Passed pawn evaluation

[Daniel Mehrmann]

Hello !

Yesterday i had a difficult game versus Cheng during HGM's online tourney.

Fruit vs Chang after move 25...a2:

[d]r2q2r1/3b2nk/4p1pp/1nBpPp2/2pP1PP1/2P1N3/p1PQB2P/R4R1K w - -

Well, i think white is already lost in a long term view and it seems to be pretty difficult to evaluate the passed pawn on A2 correctly.
Basicly the rook on A1 is catched and out of the game. But in endings it could be still a draw and the pawn looks weak at this moment.

Also Komodo has a lot of trouble to evaluate this position. If you go through the moves Komodo shows a lot shuffle moves like Fruit itself.

I'm thinking about to give a small bonus for the passer in the opening if a piece is catched by the pawn.

What do you think is a good evaluation in this case ?

PGN:

[pgn][Event "ICS rated blitz match"]
[Site "winboard.nl"]
[Date "2015.12.12"]
[Round "-"]
[White "FruitReload"]
[Black "ChengRB"]
[WhiteElo "1969"]
[BlackElo "2102"]
[Result "0-1"]

1. e4 {+0.01/1 23} e6 2. d4 {+0.01/1} d5 3. Nc3 {+0.01/1}
Bb4 4. e5 {+0.01/1} Ne7 5. a3 {+0.01/1} Bxc3+ 6. bxc3
{+0.01/1} c5 7. Nf3 {+0.01/1} O-O {-0.05/20} 8. Bd3
{+0.28/20 13} h6 {-0.10/22} 9. O-O {+0.35/20 8} a5
{-0.12/20} 10. Nh4 {+0.53/20 8} c4 {-0.17/21} 11. Be2
{+0.46/21} Na6 {-0.17/20} 12. a4 {+0.60/20 8} Nc7
{+0.00/21} 13. Bh5 {+0.47/18 8} Kh7 {-0.10/21} 14. Ba3
{+0.62/21 8} Bd7 {+0.00/22} 15. Bc5 {+0.62/22 0.5} g6
{+0.00/22} 16. Bg4 {+0.63/20 10} b5 {-0.07/22} 17. axb5
{+0.55/21 10} Nxb5 {-0.12/24 8} 18. Qd2 {+0.54/21 6} Rg8
{-0.09/22 8} 19. g3 {+0.54/21 8} a4 {-0.10/22 17} 20. Ng2
{+0.58/19 7} Nf5 {-0.05/21 10} 21. Be2 {+0.61/19 9} f6
{+0.18/22} 22. g4 {+0.63/21 0.7} Ng7 {+0.07/22} 23. f4
{+0.63/20 0.4} f5 {+0.05/21 9} 24. Ne3 {+0.35/18 7} a3
{+0.64/22 6} 25. Kh1 {+0.33/18} a2 {+0.67/22 11} 26. Rg1
{+0.24/19 0.2} Be8 {+0.76/20 3} 27. Bf3 {+0.49/10 10} Bf7
{+0.85/21 15} 28. Qe1 {+0.37/18} Ra4 {+0.87/23 2.5} 29. Bb4
{+0.00/17 13} Ne8 {+0.91/23} 30. g5 {+0.23/18 6} Nec7
{+1.21/23 2.8} 31. Nd1 {+0.06/19 2.5} h5 {+1.21/25 9}
32. Qe2 {+0.00/21 0.5} Qd7 {+1.21/26 6} 33. Qf2 {+0.00/22}
Na7 {+1.21/27 8} 34. Nb2 {-0.14/19 9} Ra6 {+1.21/27}
35. Qd2 {-0.23/21 2.6} Rb8 {+1.21/26 6} 36. Nd1 {-0.37/21
10} Nc6 {+1.20/27 4} 37. Be2 {-0.45/21 7} Nb5 {+1.20/23 6}
38. Bc5 {-0.31/21 7} Qd8 {+1.21/24 3} 39. Qe3 {-0.39/21 7}
Qa5 {+1.27/22 7} 40. Qg3 {-0.43/21 1.4} Rg8 {+1.25/22 4}
41. Re1 {-0.55/20 5} Raa8 {+1.55/20 0.1} 42. Qe3 {-0.96/20
9} Rgb8 {+1.69/21 6} 43. Bf3 {-1.32/21 32} Rb7 {+1.67/22
2.5} 44. Rg1 {-1.28/22} Rab8 {+1.89/22 6} 45. Qf2 {-1.51/20
5} Nc7 {+1.89/25 2.8} 46. Bb4 {-1.50/20} Nxb4 {+1.94/26}
47. cxb4 {-1.51/22} Rxb4 {+1.87/28 4} 48. c3 {-1.51/23 0.1}
Ra4 {+1.89/26 3} 49. Rg2 {-1.71/22 0.5} Ra8 {+1.89/26 7}
50. Qe3 {-1.51/22 4} Be8 {+1.89/28 0.5} 51. Rb2 {-1.51/22
5} Ra3 {+1.89/26 2.3} 52. Rc2 {-1.51/20 21} Nb5 {+1.89/28
4} 53. Bg2 {-1.51/22 4} Qa7 {+1.89/23 6} 54. Qd2 {-1.51/22
4} Qa4 {+1.90/24 0.5} 55. Rb2 {-1.51/9 4} Bc6 {+1.89/26}
56. Rc2 {-1.51/21 4} Kg7 {+1.89/25 5} 57. Rb2 {-1.51/22
2.6} Kg8 {+1.90/25 7} 58. Bh3 {-1.51/21 2.9} Qa6 {+1.89/24
