avoid-move test positions

[bob]

Does anyone have a smallish set of "avoid-move" type positions? Those where a move looks good until it ultimately fails low and gets replaced, or just gets replaced without failing low? Preferably positions that don't need way-deep searches.

I am still playing around with ideas to recognize positions that are the opposite of "easy moves" and realized that I didn't have anything but "easy-move" type positions saved away...

The ultimate would be positions where the best move looks worse and worse until it is replaced, as that is a key type of position to recognize and say "use more time here..."

[sje]

Perhaps the best way of producing these test positions is via a scan of GM middlegames with a long think for each move. Apply a filter for those with a late selection change AND with a significant evaluation gain.

[Rebel]

Perhaps too old, but you can have a look.

http://www.top-5000.nl/epd/am.zip

[JVMerlino]

The Eigenmann Endgame suite also has a handful of am positions, if you hadn't already thought of those:

4k3/8/3PP1p1/8/3K3p/8/3n2PP/8 b - - am Nf1; id "E_E_T 010 - B vs S";
6k1/5p2/P3p1p1/2Qp4/5q2/2K5/8/8 b - - am Qc1+ Qe5+; id "E_E_T 011 - D vs D";
r5k1/3R2p1/p1r1q2p/P4p2/5P2/2p1P3/5P1P/1R1Q2K1 w - - am Rbb7; id "E_E_T 019 - D&T&T vs D&T&T";
8/1p4k1/pK5p/2B5/P4pp1/8/7P/8 b - - am g3; id "E_E_T 020 - L vs B";
8/8/2p1K1p1/2k5/p7/P4BpP/1Pb3P1/8 w - - am Kd7; id "E_E_T 023 - L vs L";
8/p4p2/1p2k2p/6p1/P4b1P/1P6/3B1PP1/6K1 w - - am Bxf4; id "E_E_T 025 - L vs L";
3b3k/1p4p1/p5p1/4B3/8/7P/1P3PP1/5K2 b - - am Bf6; id "E_E_T 027 - L vs L";
8/3k1p2/n3pP2/1p2P2p/5K2/1PB5/7P/8 b - - am Kc6 b4; id "E_E_T 029 - L vs S";
8/8/5p2/5k2/p4r2/PpKR4/1P5P/8 w - - am Rd4; id "E_E_T 054 - T vs T";
8/6p1/1p5p/1R2k3/4p3/1P2P3/1K4PP/3r4 b - - am Rd5; id "E_E_T 059 - T vs T";
8/1pp1r3/p1p2k2/6p1/P5P1/1P3P2/2P1RK2/8 b - - am Rxe2+ Re5; id "E_E_T 066 - T vs T";
4k1r1/pp2p2p/3pN1n1/2pP4/2P3P1/PP5P/8/5RK1 b - - am Nf8; id "E_E_T 16b - T&S vs T&S";
8/3k3p/1p2p3/p4p2/Pb1Pp3/2B3PP/1P3P2/5K2 w - - am Bxb4; id "E_E_T 26b - L vs L";

jm

[jdart]

A few from my latest test suite:

8/5pk1/3p1bp1/1B5p/2P1PP2/3Q2PK/5q2/8 w - - am Qxd6; id "arasan16.14"; c0 "Arasan-Rookie, ICC 2012";

2kr3r/pp4pp/4pp2/2pq4/P1Nn4/4Q3/KP2B1PP/2RR4 b - - am Qxg2; id "arasan16.184"; c0 "Eklund-Berthelsen, DSU Jubilee 100-B corr 2003";

8/5q2/4kp1p/7R/6P1/5P2/6K1/8 w - - am Rxh6; id "arasan16.213"; c0 "Salov-Korchnoi, Wijk aan Zee 1997";