http://www.youtube.com/watch?v=XyMrK...ailpage#t=770s

Just for fun, I'm running Houdini 1.5a alongside this demonstration, and at this point, there was a wild disagreement between them. Which is interesting because apparently the Tablebase is the equivalent of God, and Houdini thinks its moves are massive blunders, but it's hard to see why because the tablebase demonstration doesn't include the refutations of Houdini's preferred moves.

Here's what Houdini thinks after a few minutes..



The tablebase then shows white's next optimal move as Be7.

Houdini drops the evaluation by a full 10 points to 0.55 for white and recommends Bxc3 for black.

The tablebase shows the optimal move for black as Re1.

When shown ...Re1, Houdini flips out and goes to 99.9...which to my understanding means a forced mate for White.

So...what's going on here? It's a bit hard for me to even know what I'm asking, as I have so little experience with all of these things...but I'm scratching my head about something.

I triple-checked to make sure I had the initial position correct between the two sources, so assuming I did...

1.) How can Houdini and the Tablebase differ by more than 10 points (!), in other words a full Queen's worth of value, in their evaluation? I mean, if I'm not mistaken, Houdini just said "What are you doing you idiot, you hung your queen" and the Tablebase said "No actually, I won."

2.) Is this common?

3.) Is it known that the super-engines by themselves can be this inaccurate in the endgame?

4.) Is there a way to get a hint at the refutation of Houdini's preferred ...Bxc3 that the tablebase apparently sees but Houdini is totally unaware of?

*head explode*

image not working

Weird, I tried using turboimagehost. I replaced it with one uploaded to Flickr.

I figured I can try to acquire the 6-piece tablebase for the position that occurs after Be7...BxC3. Since it will be a 6-piece tablebase with white to move (and a subset of the 7-game base)...and Houdini believes it is almost even at that point..and I can see how Houdini's preferred moves are refuted.

Now...I just need to figure out how and where to acquire and view this tablebase...

In another thread I showed a decent number of 'simple' 6 and even 5 piece endings that Houdini butchers. Houdini, relative to perfect chess, is terrible - something a large number of people still fail to grasp. And it's not just endgames that it's terrible at - it's just endgames where we can trivially prove that it's terrible.

I think your computer might be a bit slow. I showed the position to my Houdini 1.5.

I am not sure if 1. Be7 is indeed the strongest move, at least my Houdini wants to play 1.Nb5 with mate in 18 at depth 29. However, if 1. Be7 is given, he goes back to your 0.55 evaluation for some seconds but very quickly is back to mate in 22 for White (around depth 23 or 24).

Which is also not true (as the tablebase says it's mate in 17 after 1..Bxc3 2. Rxc3. But finding the quickest mate is not what these engines are built for (in a normal setup), and the evaluation (clear white win) is always given. He just has to get to his calculating horizon.

http://www.k4it.de/index.php?topic=egtb&lang=en

Online 6-man tablebase.

You should look up the position after 1 Be7 Bxc3 2. Rxc3 in an endgame manual. This kind of position is nearly always won when the rook has to defend on the first rank. There were also 2 excellent videos on chess.com about this R+B vs R endgame.

I see. Is Rybka better?

This probably explains it. 1. Be7 is what the tablebase displays, and I was under the impression that the tablebase plays perfectly. Perhaps there's actually a faster mate after Be7 than Houdini finds...

Thanks to both, will check them out.

Nope, I mentioned Houdini in particular since I think its generally agreed to be the strongest program. Table bases show very clearly that computers aren't necessarily good at chess, rather humans are just particularly bad.

Chess is hard.

hey guys. I just solved chess

Cool. Who wins?