7} 59. Bg2 {-1.51/23 4} Qa5 {+1.89/25 6} 60. Rc2 {-1.51/25
8} Kg7 {+1.89/27 5} 61. Bf3 {-1.51/24 4} Qa4 {+1.89/24 4}
62. Rb2 {-1.51/21 1.8} Be8 {+1.90/23 4} 63. Bg2 {-1.51/20
3} Bd7 {+1.90/24 4} 64. Bf3 {-1.51/21 2.2} Qa5 {+1.89/25 6}
65. Rc2 {-1.51/21 6} Kg8 {+1.90/26 5} 66. Bg2 {-1.51/21
2.1} Kh7 {+1.90/26 0.2} 67. Bf3 {-1.51/20 1.7} Bc6
{+1.90/25 4} 68. Bg2 {-1.51/20 1.8} Qa4 {+1.90/24 4}
69. Rb2 {-1.51/21 2.4} Bd7 {+1.90/25 3} 70. Bf1 {-1.51/22
2.0} Kg8 {+1.89/26 4} 71. Bg2 {-1.51/22 2.6} Qa6 {+1.89/25
1.5} 72. Bh3 {-1.51/20 1.4} Kg7 {+1.89/25 3} 73. Rc2
{-1.51/21 1.5} Qa5 {+1.90/26 0.3} 74. Bg2 {-1.51/23 2.3}
Ra6 {+1.90/26 1.9} 75. Bf3 {-1.51/21 1.9} Bc6 {+1.90/23 3}
76. Be2 {-1.51/20 2.4} Kg8 {+1.90/26 4} 77. Bf1 {-1.51/21
1.3} Kh7 {+1.89/24 3} 78. Qf2 {-1.51/22 1.7} Kg7 {+1.89/26}
79. Qd2 {-1.51/21 1.6} Ra8 {+1.89/24 3} 80. Bh3 {-1.51/22
2.3} Qa4 {+1.90/24 2.9} 81. Rb2 {-1.51/22 2.4} Qa6
{+1.89/23} 82. Bg2 {-1.51/22 2.0} Qa5 {+1.89/26 0.1}
83. Rc2 {-1.51/23 1.2} Kf7 {+1.90/25 3} 84. Bf3 {-1.51/20
2.2} Bb7 {+1.89/23 2.8} 85. Rb2 {-1.51/19 1.4} Kg7
{+1.90/22 2.1} 86. Be2 {-1.51/20 1.7} Ra6 {+1.89/22 4}
87. Bf3 {-1.51/18 1.1} Bc6 {+1.84/21 4} 88. Rc2 {-1.51/21
1.8} Be8 {+1.80/21 0.4} 89. Bg2 {-1.51/21 1.3} Kf7
{+1.62/22 4} 90. Bh3 {-1.51/20 1.1} Kg8 {+1.60/22 3}
91. Qe3 {-1.67/21} Qb6 {+1.82/21 2.2} 92. Bf1 {-1.51/20
1.2} Nc7 {+1.87/22 3} 93. Rb2 {-1.51/20 1.7} Qa7 {+1.80/24
3} 94. Qd2 {-1.23/17 1.0} Ba4 {+1.86/24 2.2} 95. Nf2
{-1.37/19 1.8} Bb3 {+2.28/22 41} 96. Nh3 {-1.39/22} Nb5
{+2.16/22 1.9} 97. Ng1 {-1.39/21 1.1} h4 {+1.92/21 1.8}
98. Ne2 {-1.39/21 0.5} Qb7 {+1.96/20 1.5} 99. Bh3 {-1.39/20
1.7} Ra7 {+2.32/22 0.4} 100. Bg2 {-1.39/19 1.2} Qc7
{+2.39/23 5} 101. Qe1 {-1.39/19 1.5} Qa5 {+2.38/22 1.1}
102. Bf1 {-1.39/20} Rh7 {+2.38/22 1.6} 103. Kg1 {-1.39/18
5} Rg7 {+2.38/24 1.3} 104. Bh3 {-1.39/19 1.5} Rb7 {+2.39/22
1.1} 105. Bf1 {-1.39/19 0.9} Bd1 {+3.23/19 0.2} 106. Rbxa2
{-2.75/16 1.7} Rxa2 {+4.63/23 1.6} 107. Rxd1 {-4.42/20} Ra7
{+5.01/23 1.3} 108. Rb1 {-4.63/20 1.1} Ra1 {+5.65/24 2.8}
109. Kg2 {-5.11/20 1.8} Qa2 {+6.94/24 1.3} 110. Rxa1
{-6.39/22 7} Qxa1 {+8.81/28 1.5} 111. Kf2 {-7.17/22 1.1}
Qxe1+ {+9.97/27 1.2} 112. Kxe1 {-9.01/25} Ra1+ {+10.89/26
1.0} 113. Kf2 {-9.11/26} Ra3 {+11.13/25} 114. Bg2
{-11.95/25 2.5} Nxc3 {+13.68/25 1.0} 115. Bf3 {-15.25/22
0.2} Nxe2 {+13.97/22 0.9} 116. Bxe2 {-19.69/25} Ra1
{+29.63/26 0.8} 117. h3 {-20.34/26 4} c3 {+99.69/27 1.3}
118. Bd3 {-21.23/21} Rh1 {+99.70/27 0.8} 119. Ke2
{-20.04/22} Rh2+ {+99.73/26 1.2} 120. Kf1 {-25.69/20 2.5}
Rd2 {+99.75/30 1.1} 121. Bb1 {-99.76/20 0.9} Rd1+
{+99.76/30 0.8} 122. Ke2 {-99.78/22} Rxb1 {+99.79/34 0.1}
123. Kd3 {-99.80/32} Rc1 {+99.80/37 0.1} 124. Ke2
{-99.82/46} Rh1 {+99.83/48 0.1} 125. Kd3 {-99.84/63} Rxh3+
{+99.84/61 0.1} 126. Kc2 {-99.86/63} Rf3 {+99.87/82 0.1}
127. Kb3 {-99.88/63 0.1} c2+ {+99.88/120 0.1} 128. Kb2
{-99.90/63} h3 {+99.91/120 0.1} 129. Kc1 {-99.92/63} h2
{+99.93/120 0.1} 130. Kd2 {-99.94/63} c1=Q+ {+99.94/120
0.1} 131. Kxc1 {-99.96/63} Rf2 {+99.97/120 0.1} 132. Kb1
{-99.98/63} h1=Q# {FruitReload checkmated} 0-1[/pgn]

