TCEC S17 SUper FInal report

by @glbchess64

thanks to @dtracers and @ildolphino for correction and rewriting.

Leela won the SuFi, 52.5 to 47.5 (+5 wins). This was a difficult match that showed the qualities and the weaknesses of the two engines. Leela, with network SV-t60-3010, is by most people considered as a positional engine who likes to grind down the opponent instead of going for the quickest mate, but of course she is also a tremendous tactical player. Still she has tactical blunders and sometimes has difficulties with correctly estimating the opponent’s queen play, especially in open positions. Sometimes she does not succeed in converting a winning position because of falling prey to a perpetual check (note: the developers are working hard on this issue and a few enhancements are in the pipeline). StockFish, with 176 threads, is a fast and impressive tactician with a very solid play. But, even non-grandmaster players permit themselves to call its positional play dubious at times. It looks like a very clever player, it calculates good combinations but sometimes does not know how to coordinate pieces with pawns. In other words StockFish attempts to solve chess by brute force calculation, very clever search heuristics and a fast evaluation function. Sometimes it even helps Leela in some positions sacking pawns in a desperate way while Leela had a hard time finding the win!

Odd number games SF (Stockfish 20200407DC, 176 threads) with white, even number Leela (LCZero v0.24-sv-t60-3010, 4xRTX 2080Ti) with white. TC 90’+5" (note: often short for Leela in the endgame).

Leela fans love to understand what happens on the board, love simple moves that chain with logic, love to see plans.

StockFish fans love brilliant combinations, crushing games, even if this kind of play tends to be hard to understand.

It is difficult to be both a Leela fan and a StockFish fan, even though exceptions exist. TCEC chat was passionate.

Jeroen’s work on the openings resulted in fascinating games. Very open games lead to quick draws with a lot of pieces and pawns swapped (for this reason there was not any King Gambit Accepted in the book), closed games or semi-open games were more interesting and often decisive.

Even if Leela did not crush StockFish, she clearly dominated the match. With black she mostly equalized quickly. With white she tortured StockFish during very long positional struggles until StockFish accumulated decisive mistakes. Leela’s play on a color or with pawns was simply impressive.

This match showed that you cannot play perfect chess by calculating combinations only, understanding the game is also necessary.

“Les pions sont l’âme des échecs,” Philidor (pawns are the soul of chess)

Game 1: QGD, exchange variation.

Book 1.d4

Leela equalizes before move 10 and takes an advantage with better pawn structure. She takes a pawn in the center but returns her h pawn to deny 0-0 to SF. She gets a better pawn structure and a majority on the q-side where she castles with SF king in the center. Black has a good edge. But as the queens remain on the board, SF manages, with dynamic play, to equalize and gets a draw by perpetual check.

Game 2: QGD, semi-slav.

Leela loses a small beginning advantage and the game is a short draw.

Game 3: Sicilian, Keres attack (SF +1)

Book 1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. g4 a6

The last move of book 6…a6 instead 6…h6 made this line may be winning for white. SF mobilizes the h-pawn and gets a fawn in h6. SF used it, after exploding the center, to get a winning pawn endgame. This game looks like a Leela game but the next one is more positional.

Game 4: (Leela equalizes)

Leela gets a strong advantage even more quickly than SF. She pushes the f pawn and with Rg3 in offensive and defensive position she gets a semi closed center (black d6-e5 and white e4) and castles long while SF king remains in the center. SF decides for the thematic Rxc3 exchange sac. Swapping the queen and getting the e4 pawn as a minor compensation. The ending that results with white majority on the q-side was absolutely hopeless. Finally a very simple game : pure Leela style.

Game 5: KID, orthodox

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Nd3 f5 11. Bd2 Nf6 12. f3 f4

SF decides for a quick a-pawn push to a6. Leela finds usual counter play on the k-side and the center. Lots of pawns and pieces swapping, to end queens and a knight pair, SF a pawn up. But with so few pawns on the board and kings in open space it is a logical draw.

Game 6:

3 pawns swap, then on move 25 Leela dominates the q-side and the center. SF holds the k-side and gives a 0 eval. Pieces begin to swap and Leela gets a strong passer in the center but Fish has counter play against Leela King. Draw.

Game 7: Vienna Game: Frankenstein-Dracula Variation (SF +1)

Book 1. e4 e5 2. Nc3 Nf6 3. Bc4 Nxe4 4. Qh5 Nd6 5. Bb3 Nc6 6. Nb5 g6 7. Qf3 f5

Likely a 1-0 opening. Black gives the Ra8 for development and space advantage. But taking back the Na8 loses two tempos, black pieces are not well coordinated and black king in d8. SF plays very clever moves like Qf3-h3-f3-d1-g4 at the right tempo, gets well coordinated pieces and wins with strong threats on the black king in the center. A masterpiece for SF.

Game 8: (SF +1)

Leela plays nearly the same setting as SF in the previous game but not at the right tempo, and finally all advantages evaporate. A draw.

Game 9: Dutch, Leningrad variation (SF +1)

Book 1. d4 f5 2. g3 Nf6 3. Bg2 g6 4. Nf3 Bg7 5. O-O O-O 6. c4 d6 7. Nc3 e6

This opening gives a good edge to white. Leela chooses a stonewall structure and plays an attack on the k-side. SF reacts on the q-side but strangely closes the game on that side to contest k-side. Interestingly both engines thought they were better. Finally after many swaps the game ends in a draw.

Game 10: (SF +1)

Long positional torture by Leela. SF chooses the stonewall. Leela attacks on the q-side and SF decides to defend, Leela does not manage to explode the q-side so she goes for the center and SF reacts at k-side and manages to get enough counter play. A fighting draw.

Game 11: French, Winawer variation (SF +1)

Book 1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 b6

The worst opening for SF. Playing the anti positional c4 move SF proves one more time that it does not understand this opening. Leela equalized with a clever Kd7 move after sacrificing g7 pawn for counter play on k-side. Draw.