1 something is wrong with the installation of houdini and the tablebase on your pc

2 yes its common to get rubbish out if the programs are not set up properly

3 yes, a program may say +24 when actually its a mate in 45 :-)

4 From the starting position nb5 leads to a mate in about 16 and be7,bg5,bg3,bd1 lead to a mate in about 20


FEN: 8/8/8/4b3/7B/2N2K2/2R5/r4k2 w - - 1 1

Houdini_15a_x64:
4/14 00:00 1,223 1,223,000 +0.69 Rc2-f2+ Kf1-g1 Rf2-g2+ Kg1-h1 Nc3-b5
5/17 00:00 4,641 928,000 +0.59 Rc2-f2+ Kf1-g1 Nc3-e2+ Kg1-h1 Kf3-e4 Ra1-a5 Rf2-f1+ Kh1-g2 Rf1-g1+ Kg2-h3
5/17 00:00 5,966 662,000 +0.77 Nc3-d5 Be5-d4 Kf3-e4 Ra1-d1 Bh4-g3
6/17 00:00 7,745 595,000 +0.73 Nc3-d5 Be5-d4 Bh4-g3 Ra1-a3+ Kf3-g4 Ra3-a8
7/17 00:00 9,085 534,000 +0.68 Nc3-d5 Be5-d4 Bh4-g3 Ra1-a3+ Kf3-g4 Ra3-a5 Rc2-c1+ Kf1-e2
8/17+ 00:00 16,277 739,000 +0.75 Nc3-d5
8/17 00:00 17,198 661,000 +0.78 Nc3-d5 Be5-d4 Rc2-c4 Ra1-d1 Kf3-e4 Bd4-a7 Rc4-c3 Kf1-g2 Nd5-f4+ Kg2-f1
9/17 00:00 21,484 671,000 +0.78 Nc3-d5 Be5-d4 Rc2-c4 Ra1-d1 Kf3-e4 Bd4-a7 Rc4-c3 Kf1-g2 Nd5-f4+ Kg2-f1
9/23 00:00 82,466 1,555,000 +0.81 Nc3-e4 Ra1-a3+ Kf3-g4 Ra3-a4 Kg4-f5 Be5-g7 Bh4-f2 Ra4-a5+ Kf5-g4 Ra5-a4 Kg4-f5 Ra4-a5+
10/23 00:00 96,451 1,461,000 +0.84 Nc3-e4 Ra1-a3+ Kf3-g4 Ra3-a4 Kg4-f5 Be5-g7 Bh4-f2 Ra4-a5+ Bf2-c5 Ra5-a4 Rc2-f2+ Kf1-e1 Bc5-e3 Ra4-a5+ Kf5-g4
11/26- 00:00 159,178 1,749,000 +0.77 Nc3-e4 Ra1-a3+
11/26 00:00 203,215 1,814,000 +0.84 Nc3-d5 Be5-d4 Rc2-c4 Ra1-a3+ Kf3-e4 Bd4-a7 Bh4-f6 Ra3-a2 Nd5-f4 Kf1-f2 Bf6-h4+ Kf2-g1 Bh4-e1
11/26+ 00:00 265,403 1,995,000 +0.90 Nc3-b5
11/26+ 00:00 275,601 1,968,000 +1.04 Nc3-b5
11/26 00:00 333,468 2,084,000 +1.20 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-e2 Be5-h2 Re2-e6 Ra1-d1 Re6-g6 Rd1-d8 Bf2-c5 Rd8-d5
12/26 00:00 363,883 2,103,000 +1.15 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c4 Be5-h2 Rc4-h4 Ra1-a5 Rh4-h5 Ra5-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d4
13/26 00:00 422,324 2,154,000 +1.17 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c4 Be5-h2 Rc4-h4 Ra1-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d5 Re8-d8+ Nb5-d6
14/26+ 00:00 515,023 2,299,000 +1.26 Nc3-b5