[stegemma]

I think that there's no way to solve this problem just adding a bonus. the position could be lost or win depending on the situation on the king side, where white could attack with g5. This move is not good, now, for tactical reasons but just moving the black knight could becomes good. so it is more a tactical problem than a positional one, IMHO.

[whereagles]

Didn't go through it deeply, but seems like if the king side can be blocked, maybe the white N can go to c2 and exert some pressure. Trouble is, black will bring the heavy guns into the Q-side and blast it

[hgm]

Isolated 8th-rank Pawns can go either way. You lose them, or the opponent has to give a minor for them when they promote. (And in the mean time they bind at least a minor.) That is a difference between -1 and +2.25.

I don't think much can be gained by trying to tune a bonus for this. Whatever it is, you will be off by more than 1.5 if the unfavorable case happens, and if the bonus is near the average, because statistically both outcomes occur eaually often, you will always be off by > 1.5, except that you don't know in which direction.

If you really want to solve this, the only way is to let the search solve it, by searching branches where the Pawn is on 7th rank deeper.

[Ferdy]

Hello !

Yesterday i had a difficult game versus Cheng during HGM's online tourney.

Fruit vs Chang after move 25...a2:

[d]r2q2r1/3b2nk/4p1pp/1nBpPp2/2pP1PP1/2P1N3/p1PQB2P/R4R1K w - -

Well, i think white is already lost in a long term view and it seems to be pretty difficult to evaluate the passed pawn on A2 correctly.
Basicly the rook on A1 is catched and out of the game. But in endings it could be still a draw and the pawn looks weak at this moment.