Game 12: (Leela equalizes)

The game all Leela fans waited for. Classical play by Leela on k-side with g4-h5, Rh3 setup. SF reacts by closing the q-side with c4? and then a4?? At this point he thinks white has a slight edge whereas Leela and all club players know that the position is hopeless for black. The remainder of the game is just technical for Leela. A game to analyse by the SF dev team.

Game 13: Benoni defense

Book 1. d4 Nf6 2. c4 c5 3. d5 e6 4. Nc3 exd5 5. cxd5 d6 6. Nf3 g6 7. Nd2

Another opening that SF does not understand but Leela does. SF plays on the q-side and keeps the center closed. Since in Benoni black must play on q-side and white in the center this results in a fighting draw where SF gets a pawn up but nothing to convert.

Game 14: (Leela +1)

New nightmare for SF. It swaps, early in the game, the Bg7 against the Nc3 and allows white to block q-side with Nb6. The only compensation it gets is the e-pawn. All Benoni players know that it is not enough. Weaknesses on DS and closed q-side is too much. At this point, on move 19, the game is positionaly lost and SF eval is just +0.7 Leela fights to take control of the e-file, SF gives the pawn back on the a-file to delay the logical result : Leela wins with a devastating rook on e6.

Game 15: Modern, Robatsch defense (Leela +1)

Book 1. e4 g6 2. d4 d6 3. Nc3 a6

SF chooses the 3 pawns attack, Leela reacts immediately with a6-b5 and, for a pawn, gets the control of LS on the q-side and the center. On move 20, she has enough counter play to get an easy draw in a simple endgame. Good positional play for Leela

Game 16: (SF equalizes)

Drama Leela blunders and loses the game. She chooses a Be3, h4, 0-0-0 setup and begins a long positional torture. But SF manages to resist and Leela blunders.

Game 17: Budapest gambit

Book 1. d4 Nf6 2. c4 e5 3. dxe5 Ng4

SF chose a solid a2-b3-c4, e3-f4 and Bd2 setup. Leela reacts with g6 and Bg7 and begins to attack the q-side. After all pawns disappear on the q-side and center and some piece swap, SF wins a pawn but with the pawns only on k-side this is not enough to win. Good positional evaluation of the game for Leela.

Game 18:

A masterpiece of positional play. May be one of the best games of the match. A very complex game to watch absolutely and to analyse with an engine. Leela decides for the same pawn setting that SF chose in the previous game. But she puts the pieces on other squares compared to SF and begins a long positional domination of all parts of the board. She manages to get SF queen for two rooks and a pawn, and something that seems to be a drawn endgame. She continues the pressure and gets a likely won game threatening to infilter her king on weak LS in the k-side. SF sees the win and she misses it in zeitnot. 5 second increment is not enough for her to play such a difficult attack. Draw.

Game 19: Sicilian

Book 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Qb6 5. Nb3 Nf6 6. Nc3 e6 7. Qe2

Leela 7…Be7 seems to be a new move. She manages to equalize pushing the d7 pawn to d5. After a fight around isolated d5, SF gets nothing. Draw.

Game 20:

Long positional torture. Leela gets the d-file and space on the k-side. SF gets enough counter play on semi-open b-file after Leela castles long. Draw.

Game 21: KID, Saemisch variation

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. f3 O-O 6. Be3 Nc6 7. Nge2 a6 8. Qd2 Rb8 9. Rc1 Bd7

SF attacks the k-side but Leela reacts opening lines and sacking pawns in the q-side and center. She gets a piece for 4 pawns and a draw in the endgame.

Game 22:

A masterpiece of positional play. Leela gets DS control on the q-side, a dangerous passed pawn on d6 and an attack on the k-side. Dominating all the board with the king in the center. SF finally manages to sac the exchange and win the d6 pawn and gets enough counter play against the exposed white king. A fighting draw.

Game 23: Ruy Lopez, Schliemann variation

Book 1. e4 e5 2. Nf3 Nc6 3. Bb5 f5 4. Nc3 fxe4 5. Nxe4 d5

An opening typical of correspondence play and very risky, quasi-winning variation for white Leela manages to hold against SF in a brilliant way : sacrificing 2 pawns for attack and activity, playing good attacking and defensive moves, she manages to outclass SF on its own ground. Really impressive. The key of the defense was the control of the e-file, LS on queen-side and a dominating knight in the center. After that, SF eval goes to 0. Draw.

Game 24:

SF chooses one of the main lines that allows it to keep a long time a very annoying e4 pawn against two pawns. This pawn is a pain because it prevents normal development for Bc1 and Ra1. Leela plays one of the usual lines known from correspondence games and the game ends in a draw.

Game 25: Slav defense, Alapin variation.

Book 1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 dxc4 5. a4 Bg4 6. Ne5 Bh5 7. f3 Nfd7 8. Nxc4 e5

A very biased and complex opening that looks like winning for white. White has a lot of space and Leela as black lacks room for her pieces. Playing e5 at the end of book she decides to sac a pawn to gain a tempo to develop the DSB perhaps her best piece and acting on the center with a temporary pin of the Nc3. At this stage black position looks horrible. Nevertheless white chooses a line that causes him some issue to develop and black manages to take control on the d-file.

At that point SF decides to return the pawn to fluidify its development and swaps a knight and a rook. After this it seems that black takes an option on DS and that swapping better favours black than white.

In fact, Leela and SF often have the same PV as if SF is playing against itself. This results in an eval that decreases slowly, numbers of forced moves for Leela and a position virtually drawn. The decision to return the pawn seems, now, very doubtful since it results in a 3 to 1 pawn advantage for Leela on q-side. White has no play on that wing and this limits win possibilities.

Game 26: (Leela +1)

Contrary to SF, Leela chooses to swap the good black DSB and the queen and enters an endgame a pawn up and with a better pawn structure on q-side. With space and development advantage. Then she takes her time in a long positional torture typical of her play and finally wins. After 1/4 of games played there is no doubt that Leela dominates this match and that SF only resists because of the incredible calculus power given by the 176 threads.

Game 27: QID (SF equalize)

Book 1. e4 b6 2. d4 Bb7 3. Bd3 e6 4. Nf3

The game early transposed to a French defense, advance variation. Good defensive play by Leela, very clever and not so bad play by SF until Leela blunder in zeitnot.