14/26 00:00 532,847 2,277,000 +1.17 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c4 Be5-h2 Rc4-h4 Ra1-b1 Nb5-d4 Rb1-b7 Bf2-g3 Rb7-f7+ Kf3-e4 Rf7-e7+ Ke4-d3
15/26 00:00 703,073 2,502,000 +1.24 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c4 Be5-h2 Rc4-h4 Ra1-a5 Rh4-b4 Bh2-e5 Bf2-c5 Ra5-a1 Rb4-h4+ Be5-h2 Nb5-c3 Ra1-a5 Rh4-h5
16/28+ 00:00 920,780 2,700,000 +1.33 Nc3-b5
16/28+ 00:00 1,033,157 2,769,000 +1.50 Nc3-b5
16/36+ 00:00 1,207,714 2,896,000 +1.91 Nc3-b5
16/37 00:01 1,956,106 2,661,000 +1.27 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c4 Ra1-d1 Rc4-c5 Be5-h2 Nb5-d4 Rd1-d3+ Bf2-e3 Rd3-d1 Rc5-g5 Rd1-f1+ Be3-f2 Rf1-d1 Nd4-e6 Rd1-a1 Bf2-d4 Ra1-a3+ Kf3-e4 Bh2-d6 Rg5-g1+ Kh1-h2
17/37+ 00:01 2,275,012 2,791,000 +1.36 Nc3-b5
17/37+ 00:01 2,338,013 2,800,000 +1.54 Nc3-b5
17/43+ 00:01 3,520,345 3,115,000 +5.92 Nc3-b5
17/43 00:01 4,418,699 3,280,000 +6.33 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Nb5-d4 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Bh2-g1+ Ke3-e4 Kh1-g2 Rc5-c2+ Kg2-h3 Nd4-e6 Kh3-h4 Rc2-g2 Bg1-a7 Ne6-f4 Ba7-b8 Ke4-f3 Bb8xf4 Kf3xf4 Kh4-h5 Kf4-e5 Kh5-h4
18/43+ 00:02 6,231,563 3,473,000 +6.60 Nc3-b5
18/43 00:03 10,135,576 3,663,000 +6.88 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a3 Bf2-d4 Ra3-a8 Bd4-e5 Ra8-f8+ Kf3-e3 Rf8-f2 Nc3-e4 Rf2-a2 Ne4-d2 Ra2-a3+ Ke3-f2 Ra3-h3 Rh5-g5 Rh3-h4 Nd2-f3 Rh4-a4 Be5xh2 Ra4-g4 Rg5-h5 Rg4-g5 Rh5-h7 Rg5-g2+ Kf2-e3
19/43 00:03 11,308,726 3,690,000 +6.84 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a3 Bf2-d4 Ra3-a8 Bd4-e5 Ra8-f8+ Kf3-e3 Rf8-f2 Nc3-e4 Rf2-a2 Ne4-d2 Ra2-a3+ Ke3-f2 Ra3-h3 Rh5-g5 Rh3-h4 Nd2-f3 Rh4-a4 Be5xh2 Ra4-g4 Rg5-h5 Rg4-g5 Rh5-h6 Rg5-g6 Rh6-h7
20/46 00:04 16,252,955 3,831,000 +6.92 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d4 Kh1-g2 Bg3xh2 Kg2-f3 Bh2-e5 Kf3-g4 Rh5-h1 Kg4-f3 Rh1-h3+ Kf3-g4 Rh3-g3+ Kg4-f5 Nc3-d5 Re8-a8 Nd5-e3+ Kf5-e6 Rg3-g6+ Ke6-f7 Rg6-f6+ Kf7-g8 Kd4-e4
21/46 00:05 21,954,164 3,900,000 +6.94 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d4 Kh1-g2 Bg3xh2 Kg2-f3 Bh2-e5 Re8-d8+ Nc3-d5 Kf3-g4 Rh5-h7 Kg4-f3 Rh7-f7+ Kf3-e2 Kd4-e4 Rd8-e8 Nd5-c7 Re8-g8 Nc7-e6 Rg8-e8 Ke4-d5 Re8-a8