Also Komodo has a lot of trouble to evaluate this position. If you go through the moves Komodo shows a lot shuffle moves like Fruit itself.

I'm thinking about to give a small bonus for the passer in the opening if a piece is catched by the pawn.

What do you think is a good evaluation in this case ?

Talking about a passer, normally a passer has a bonus, and when this passer is blocked, this bonus must be reduced. So in this case the passer bonus in A2 should be reduced.
It is the passer in A2 that is catched by the rook in A1 and not the rook is catched by the pawn in A2 because the rook can still move and the pawn cannot.

Now when we analyze the rook in A1, its mobility is reduced as its vertical movement has been limited, but it is not really too bad as that rook can still move horizontally.
This rook will surely get penalized in its mobility evaluation, and if you want to penalize it more, perhaps consider more penalty for having zero mobility on its vertical movement assuming this was not considered in the rook mobility.

On the other hand the penalty of blocked passer depends on the rank it is blocked, depends on the bonus this passer gets at that location, depends on the phase of the game and depends on what kind of piece has blocked this passer.

[Fabio Gobbato]

It is difficult IMHO to have a perfect passed pawn evaluation for all the positions.
I found useful in my engine to give a big malus (related to the rank) if the passed pawn's stop square is blocked or attacked and a smaller malus if the path to promotion is blocked or attacked.

[hgm]

Now when we analyze the rook in A1, its mobility is reduced as its vertical movement has been limited, but it is not really too bad as that rook can still move horizontally.

The Rook is nearly useless, because it is confined to the back rank. This is far, far worse than just missing part of the mobility. If there would have been a white Pawn on a2, blocked by a black Pawn on a3, Ra1 would have the same mobility as it has now, but two moves later it could be anywhere. It is available for use, and the penealty for low mobility is just a temporary setback, which costs you only an extra move to correct, and is tuned accordingly. You cannot threaten anything with a passer-blocking Rook from the back-rank (except for checking a King).

It really deserves a far larger penalty than that derived from mobility. That Rook is not even worth half a normal one... If you would block a passer with a Knight or a Bishop the passer would not block any of the moves. Yet the piece is effectively tied down by the passer.

[Sven]

Now when we analyze the rook in A1, its mobility is reduced as its vertical movement has been limited, but it is not really too bad as that rook can still move horizontally.

The Rook is nearly useless, because it is confined to the back rank. This is far, far worse than just missing part of the mobility. If there would have been a white Pawn on a2, blocked by a black Pawn on a3, Ra1 would have the same mobility as it has now, but two moves later it could be anywhere. It is available for use, and the penealty for low mobility is just a temporary setback, which costs you only an extra move to correct, and is tuned accordingly. You cannot threaten anything with a passer-blocking Rook from the back-rank (except for checking a King).

It really deserves a far larger penalty than that derived from mobility. That Rook is not even worth half a normal one... If you would block a passer with a Knight or a Bishop the passer would not block any of the moves. Yet the piece is effectively tied down by the passer.

I think this is not quite correct. Whether the Ra1 can move horizontally without losing immediately does in fact depend on a lot of other factors, including for instance the possibility to move away by giving check or threatening something serious, or including any other white pieces also defending a1. For the latter, simply imagine another white rook or queen on the first rank (d1 up to h1), a diagonally moving piece on the diagonal b2-h8 defending a1, or a knight on c2/b3, or a king on b2. Additionally, sometimes there are tactical shots (mostly in endgames, though) involving "rook takes pawn" surprises. So it is certainly correct that the Ra1 deserves a penalty compared to a rook with more mobility and active resources, but I would restrict that penalty mostly to the effect of reduced mobility and less to the situation of blocking a passed pawn.

[Evert]

I think the passed pawn is the least of it, to be honest. The doubled pawn on the c-file and the corresponding loss of control of the b-file (including the important b1), the lack of space on the a file (which,to be fair, is due to the passer on a2) and the inferiority of white's bishop pair seem much more relevant to me.

Sure, some of these are linked to the passer, but they are secondary effects of the passer, and they would have to be scored separately from it.

[Ferdy]

The Rook is nearly useless, because it is confined to the back rank.

Not really, as rooks should be developed late in the game, also the rook is not only blockading there it actually attacks the passer, unlike a blockade from minors. This would mean that the passer needs a constant protection when the blockader is a rook.