Game 28: (Leela +1)

The game transposes also in French advance. SF plays the anti positional c4? and Leela a clever h4-h5, Rh3-g3, Kf1-g1. Black gets the q-side and white the k-side. But white has more space for the attack than black (c4 limits the attacking possibilities). Finally Leela manages to immobilise Bf8 and Rh8 with simple threats on g7 with the Rg3, plays f5, gets the center, swaps good black pieces and wins the endgame despite time pressure when black has only its bad pieces, DSB and Rh8.

Game 29: QID (SF equalize)

Book 1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. a3 c5

A position known from A0 games is on the board! Leela plays a better defense than SF8. Positional fight. But on move 30 SF is a pawn up without compensations for black. 10 moves later SF gets the a-pawn and a passer. Maybe a Leela blunder but the position looks difficult. Now it seems desperate for Leela that tries a dynamic defense against the white king in the center and pushes a fawn in h3. But SF eval more than 3 lets few hope. All piece swapping is good for white and the a-passer is also dangerous. Leela can not answer all the threats and loses.

Game 30:

Same A0 position on the board. Like Leela in the previous game SF declines the d pawn gambit and Leela contrary to the Fish attack with h4-h5-hxg6. In pure A0 style, she then transfers the queen to the k-side with Qa4-h4 but SF manages to swap queens and enters a positional endgame. Leela dominates the q-side and center with a protected passer and SF has an edge on the k-side. The position remains very closed. After 31.b4 Leela enters a RB+P vs RB+P endgame that SF seems to think is winning but she does not succeed to convert. Great positional game but draw.

Game 31:

Book 1. e4

SF plays the Ruy Lopez and Leela chooses the Berlin defense. Leela, thanks to the bishop pair and some weaknesses in white pawn structure, gets a small edge and after sacking the exchange for two pawns manages to get a good endgame. But the game ended as a draw. Good positional play.

Game 32:

A Sicilian Najdorf with e5 where SF sacs the exchange for the e4 pawn and gets a strong center. After a complex middle game Leela gets an endgame where the exchange-up was more sensible but does not succeed to win. Draw.

Game 33: KID, classical (SF +1)

Book 1. d4 Nf6 2. c4 g6 3. Nf3 Bg7 4. g3 d6 5. Bg2 O-O 6. Nc3 Nbd7 7. O-O e5 8. e4 c6 9. h3 Qb6 10. c5 dxc5 11. dxe5 Ne8

Sub variation where pawns repartition is very unbalanced after exchange. White has a strong majority on k-side and black on q-side. This majority gives the game to white on move 29. Leela sees only on move 35 that there may be a problem (SF eval >7) and on move 38 that the game is lost. This shows the power of the 176 threads and also why she has so much difficulty to lead despite a better positional play in almost all games.

Game 34: (Leela equalize)

Leela played 12.e6 at the end of the book and closed the center. Later on, SF plays the strange new move 17…Rd7 that seems ugly and forces Leela to put her queen on a better square. Later on, bis repetita, 25…Ba6 forced an unexpected rook move 28.Rf2 and the SF eval began to rise (if SF, I would have played the Leela move). SF got a position with a horrible bad bishop, an attack against its king and rooks and queen blocked by pawns. Only a strong knight in the center. Finally SF sacs the exchange to get rid of a strong Nd6 and loses a technical endgame.

Game 35: Sicilian, Dragon, Yugoslav Attack

Book 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2 Nc6 9. Bc4 Qa5 10. O-O-O Bd7

Generally this is a 1-0 opening for computers. SF plays it in a very personal fashion. Leela went for the thematic exchange sac in c3 but SF declines it and swaps the queens. But Leela had to sac a piece for a dangerous h7 pawn. And SF manages to take the Ra8 for 3 pawns. It seems it will win even if Leela had 2 connected and protected passers on k-side. SF wins a pawn and sac a piece for 2 pawns. Now it is an exchange up. But it had to return the exchange for the two passers and loses a central pawn and finally gets a BPPP vs NPP endgame. All the q-side pawns remain and are in the starting position. And Leela holds… What a game!

Game 36: (Leela +1)

There is little doubt that Leela will win this game. She plays the thematic Nd5 move to inflict structural damages to SF swapping the Nf6 and the Qa5 for double f-pawns and an isolated d-pawn. Advancing pawns SF allows Leela to trade a rook pair and LSB for the knight. Some pawn exchanges later she gets a winning ending: opposite color bishops and rooks but 3 connected passers. She trolls at the last move with a Rg4 sac to get a 6 men TB win!

Game 37: (Leela +1)

Book 1. c4 Nc6

Starting from an English opening the game evolves to a QGD, exchange variation structure with a weak c6 square since Leela played b6. SF decides for a3-b4 pushes and got a very bad DSB. So Leela chooses to “waste” 3 tempos to swap LSB and fight for LS control. After a pair of rook swap along a-file, began a long sequence of piece moves for better positions as if SF is not aware of its bad DSB. Finally Leela equalizes, gets a pawn and two connected passers and SF is forced to take a perpetual. Draw.

Game 38: (Leela +2)

May be the worst SF game in TCEC history. SF chooses a system that gives a good advantage to white: the bishop pair, a good center and DS control. As a compensation black has a short development advantage and an advanced e-pawn.

White castled long and the fight is promising but with 17…c5?? Fish closes the q-side after the obvious 18.a4 that Leela plays. Its eval rises to 2 but too late. Now Leela should win. New mistake 20…Nd3?? and Leela gets a central pawn and swaps the knight. Now all endgames are losing since the a-pawn will fall in the endgame. After Be1, Ra3 Leela attacks this pawn and Fish makes a third positional mistake swapping the queens. The game is totally lost for Fish but its eval is just +3. Leela eval is close to 3.5 and everybody knows what that means.

