very difficult for us to exploit against a much stronger opponent (an engine).

Note: I wonder how can 55 to 65 rating points makes such a huge difference with engines matches, as you can see Stockfish is only about 60 rating points more than Komodo Dragon, but is beating it18 games to 10 so far https://tcec-chess.com/

Greetings.

Most of the bad exchanges can be easily detected, by any real weak player, but the Majority of the times, which is about 90% it benefit the side that is up in material, and as far as active or inactive Rook you will have to be a very beginner not to make your extra Rook actives securing either open files and putting more pressure on the opponent.

[pepechuy]

Hi.

The second game against Komodo, removing the h1 rook.
A draw.
I see that Komodo deviated from the previous game with 5. e3 (instead of 5. Nc3).
After I made it, I regretted playing 16... Na4 (going for knight exchange), and thought that 16... c4 (rounding up the b5 pawn) was better. I guess both moves are good.
I like my move 22... h6, offering a pawn in order to exchange rooks. The bad thing, I exchanged my dark squared bishop, weakening my king.
I missed both 28. e5 and 29. Nf5. Fortunately for me, Komodo only gets a perpetual after that, not a mating attack.
I see that Komodo gives me advantage in the final position. I saw that I could keep an extra piece with 31... Ke8, but did not like 32. Qh8 Ke7 33. Bf5, and preferred to take the draw.

[pgn]
[Event "banksia game"]
[Site "?"]
[Date "2022.04.12"]
[Round "?"]
[White "Komodo 12.1.1 64-bit"]
[Black "hm::Garcia Ruvalcaba, Jose de Jesus"]
[Result "1/2-1/2"]
[WhiteElo "3500"]
[Time "20:15:35"]
[TimeControl "900+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBN1 w Qkq - 0 1"]
[Termination "repetition"]

1. d4 {-3.64/24 56468 47970336} 1... Nf6 2. c4 {-3.62/23 32194 27442797} 2... c5
3. d5 {-3.28/25 23220 19097493} 3... b5 4. cxb5 {-3.59/23 37596 32148141} 4...
a6 5. e3 {-3.53/24 77111 67425985} 5... g6 6. Nc3 {-3.38/23 34386 27071307} 6...
d6 7. e4 {-3.47/25 56283 51615137} 7... Bg7 8. a4 {-3.51/24 36993 39230561} 8...
O-O 9. Rb1 {-3.54/24 36962 52471261} 9... axb5 10. axb5 {-3.55/23 26253
24337903} 10... Nbd7 11. Be2 {-3.64/24 27540 31550970} 11... Nb6 12. Kf1
{-3.69/24 96106 94701596} 12... Qc7 13. Bg5 {-3.72/24 84838 73274745} 13... Bd7
14. Nf3 {-3.37/22 29014 26873226} 14... Ra5 15. Nd2 {-3.34/23 58449 54749455}
15... Rfa8 16. Kg1 {-3.47/25 50031 68470352} 16... Na4 17. h3 {-3.61/24 10
54172728} 17... Nxc3 18. bxc3 {-3.34/21 11473 10887085} 18... Ra2 19. b6
{-3.67/27 10 99653889} 19... Qb7 20. Bd3 {-3.84/26 10 38421611} 20... Ba4 21.
Qc1 {-3.28/29 42540 42827941} 21... Bd7 22. Nc4 {-2.69/23 28119 26998888} 22...
h6 23. Bxh6 {-2.48/25 41157 42113207} 23... Bxh6 24. Qxh6 {-2.82/23 6198
10308105} 24... Ra1 25. Kh2 {-2.88/26 62696 69968694} 25... Rxb1 26. Bxb1
{-3.13/25 10 13383620} 26... Bb5 27. Ne3 {-2.86/23 22871 25036948} 27... Qxb6
28. e5 {-2.57/24 20554 20286883} 28... dxe5 29. Nf5 {-2.72/30 10 73482665} 29...
gxf5 30. Qg5+ {-2.63/23 2616 20775217} 30... Kf8 31. Qh6+ {-2.64/23 10341
21558884} 31... Kg8 32. Qg5+ {+0.00/48 76869 106853657} 32... Kf8 33. Qh6+
{-2.64/26 10 22982712} 33... Kg8 1/2-1/2
[/pgn]

Next game, against Lc0 removing the a1 rook, when I get the chance.

Greetings.

[pepechuy]

Hi.

