[Event "?"]  [Site "?"]  [Date "2017.11.16"]  [Round "?"]  [White "Weekly Chess Study #24"]  [Black "Kuznetsov study"]  [Result "1/2-1/2"]  [SetUp "1"]  [FEN "2r5/8/1P6/Pk6/1p2p3/3p4/1R1P4/1K6 w - - 0 1"]  [PlyCount "17"]   1. a6 $1 {[pgndiagram] The intended way of drawing this position. Amazingly,  the engine doesn't realize this is a draw}   (1. b7 $2 {Doesn't help} Rb8 2. a6  Kxa6 3. Rxb4 Rxb7 4. Rxb7 Kxb7 {With a lost pawn endgame})   (1. Ka2 $1 {Is not  mentioned by the author of the study and it might be the flaw. It seems to be  the simplest way to draw this study} Kxa5 2. Rb1 $1 {[pgndiagram] And I don't  see a win for Black here} Rc2+   (2... e3 3. dxe3 Rc2+ 4. Kb3) 3. Kb3 Kxb6 4. Re1  {With a comfortable draw. Thus it appears that the study is flawed. But the  main line is still worth going through.}) 1... Kxa6 2. Rxb4 e3 3. b7 Rd8 $1 {  [pgndiagram] The only try for the win} 4. Kc1 $1 {The only move}   (4. b8=N+ {  Doesn't work yet} Rxb8 5. Rxb8 exd2 {[pgndiagram] And the rook can't cover d1})  4... e2 $1 5. b8=N+ $3 {[pgndiagram] A great move} Rxb8   (5... Ka7 6. Nc6+ Ka6   (  6... Ka8 $4 7. Ra4+ Kb7 8. Nxd8+ {Even loses}) 7. Nb8+ {Is a repetition}) 6.  Ra4+ Kb5 7. Ra5+ $3 Kb4 8. Ra4+ Kc5 9. Rc4+ {[pgndiagram] And White draws with  the help of rampaging rook} 1/2-1/2