Game 39: French Defense, Classical, Steinitz variation (Leela +2)

Book 1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 a6 8. Qd2 b5 9. Nd1 Bb7

Like all good french players, Leela attacked the center and the q-side without castling. The center stabilizes with d5-e6-f5 for black and d4-e5-f4 for white. Both opponents have a bad bishop since SF chooses to keep its bad DSB when swapping around d4. Fish attacks the k-side with Rf3-h3 and Leela castles short under fire and answers g4 with b3. At this point SF understands it has nothing. The q-side explodes after Leela adds the a3 push. Then Fish makes a positional blunder, closing the k-side with g5. Now Leela can use all her pieces on the q-side. She plays a little combination to swap white LSB for a knight. She is a pawn down but with better pieces, space and perspectives. The position is equal. Draw.

Game 40: (Leela +2)

Fish, like Leela in the previous game, attacks the center and the q-side without castling. But Leela does not take the same swapping decisions in the battle around c3 and d4. She prefers keeping a pawn on c3 and uses the d4 square for a piece. A typical french treatment of the position. But she let Fish swap a knight for her LSB, which is a strange decision. Then sacs a pawn to open the center playing 17.f5 a thematic push in front of the SF king. The position looks tactically complicated and dangerous for black but Fish thinks all is OK. It castles short and Leela sacs the a new pawn in f6 while c3 is also hanging. The two sacs are declined and a fawn appears when Fish plays g6. Now Leela has an option on DS on the k-side and center. The position simplifies. Fish is a pawn up with a bad LSB and Leela has a knight and dominates DS thanks to her fawn.

Finally Leela got an endgame B+pawns vs N+pawns where she thinks to have a great advantage despite a pawn less. Fish eval is 0 since a long time. The black king is in a box, pawns are blocked. The white king can go everywhere on the board but without real threats. Draw.

Game 41: Old Benoni (Leela +2)

Book 1. d4 c5 2. d5 e5 3. e4 d6

This blocked pawn structure in the center is perfect for Leela. SF developed slowly controlling the q-side and attacking the k-side with a strong advantage, all pawns on white square. Leela swaps her bad DSB for the good white DSB and counter attacks on q-side with a pawn sac that SF takes time to accept. After many swapping the game ends in a drawn rook ending. SF pawn up in the center is not really usable.

Game 42: (Leela +2)

After seeing the previous game there is little doubt that Leela will win this game.

The game starts the same way, Leela searching to limit black play on the q-side but contrary to Fish she manages to push a5 and to play Nb6 for a total blockade. Then she turns her attention to k-side where she manages to keep control on DS with h4-g3 and keeping the bishops on the board. The position looks strong for white (Leela eval +2).

Leela keeps her king in the center, while SF manages to swap its good LSB for Leela bad DSB, she uses the DSB as a threat but refuses to swap and waits for better rooks position before the opening of k-side. Leela slowly takes control of all the board and manages to slightly modify the pawn structure in her favour. Semi-open h-file with h6 as a target, semi-open e-file with the threat to push e5 swapping d6 and creating a passer and threat to open q-side at any time pushing b-pawn. Rooks and queens still on the board, black bad DSB and white Nc4 controlling b6 (Leela eval >3).

Leela finally opens the b-file without getting any advantage and the game ends in a draw.

Game 43: Scandinavian defense (Leela +1)

Book 1. e4 d5 2. exd5 Qxd5 3. Nc3 Qd6 4. d4 Nf6 5. Nf3 a6 6. g3

An inferior variation of the Scandinavian defence and quasi-winning for white.

Leela develops with Bg4 and Nc6 and SF reacts with thematic h3-g4, Ne5, Bg2 and then h4 setup. As a result it trades the Bg2 for the Nc6 and the Ne5 for the Bg6 inflict heavy structural damages: c and g doubled pawns, isolated e-pawn. Then it castles long and gets a winning attack against Leela king in the center.

Leela desperately searches for counter play with heavy pieces on q-side. SF sacs the exchange to end Leela attack and get a N+pawns vs R+pawns endgame that it wins thanks to 2 advanced passers and the threats of forks by the knight.

Game 44: (Leela +2)

This opening seems really too biased in white favour. Leela treats the position in her own way: a positional approach. She early gets the bishop pair against two knights, castles long and pushes a fawn in h6. SF reacts attacking the q-side but around move 35 the game is nearly winning for white: not enough threats on the q-side whereas white controls all the board with her aggressive bishop and the fawn, and also threatens the f5 devastating push.

Finally she pushes g5, consolidating the fawn and the grip on the k-side. She gets the a-pawn so that all endgames are lost. Trading pieces is the most simple winning plan.

She wins the endgame trading pieces and getting more pawns.

Game 45: Benko Gambit (Leela +2)

Book 1. d4 Nf6 2. c4 c5 3. d5 b5 4. Qc2 Na6

A strange variation with the weird Na6. I suppose Qc2 idea is to decline the gambit so SF continues with 5.a3 and with 5…b4 Leela justify the knight position.

Before move 20, Leela attacks the k-side with a bishop sac that SF declines, all her pieces are on efficient squares except maybe the Na6. White is underdeveloped on the q-side. Leela equalizes.

Now she focuses on q-side and the center with the e4 push and Fish releases the tension, swapping pawns in b4. With Nxb4, Leela gives a good square to the knight. SF swaps queens and sacs the exchange for a pawn and after a long tactical sequence Leela finally gets a knight for 2 pawns, then only… But SF has a perpetual. Draw.

Game 46: (Leela +2)

Fish with black decides to attack the q-side and thinks it has equalized after a few moves. Leela Thinks she has a good edge. The q-side closes on move 16.b3. White pieces seem to occupy a better position to attack.

Now, white can turn attention to the center and k-side with for example Nh2, f4 or may be g4. But she prefers a slow setup with Be3, Re2, Kh1, Nd2, Nh2, Rfe1. In fact she has time since half of black pieces are on q-side and have no position to occupy on k-side. SF seems to understand that its 0 eval is too optimistic and Leela fans see a won game.

But the e-file opens and it appears that heavy pieces must occupy it and exchange. So that the game ends in a draw.