Now, the second game against Lc0, removing the a1 rook.
I deviated from the first game, with 3... d6 (instead of 3... Nc6).
Lc0 gave me a hard time, not allowing me to activate my rooks.
Then, the move 19. Ng6 (allowing me to activate one rook) gave me a lot of confidence in my position.
23... Rc8 is a mistake, which loses a bishop (of course I missed it). Fortunately, I was still an exchange up, with a very good position, and soon was two (and then three) pawns up too.
I did not hesitate sacrifice the exchange and enter a knight vs bishop ending, being three pawns up, with 29... Ra3 30. Kd4 Rg3
Before that, I felt that the bishops were harassing my king, my rook and my knight.
On the other hand, usually it is very easy to avoid a single bishop.
I made some mistakes afterwards (for example, I missed 43... Bf3), but by that point my advantage was overwhelming, there was never a risk of not winning this ending.

[pgn]
[Event "banksia game"]
[Site "?"]
[Date "2022.04.14"]
[Round "?"]
[White "Lc0 v0.27.0"]
[Black "hm::Garcia Ruvalcaba, Jose de Jesus"]
[Result "0-1"]
[WhiteElo "3500"]
[Time "20:44:10"]
[TimeControl "900+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "mate"]

1. Nf3 {-6.10/5 108079 6613} 1... Nf6 2. c4 {-7.73/5 40817 3694} 2... c5 3. Nc3
{-9.08/5 39400 3544} 3... d6 4. e3 {-8.27/4 27983 2561} 4... Bg4 5. Be2 {-7.18/4
26961 1873} 5... Nc6 6. d4 {-6.08/5 22729 1896} 6... cxd4 7. exd4 {-6.76/5 10104
1023} 7... d5 8. cxd5 {-5.34/6 13557 1436} 8... Nxd5 9. Qb3 {-4.96/6 9197 1530}
9... Nxc3 10. bxc3 {-5.67/6 250 1629} 10... Qb6 11. Qa4 {-5.99/6 19685 2262}
11... e6 12. O-O {-5.90/6 22741 1647} 12... Be7 13. Be3 {-7.01/7 22871 2356}
13... Qa5 14. Qxa5 {-5.92/7 9972 1782} 14... Nxa5 15. Bb5+ {-12.16/5 10609 2081}
15... Kd8 16. Ne5 {-10.82/6 3896 1974} 16... Bh5 17. g4 {-9.90/5 19305 2689}
17... Bg6 18. d5 {-9.04/5 44060 3748} 18... Bd6 19. Nxg6 {-9.33/6 6244 3181}
19... hxg6 20. dxe6 {-10.91/5 22354 2912} 20... f6 21. Rd1 {-11.85/5 41138 3785}
21... Ke7 22. Kg2 {-18.15/5 34529 3556} 22... Rxh2+ 23. Kf3 {-22.71/5 33767
2929} 23... Rc8 24. Rxd6 {-16.21/5 24271 3096} 24... Kxd6 25. Bf4+ {-24.12/6
11875 1984} 25... Kxe6 26. Bxh2 {-27.07/6 8401 1403} 26... Rxc3+ 27. Ke4
{-27.23/5 30222 2659} 27... Rc2 28. Bg3 {-27.78/5 115784 9755} 28... Rxa2 29.
Kd3 {-19.25/5 22489 2751} 29... Ra3+ 30. Kd4 {-34.92/5 8778 2540} 30... Rxg3 31.
fxg3 {-37.38/5 16781 1317} 31... Nc6+ 32. Kc5 {-29.75/5 28301 2481} 32... g5 33.
Ba4 {-26.38/5 30084 3582} 33... Ke5 34. Bc2 {-18.54/5 24824 2980} 34... Nd4 35.
Bg6 {-31.97/6 21377 3526} 35... Ne2 36. Bd3 {-48.35/5 88714 12371} 36... Nxg3
37. Kc4 {-67.12/5 15491 3287} 37... Ne4 38. Be2 {-49.50/5 23641 2493} 38... g6
39. Bd1 {-45.69/5 41607 4634} 39... f5 40. Be2 {-37.68/4 20021 2142} 40... f4
41. Kd3 {-30.66/5 20240 2050} 41... Nf6 42. Bd1 {-51.16/5 32784 3860} 42... Kd5
43. Bf3+ {-50.14/4 18339 1669} 43... Kc5 44. Bxb7 {-64.34/4 14356 1453} 44...
Nxg4 45. Bc8 {-36.67/4 16208 1763} 45... Ne5+ 46. Ke4 {-30.96/5 12591 1096}
46... Nc4 47. Ba6 {-15.25/5 14106 1664} 47... Nd6+ 48. Kf3 {-32.18/5 14831 1383}
48... Kb6 49. Bd3 {-59.00/5 19619 1764} 49... a5 50. Bc2 {-41.86/4 7778 1551}
50... Kc5 51. Bb3 {-46.90/4 18971 2339} 51... Kb4 52. Ba2 {-53.13/5 17063 1981}
52... Kc3 53. Bd5 {-55.05/4 14364 1939} 53... a4 54. Ba2 {-46.87/5 34231 4026}
54... Kb2 55. Bd5 {-72.87/4 13143 1490} 55... a3 56. Be6 {-37.87/4 10221 1374}
56... a2 57. Bxa2 {-17.34/5 6258 685} 57... Kxa2 58. Kg4 {-40.26/5 8429 1697}
58... Nf7 59. Kf3 {-47.01/5 10850 1539} 59... Kb3 60. Ke4 {-32.06/5 10495 1814}
60... Kc4 61. Kf3 {-18.26/1 10 155} 61... Kd3 62. Kf2 {-30.17/5 10563 1245}
62... g4 63. Ke1 {-39.64/5 11388 1752} 63... g3 64. Kf1 {-29.59/5 8287 989}
64... f3 65. Kg1 {-15.40/6 12208 1640} 65... Ke2 66. Kh1 {-8.90/1 10 155} 66...
Ng5 67. Kg1 {M-5/1 10 97} 67... Nh3+ 68. Kh1 {M-4/1 10 16} 68... Nf4 69. Kg1
{M-2/1 10 5} 69... f2+ 70. Kh1 {M-1/1 10 3} 70... f1=Q# 0-1
[/pgn]