Game 47: Caro-Kann Defense, advance variation (Leela +2)

Book 1. e4 c6 2. d4 d5 3. e5 Bf5 4. Nc3 e6 5. g4 Bg6 6. Nge2 c5

A good opening for a double kill by Leela. After swapping one pawn, bishops and a knight, the k-side pawn and the center structure is blocked. The SF eval remains to 0 whereas Leela eval gives her a strong advantage.

But the engines close the q-side and despite the very optimistic Leela eval it is a draw and a very deceiving game : it started from an interesting position but the engines played it like humans that wanted a draw.

Game 48: (Leela +2)

The game begins with 15 tactical moves to reach an equal endgame where Fish is a pawn up. After some piece positioning moves and a new wild long swapping sequence the game reaches a drawn simple QP vs RNPPP endgame. This game is based all along on dynamic equality. Who said that Leela is not good at tactics ?

Game 49: KID (Leela +2)

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. h3 O-O 6. Be3 c5 7. Nf3 Qa5 8. Nd2 cxd4 9. Nb3 Qxc3+ 10. bxc3 dxe3

A good starting position from Firouzja, Alireza (2682) vs Karthikeyan, Murali (2593), 18th Asian Continental, 2019/06/11, 0-1.

In this position white has a queen for knight, bishop and a pawn, but heavy structural damages. There are good outposts for black knights, strong Bg7 (white DS are very weak), bishop pair, and white double pawn on semi-open c-file. Holding this is a pain.

After swapping black LSB for a knight on e6 square, Leela takes with f-pawn opening the f-file. SF tries to find counter play on the a-file and Leela gets a dangerous passer on the a-file that she leads to a3 after pushing h-pawn to h3.

At that point after a few shuffling moves Leela eval goes to -2 and SF eval becomes negative. But in zeitnot, with 5s increment, Leela does not manage to find the win. Draw.

Game 50: (Leela +2)

Leela plays f4 and sacs the exchange and c3 pawn to get the Bg7. One of the big problems for white is DS weaknesses, now this problem is solved and she thinks she gets an edge.

She takes the e3 pawn with the king and pushes a fawn to h6 then sacs a pawn to open lines and gets a strong attack but it is not enough. Fighting draw.

Game 51: Sicilian, Kan variation (Leela +2)

Book 1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Bd3 Nf6 6. O-O Qc7

White establishes a Maróczy Bind and starts an attack on the k-side. SF sacs two pawns to open lines, Leela sacs the exchange and returns a pawn, SF gives back the exchange. The position looks strong but the engines evaluate it equal. SF sacs a new pawn to get full play for the bishop pair. Leela returns a pawn, repulses the attack and swaps white LSB. Draw.

Game 52: (Leela +2)

With white Leela lets SF establish a strong center on LS : c6-d5-e6, against the e5-c4 setup. In such a position no pawn move is possible without strong tactical reason. The plan seems to play against the bad Bc8 and to take control on DS. SF reacts with a5 and Ba6 that seems good.

After a dubious Rd4, Leela takes control of all the DS and gets a passer, sacking the h pawn and the exchange on d6. But finally she may have missed something since her eval dropped.

After some pawns sacs to open lines, SF gets a dynamic draw.

Game 53: Dutch defence, Blackburne variation (Leela +2)

Book 1. d4 f5 2. g3 Nf6 3. Bg2 e6 4. Nh3 d5

Generally in the Dutch black plays to push e5 to liberate the bad Bc8. This is not so easy with the d5-e6-f5 setup given by the book so black establishes the solid stonewall with c6 after white logical c4 and the tension remains since moving a pawn in this situation is not good. This solid center gives black the time needed to organize pieces.

But SF plays the anti-positional 8.cxd5 and with 8…exd5 Leela eval decreases a bit. This is not a major concession by SF but this helps a lot black to develop. On move 12, white has lost its initial advantage and on move 20 the position is totally equal.

Leela starts an attack on the q-side and SF panics, pushing f4 to slow down the attack but giving the e4 square to Leela. That Leela occupies with a strong knight a few moves later. Now, she has a slight advantage.

Leela sacs a pawn on q-side and destroys it, forcing Fish to take the Ne4. Then she attacks k-side, SF reacts in the center, gives the pawn back and obtains a draw endgame.

Game 54: (Leela +2)

SF with black plays 9…dxc4 to use d5 as an outpost for a knight but definitely giving e6 and the e5 square to white that have now a durable edge.

Leela exchanges the bad DSB, a pair of knights and turns attention to the backward c-pawn on a semi-open file. SF reacts on k-side. Leela sacs the backward b3 pawns to inflict two isolated doubled e-pawns to the Fish and has now a good Ne5 against bad black LSB.

The pawns structure is fixed in center and q-side to white advantage that gets definitively the c-file with a target in c7. As a compensation, Fish gets the semi open f-file but the Ne4 controls f3 and f7 and it has less value for black than c-file for white.

Fish pushes the h-pawn to h4 and Leela closes the k-side with g4. There is not enough play for white to win and the game would end in a draw by 50 moves rule. But Leela decides to sac the h-pawn to continue. With black queen in h3 she gets the c-pawn and a Rc8 but SF sac a rook on f2 to get a perpetual.

Game 55: Pirc defence (Leela +2)

Book 1. e4 d6 2. d4 Nf6 3. Nc3 Nbd7 4. g4 h6 5. h3 e5

To react to the aggressive h3-g4 white setup, Leela quickly develops the q-side and sacs two pawns. SF accepts the d6 pawn and swaps its dangerous Nf5.

Leela starts a massive swapping combination that ends in RB+pawns endgame where white is a pawn up but with no gain perspective thanks to the OCB. After swapping rooks the OCB endgame is a draw.

Game 56: (Leela +2)

In the reverse game, Leela with white tries to maintain as long as possible a close center and castle long. But SF plays Nc4 and she must swap it, opening the b-file for SF attack. Then with the logical f4 and black f5 reaction, the center explodes. Too many lines are opened, pieces swap and an inevitable drawn endgame is on the board.

Game 57: Benoni defence (Leela +2)

Book 1. d4 Nf6 2. c4 c5 3. d5 e6 4. Nc3 exd5 5. cxd5 Bd6 6. e4 O-O 7. f4

This opening gives white a very aggressive setup in the center but a slight development disadvantage. Leela reacts immediately sacking a knight for a pawn to open the center and attack the white king. SF reacts, sacking its queen for the two most aggressive black pieces, the Bd6 and the Re8.

Developing her queen Leela, trade the c4 pawn for SF’s b2. Then she gets the central d5 pawn swapping a knight pair. SF remains with a material advantage that compensates Leela’s initiative so that it has a good edge.

SF pieces are well coordinated so that it repulses the attack. Leela now thinks to defend, secures the fragmented pawns and key squares on row 6. SF forces a rook swap.

Pawns moves stop in a position where any new move creates weaknesses without a single compensation. White wants to keep the bishop pair and swapping black LSB for white knight is only good for white. Draw by 50 moves rule is likely but Leela moves a pawn in an equal position.

The game ends in a draw after two tries by Leela to promote a pawn.

Game 58: (Leela +2)

Any Leela fan expects a win when Leela plays white in the Benoni. But SF plays the same line as Leela until it deviates with 10…Qe7 instead of 10…c4 played by Leela. This means that after the queen sac, white do not get the black c-pawn.

Leela still plays the queen sac, trades Nb4 for the DSB castle long to protect her pawns and finally gets a position she evaluates much better than whatever SF got in the previous game. Is Leela stronger than SF even in open games ?

Leela sacs the d-pawn, reorganizes pieces in the center, begins a king walk to the k-side, trades a rook pair and a-pawns. The material is balanced but Leela eval is +3 and SF eval +2. SF has two doubled b-pawns and an isolated d-pawn in an open file. Good targets for Leela rook and knight.

And finally Leela pushes the f-pawns and swaps black LSB for a knight. Now all pawns are fragmented, this is the time for picking. “Les feuilles mortes se ramassent à la pelle”, Jacques Prévert, in english “Dead leaves are picked up by the shovel.” Leela is winning! But she misses the win. A queen in open space is a real problem for Leela.

Game 59: two knights defence (Leela +2)

Book 1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 d5 5. exd5 Nxd5 6. Nxf7 Kxf7 7. Qf3+ Ke6 8. Nc3 Nb4

A very weird initial position that at first SF evaluated to +5 before finding just an edge for white as Leela. The black king is in e6 under attack and white got this after sacking a knight for a pawn.

The middle game is very confusing and finally SF gets 4 pawns for the knight with two connected central passers, one in d6. Heavy pieces and OCB remain on the board. Leela, as black, is also a knight up. Leela pieces block the dangerous d6 pawn and king is safe in h7. It seems that she will hold. After rooks swap in the open e-file the game ends in a draw.

Game 60: (Leela +2)

Leela plays the clever 11.Qd1 instead of Fish 11.Qe2 in the previous game. The advantage of this move is to maintain the e-file open for the attack. The fish has difficulty regrouping its pieces and sacs 3 pawns. But Leela does not have enough play. Draw.

Game 61: Trompowsky attack (Leela +2)

Book 1. d4 Nf6 2. Bg5 c5 3. d5 Qb6 4. Nc3

As often in tactical open fight, a lot of pieces swap in a wild and quick fight. When smoke disappears Fish has a queen for rook, bishop and pawn. Leela pawn structure is sound and looks like a fortress. On move 28 this is a draw but the game will continue until move 190.

Game 62: (Leela +2)

As in the previous game the lines open quickly. The opposite castle and a wild white attack forces SF with black to take a perpetual on move 21.

Game 63: king indian attack (Leela +2)

Book 1. e4 e6 2. d3 d5 3. Nd2 c5 4. g3 Nc6 5. Bg2 Nf6 6. Ngf3 Be7 7. O-O O-O 8. Re1 b5 9. e5 Nd7 10. Nf1

At the end of the book black dominates the q-side and the center is blocked but not robust. white as a small advantage on the k-side.

Leela, as black, attacks the center and the q-side sacking the a-pawn. SF plays to maintain the center and get a strong control on DS. Leela has just a slight edge. Pawns swap and pieces remain on the board. Finally an OCB drawn endgame is reached.

Game 64: (Leela +2)

Leela, as white, first plays to slow down black attack on the q-side. Then she reinforces the center and launches a wild attack on the k-side, taking control of the DS.

She abandons the q-side where she sacs two pawns. She pushes a fawn in f6. SF launches a furious counter attack sacking a knight for the fawn. Pawns and pieces do not count. Leela gets a mate threat and Fish has to take a perpetual in a position where it has sacked the exchange for 6 pawns.

Game 65: KID (Leela +2)

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7

The book gives 6 plies in the King Indian Defence. SF pushes an early d5 and launches an attack on the k-side with h3-g4. Leela reacts attacking the q-side and the center where she sacs the Nf6 for 2 pawns.

She finally gets a rook endgame where SF has a bishop for 3 pawns. Draw.

Game 66: KID (Leela +3)

Leela gets a bind with c4-e4-f3 and Fish swaps the Bg3 for Nc3 inflicted doubled d-pawns to Leela.This is a totally anti-positional decision. The c-file is not open, there is no means to attack the double pawns and now Leela dominates the DS.

She finds a combination that swaps pieces and gets a rook endgame with a bishop for 4 fragmented pawns. This is the picking season : Leela grabs each pawn, one after another, keeping her g-pawn for queening. Nice win from a standard position.

Game 67: Sicilian, Kan (Leela +3)

Book 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Qc7 5. Nc3 e6 6. Ndb5 Qb8 7. Be3 a6 8. Bb6 axb5 9. Nxb5 Bb4+

A sub variation that forces Leela to give up the queen and a pawn for the two knights and the DSB. Fish attacks with pawns on all the board and gets a drawn endgame.

Game 68: (Leela +3)

Leela in a more positional style decides to not take the queen to reach a position with a good space advantage but where SF is a knight up for two pawns and has lost castling right.

Leela gets 3 connected passers that leads to a drawn endgame.

Game 69: QGD, Slav variation (Leela +3)

Book 1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 dxc4 5. e4 b5

A very sharp opening. White gives a pawn for better development and a very huge center. Black compensation is a 4 to 2 pawn advantage on the q-side.