Next game is against Stockfish, removing the h1 rook.
To be played when I get the chance.

Greetings.

[Chessqueen]

Hi.

Now, the second game against Lc0, removing the a1 rook.
I deviated from the first game, with 3... d6 (instead of 3... Nc6).
Lc0 gave me a hard time, not allowing me to activate my rooks.
Then, the move 19. Ng6 (allowing me to activate one rook) gave me a lot of confidence in my position.
23... Rc8 is a mistake, which loses a bishop (of course I missed it). Fortunately, I was still an exchange up, with a very good position, and soon was two (and then three) pawns up too.
I did not hesitate sacrifice the exchange and enter a knight vs bishop ending, being three pawns up, with 29... Ra3 30. Kd4 Rg3
Before that, I felt that the bishops were harassing my king, my rook and my knight.
On the other hand, usually it is very easy to avoid a single bishop.
I made some mistakes afterwards (for example, I missed 43... Bf3), but by that point my advantage was overwhelming, there was never a risk of not winning this ending.

[pgn]
[Event "banksia game"]
[Site "?"]
[Date "2022.04.14"]
[Round "?"]
[White "Lc0 v0.27.0"]
[Black "hm::Garcia Ruvalcaba, Jose de Jesus"]
[Result "0-1"]
[WhiteElo "3500"]
[Time "20:44:10"]
[TimeControl "900+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "mate"]

1. Nf3 {-6.10/5 108079 6613} 1... Nf6 2. c4 {-7.73/5 40817 3694} 2... c5 3. Nc3
{-9.08/5 39400 3544} 3... d6 4. e3 {-8.27/4 27983 2561} 4... Bg4 5. Be2 {-7.18/4
26961 1873} 5... Nc6 6. d4 {-6.08/5 22729 1896} 6... cxd4 7. exd4 {-6.76/5 10104
1023} 7... d5 8. cxd5 {-5.34/6 13557 1436} 8... Nxd5 9. Qb3 {-4.96/6 9197 1530}
9... Nxc3 10. bxc3 {-5.67/6 250 1629} 10... Qb6 11. Qa4 {-5.99/6 19685 2262}
11... e6 12. O-O {-5.90/6 22741 1647} 12... Be7 13. Be3 {-7.01/7 22871 2356}
13... Qa5 14. Qxa5 {-5.92/7 9972 1782} 14... Nxa5 15. Bb5+ {-12.16/5 10609 2081}
15... Kd8 16. Ne5 {-10.82/6 3896 1974} 16... Bh5 17. g4 {-9.90/5 19305 2689}
17... Bg6 18. d5 {-9.04/5 44060 3748} 18... Bd6 19. Nxg6 {-9.33/6 6244 3181}
19... hxg6 20. dxe6 {-10.91/5 22354 2912} 20... f6 21. Rd1 {-11.85/5 41138 3785}
21... Ke7 22. Kg2 {-18.15/5 34529 3556} 22... Rxh2+ 23. Kf3 {-22.71/5 33767
2929} 23... Rc8 24. Rxd6 {-16.21/5 24271 3096} 24... Kxd6 25. Bf4+ {-24.12/6
11875 1984} 25... Kxe6 26. Bxh2 {-27.07/6 8401 1403} 26... Rxc3+ 27. Ke4
{-27.23/5 30222 2659} 27... Rc2 28. Bg3 {-27.78/5 115784 9755} 28... Rxa2 29.
Kd3 {-19.25/5 22489 2751} 29... Ra3+ 30. Kd4 {-34.92/5 8778 2540} 30... Rxg3 31.
fxg3 {-37.38/5 16781 1317} 31... Nc6+ 32. Kc5 {-29.75/5 28301 2481} 32... g5 33.
Ba4 {-26.38/5 30084 3582} 33... Ke5 34. Bc2 {-18.54/5 24824 2980} 34... Nd4 35.
Bg6 {-31.97/6 21377 3526} 35... Ne2 36. Bd3 {-48.35/5 88714 12371} 36... Nxg3
37. Kc4 {-67.12/5 15491 3287} 37... Ne4 38. Be2 {-49.50/5 23641 2493} 38... g6
39. Bd1 {-45.69/5 41607 4634} 39... f5 40. Be2 {-37.68/4 20021 2142} 40... f4
41. Kd3 {-30.66/5 20240 2050} 41... Nf6 42. Bd1 {-51.16/5 32784 3860} 42... Kd5
43. Bf3+ {-50.14/4 18339 1669} 43... Kc5 44. Bxb7 {-64.34/4 14356 1453} 44...
Nxg4 45. Bc8 {-36.67/4 16208 1763} 45... Ne5+ 46. Ke4 {-30.96/5 12591 1096}
46... Nc4 47. Ba6 {-15.25/5 14106 1664} 47... Nd6+ 48. Kf3 {-32.18/5 14831 1383}
48... Kb6 49. Bd3 {-59.00/5 19619 1764} 49... a5 50. Bc2 {-41.86/4 7778 1551}
50... Kc5 51. Bb3 {-46.90/4 18971 2339} 51... Kb4 52. Ba2 {-53.13/5 17063 1981}
52... Kc3 53. Bd5 {-55.05/4 14364 1939} 53... a4 54. Ba2 {-46.87/5 34231 4026}
54... Kb2 55. Bd5 {-72.87/4 13143 1490} 55... a3 56. Be6 {-37.87/4 10221 1374}
56... a2 57. Bxa2 {-17.34/5 6258 685} 57... Kxa2 58. Kg4 {-40.26/5 8429 1697}
58... Nf7 59. Kf3 {-47.01/5 10850 1539} 59... Kb3 60. Ke4 {-32.06/5 10495 1814}
60... Kc4 61. Kf3 {-18.26/1 10 155} 61... Kd3 62. Kf2 {-30.17/5 10563 1245}
62... g4 63. Ke1 {-39.64/5 11388 1752} 63... g3 64. Kf1 {-29.59/5 8287 989}
64... f3 65. Kg1 {-15.40/6 12208 1640} 65... Ke2 66. Kh1 {-8.90/1 10 155} 66...
Ng5 67. Kg1 {M-5/1 10 97} 67... Nh3+ 68. Kh1 {M-4/1 10 16} 68... Nf4 69. Kg1
{M-2/1 10 5} 69... f2+ 70. Kh1 {M-1/1 10 3} 70... f1=Q# 0-1
[/pgn]

Next game is against Stockfish, removing the h1 rook.
To be played when I get the chance.

Greetings.

Well done, even early on you could have traded your Bishop for Lco Knight on f3 after 5. Be2...5.BxNf3, and place your Knight on c6 after the exchange, with a Rook Odds the earlier you trade the less your opponent Engine has a chance of using his tactical superiority. Also on move 22....Rxh2? you should have taken the pawn with your Bishop 22.......Bxh2! to avoid the engine from taking your Bishop and and pinning your Rook