Leela returns the pawn to accelerate development and gets a drawn endgame.

Game 70: (Leela +3)

Leela varies from the previous game with the interesting 8.Qc2. Then, playing 10…b4 SF lost all advantage since the q-side black pawns are now impossible to defend. Leela equalizes but she does not succeed to take an advantage. Draw.

Game 71: Nimzovich defense (Leela +3)

Book 1. e4 Nc6

A perfect opening for a double kill by Leela. SF plays badly in the opening. The weird Bh5+ leads to the Karpovian Kd7 by Leela. The anti-positional g4 blocks the Bh5. So at the end of the opening phase, Leela gets a dangerous protected passer in e4 and she equalizes easily sacking the queen for rook and knight.

A few moves later she sacs the exchange for a pawn and begins to advance the e4 pawn. As SF gives the exchange back for the pawn, the endgame is a draw.

Game 72: (Leela +3)

The opening is very orthodox for a closed French like pawn structure. SF castles long and Leela short. The two opponents begin a race against the king, alternating defensive and offensive moves. Fighting draw.

Game 73: Benoni defense (Leela +3)

Book 1. d4 Nf6 2. c4 c5 3. d5 e6 4. Nc3 exd5 5. cxd5 d6 6. e4 g6 7. f4 Bg7 8. Bb5+ Nfd7 9. Be2

An opening that generally Leela wins as white and draws as black. Leela decides to swap the good Bg7 for the Nc3 to win the e4 pawn. Is this not too risky ? She won a game as white when SF did the same.

No, because there are two big differences. Her LSB is not blocked in c8 but is in g4 and she swaps it for the Nf3. The q-side is not blocked and she immediately opens it and gets a passer in the a-file. She trades a pair of rooks in the e-file, sacs the a-pawn and trades the rooks in the a-file. Now there is not enough play for SF. Draw.

Game 74: (Leela +3)

In the opening SF plays the ugly Na6 that blocks its q-side so that Leela has not to worry about that side. She launches a pawn attack on the k-side, sacs the e4 pawn to get a fawn in f6 and to trap the Bh8. She then sacs the d5 pawn to swap the queens and the LSB.

Is it enough to win ? SF sacs the Bh8 for 2 pawns and gets a drawn rook ending where it has 3 pawns for a knight.

Game 75: Pirc (Robatsch) defense (Leela +3)

Book 1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2

SF plays badly in the opening. It starts with an aggressive Bh6 but then plays the slow h3 instead of continuing with moves like Bc4. Leela counter attack in the center and at move 11 the game is nearly a draw.

After queen and central pawns swap. The rooks exchange in the open central files and the endgame with symmetric pawns structure leads to an inevitable draw.

Game 76: (Leela +3)

Leela manages to get a strong bind but does not succeed in converting to win. Draw.

Game 77: QGD, Chigorin defense (Leela +2)

Book 1. d4 d5 2. c4 Nc6 3. Nf3 Bg4 4. cxd5 Bxf3 5. gxf3 Qxd5 6. e3 e5

This variation gives a strong edge to white since the Nc6 blocks the useful c-pawn. By sacking the b-pawn Leela manages to double her rooks in the b-file to get a very bad rook ending that SF manages to win.

Game 78: (Leela +2)

Leela gets a rook endgame with a bishop pair against knights. She has a strong central pawn mass but SF manages to block the pawns. Draw.

Game 79: Ruy Lopez (Leela +2)

Book 1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 5. O-O Be7 6. Re1 b5 7. Bb3 d6 8. c3 O-O 9. h3 Bb7 10. d4 Re8 11. Nbd2 Bf8 12. a4 h6 13. Bc2 exd4 14. cxd4 Nb4 15. Bb1 bxa4

A very long opening line that looks like a draw at the end of the book. Draw.

Game 80: Ruy Lopez (Leela +2)

Leela gets more pressure with white but it is still a draw.

Game 81: KID (Leela +2)

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. f3 O-O 6. Be3 e5 7. d5 Nh5 8. Qd2 Qh4+ 9. g3 Nxg3 10. Qf2 Nxf1 11. Qxh4 Nxe3 12. Ke2 Nxc4

A very long opening line where black sacked the queen for the bishop pair and two pawns. Begins a furious dog fight where pawns and pieces count less than threats and speed and that ends in a drawn endgame.

Game 82: KID (Leela +2)

In the reversed game, Leela as often is more positional and the play is slower and more understandable. But after the inevitable 20.Ne3 Nxe3 21.Kxe3 SF remains with a strong Nc5 that controls the key squares and prevents the queen from entering behind black lines.

It is likely that white will be forced to sac the exchange for that knight. But Leela does not choose this solution, she prefers to place her pieces on good squares. At this point no progress is possible and she simplifies in a drawn endgame with the exchange for two pawns.

Game 83: Sicilian (Leela +1)

Book 1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Qb6 5. Nb3 a6 6. Nc3 d6 7. Bf4 Nf6 8. Qd2 Be7 9. O-O-O O-O

A sharp opening where white looks like winning at the end of the book : the Qb6 is misplaced in an opposite castle situation. Logically Fish attacks the k-side with pawns and Leela must replace the Q to c7. Then Fish plays the deadly 15.g6 and wins.

Game 84: Sicilian (Leela +2)

Leela plays this game in Leela positional fashion. She pushes a fawn to h6 and then after SF e5 plays against the d6 backward pawn. She slowly ameliorates piece positions and finally wins a pawn.

After a long shuffle time she finally sacs the exchange to win a pawn and enter into black position. The black pawns fall one after the other. Win.

Game 85: Dutch defense (Leela +2)

Book 1. d4 f5 2. Nf3 d6 3. c4 c6 4. Nc3 Nf6

Leela occupies the center and attacks the k-side. SF attacks the q-side, too slowly. On move 21, the game is equal. Leela does not have enough play to win. Draw.

Game 86: (Leela +2)

SF attacks the q-side. Leela contains the attack and stabilizes the center. She launches an attack on the q-side. Finally she plays on the two wings but she does not have enough threats to win. Draw.

Game 87: Modern (Robatsch) defense (Leela +1)

Book 1. e4 g6 2. d4 c6

The opening transposes to a KID with classical plan : SF attacks the q-side and Leela the k-side. On move 33 all pieces are still on board and the position is equal. Pieces swap, the e-file opens, and Leela blunders. SF wins.

Game 88: (Leela +2)

The reversed game also transposes to a classical KID. But SF as black decides to attack the q-side. This is a dubious plan. The c-file opens and Leela gets a strong advantage in that file.

Leela exchanges the LSB and a pair of knights and gets a winning endgame : all her pawns are on LS so that SF DSB has no target.

Game 89: Nimzo Indian defense (Leela +2)

Book 1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. e3 O-O 5. Bd3 b6 6. e4 Bb7 7. e5

This opening honours Leela since it had been discovered by ID395 against Gull. A video on it.

SF has an attack on k-side for an exchange sac. Leela repulses the attack and gives back the exchange to reach a drawn endgame.

Game 90: (Leela +2)

Leela leads the attack against black king in a slightly different way and finally gets a small edge. But SF finds a perpetual. Draw.

Game 91: Two knight defense (Leela +2)

Book 1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 Bc5

A very sharp opening likely winning for white. SF plays 5.Bxf7+ getting a pawn and denying castling for black.

SF lets Leela open the g-file and the d-file and to advance the a-pawn in a3. Difficult to say what was wrong in the Fish attack. With this pawn structure, Leela manages to regroup pieces on good squares and she holds.

She finally gets a drawn queen endgame. What a game!

Game 92: (Leela +3)

Leela plays the same 5.Bxf7+ Ke7 line. Have a look at the position : white is a pawn up (f7), black lost castling right (Ke7) and there is nothing as a compensation. If Leela plays her positional usual style, she must win. A giuoco piano with great advantages.

Leela simplifies to a winning endgame swapping a rook and a pawn for bishop and knight. She has a good pawn structure, the center and the bishop pair. In particular she is very strong on LS. The rest is a technical matter.

Game 93: QGD, Chigorin variation (Leela +2)

Book 1. d4 d5 2. Nf3 Nc6 3. g3 Bg4 4. Bg2 Qd7 5. O-O O-O-O 6. c3

A new very sharp opening likely winning for white. Long castling is always dangerous for black and with opposite castles it is often deadly.

T8.T40.610 lost this opening against “the engine that can not be named”, version 3, the winner of TCEC S4, in a S15 bonus with all champions.

At the end of the book, Leela plays the slow 6…f6 and it is likely that 6…h4 is the only playable move if it is possible to hold this : this game is a speed race. SF totally outclasses Leela in that game.

Game 94: (Leela +3)

White has more time to play the position. Leela stabilises the center, pushes her pawns toward SF king, and swap the Nc6. Slow but efficient. SF swaps the Nf3 for the Bg4 and opens the h-file but lacks pieces to continue.

Leela pushes a fawn in a6, SF sac the exchange for Leela strong LSB. She reduces material to a winning R+pawns vs N+pawns endgame, thanks to the fawn.

Game 95: French, advance variation (Leela +4)

Book 1. e4 e6 2. d4 d5 3. e5 b6

A dubious variation of the french defense that gives a good edge to white. The recommended move is 4.Bb5+ but SF plays the not so good c3.

Leela plays like in the manual : she attacks the head and the tail of the central pawn chain, sacs a pawn to activate her bad LSB, pushes g5 pawn in front of the enemy king. On move 20 the position is roughly equal.

SF makes a positional blunder that gives Leela two linked passers on the q-side and has to sac the LSB for the pawns. Leela wins the endgame.

Game 96: (Leela +5)

Leela does not play the recommended 4.Bb5+ but pushes h4 her speciality. She gets that way a big space advantage on the k-side and prevents SF to castle short. SF reacts by opening the c-file and Leela sac the d4 pawn to use the square as an outpost for a knight.

She gets a quasi-winning endgame that she converts, SF helping a lot sacking two pawns under pressure and self-trapping a bishop.

By the way she wins the SuFi since there are not enough games for SF to recover.

Game 97: KID, Mar Del Plata variation (Leela +5)

Book 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Be3 f5 11. f3 f4 12. Bf2 g5

Classical plans, SF attacks the q-side, Leela the k-sides. Leela gives a pawn, pieces swap, she gets a RB+pawns vs RN+pawns drawn endgame because of a bad DSB.

Game 98: (Leela +6)

Leela sacs the c-pawn to open lines on the q-side. Then, on move 20 she plays the good g4 that stop the black attacks on k-side. Fish thinks it has an edge after that move. In fact this is the contrary. This g4 move closes the k-side and only Leela has threats now.

After some long Leela maneuvers to place pieces on good squares, Fish commit suicide playing a5 that open more lines in white favour. The rest of the game is easy for Leela.

Game 99: Sicilian, English attack (Leela +5)

Book 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Be3 e6 7. f3 Be7 8. Qd2 Nc6 9. O-O-O O-O

The English attack is characterized by e4-f3-g4-h4 with 0-0-0, Qd2, Be3 and considered very dangerous. Fish plays 22.Kb1 and Leela thinks the position is very bad. A few moves later, she plays g6, Fish opens the k-side and wins.

Game 100: (Leela +5)

Leela also chooses the classical setup of the English attack and plays an early prophylactic Kb1. SF is not impressed and thinks the position is equal.

Leela pushes a fawn in h6! After Leela played a3, SF swaps the DSB for the Nc3, inflicting doubled isolated c-pawns to Leela but also giving more value to the fawn. Now, white has a simple plan : reduce material and win an endgame based on the two weaknesses principle, exploiting the q-side majority and the h6-fawn with a strong play on DS.

She applies my plans ! In fact it is Leela’s plan, I saw so many games she played like this that I anticipated her moves. SF sacs the exchange to get the g5 pawn that reinforces the fawn. Queen swaps and she gets a very good RRB+pawns vs RBN+pawns endgame with the a-passer. But she draws.