22/46 00:06 25,417,343 3,952,000 +6.94 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d4 Kh1-g2 Bg3xh2 Kg2-f3 Bh2-e5 Re8-d8+ Nc3-d5 Kf3-g4 Rh5-h7 Kg4-f3 Rh7-f7+ Kf3-e2 Kd4-e4 Rd8-e8 Rf7-h7 Ke2-f2 Rh7-h1 Kf2-g2 Rh1-h5 Kg2-g1 Rh5-h3 Kg1-f2 Rh3-b3 Kf2-e2 Rb3-c3 Ke2-d2 Rc3-b3 Kd2-e2 Rb3-c3
23/46 00:08 31,237,091 4,048,000 +6.94 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-b8 Rc5-h5+ Bb8-h2 Nb5-c3 Ra1-a8 Bf2-g3 Ra8-f8+ Kf3-e4 Rf8-e8+ Ke4-d4 Kh1-g2 Bg3xh2 Kg2-f3 Bh2-e5 Re8-d8+ Nc3-d5 Kf3-g4 Rh5-h7 Kg4-f3 Rh7-f7+ Kf3-e2 Kd4-e4 Rd8-e8 Rf7-h7 Ke2-f2 Rh7-h1 Kf2-g2 Rh1-h5 Kg2-g1 Rh5-h3 Kg1-f2 Rh3-b3 Kf2-e2 Rb3-c3 Ke2-d2 Rc3-b3 Kd2-e2 Rb3-c3
24/49+ 00:16 71,264,990 4,376,000 +7.23 Nc3-b5
24/57+ 00:20 86,063,808 4,365,000 +7.81 Nc3-b5
24/69+ 00:36 162,057,080 4,524,000 +14.13 Nc3-b5
24/69 00:52 287,138,447 5,500,000 +M25 Nc3-b5 Be5-h8 Bh4-f2 Bh8-f6 Bf2-e3 Kf1-e1 Nb5-a3 Bf6-d4 Na3-c4 Ke1-d1 Rc2-d2+ Kd1-c1 Be3xd4 Ra1-a4 Bd4-b2+ Kc1-b1 Nc4-e3 Ra4-a7 Ne3-d5 Ra7-f7+ Kf3-e4 Kb1-a2 Bb2-c1+ Ka2-b3 Rd2-b2+ Kb3-a4 Bc1-d2 Ka4-a3 Rb2-b1 Ka3-a4 Rb1-b4+ Ka4-a5 Rb4-b7+ Ka5-a4 Rb7xf7 Ka4-b5 Rf7-a7 Kb5-c4 Ra7-b7 Kc4-c5 Rb7-c7+ Kc5-b5 Rc7-c2 Kb5-a4 Rc2-b2 Ka4-a3 Rb2-b5 Ka3-a2 Nd5-c3+ Ka2-a3 Bd2-c1+
25/69 00:56 318,337,121 5,648,000 +M25 Nc3-b5 Be5-h8 Bh4-f2 Bh8-f6 Bf2-e3 Kf1-e1 Nb5-a3 Bf6-d4 Na3-c4 Ke1-d1 Rc2-d2+ Kd1-c1 Be3xd4 Ra1-a4 Bd4-b2+ Kc1-b1 Nc4-e3 Ra4-a7 Ne3-d5 Ra7-f7+ Kf3-e4 Kb1-a2 Bb2-c1+ Ka2-b3 Rd2-b2+ Kb3-a4 Bc1-d2 Ka4-a3 Rb2-b1 Ka3-a4 Rb1-b4+ Ka4-a5 Rb4-b7+ Ka5-a4 Rb7xf7 Ka4-b5 Rf7-b7+ Kb5-c6 Rb7-c7+ Kc6-b5 Rc7-c2 Kb5-a4 Rc2-b2 Ka4-a3 Rb2-b5 Ka3-a2 Nd5-c3+ Ka2-a3 Bd2-c1+
26/69 01:00 345,142,683 5,747,000 +M25 Nc3-b5 Be5-h8 Bh4-f2 Bh8-f6 Bf2-e3 Kf1-e1 Nb5-a3 Bf6-d4 Na3-c4 Ke1-d1 Rc2-d2+ Kd1-c1 Be3xd4 Ra1-a4 Bd4-b2+ Kc1-b1 Nc4-e3 Ra4-a7 Ne3-d5 Ra7-f7+ Kf3-e4 Kb1-a2 Bb2-c1+ Ka2-b3 Rd2-b2+ Kb3-a4 Bc1-d2 Ka4-a3 Rb2-b1 Ka3-a4 Rb1-b4+ Ka4-a5 Rb4-b7+ Ka5-a4 Rb7xf7 Ka4-b5 Rf7-b7+ Kb5-c6 Rb7-c7+ Kc6-b5 Rc7-c2 Kb5-a4 Rc2-b2 Ka4-a3 Rb2-b5 Ka3-a2 Nd5-c3+ Ka2-a3 Bd2-c1+
27/69 01:08 402,176,515 5,940,000 +M25 Nc3-b5 Be5-h8 Bh4-f2 Bh8-f6 Bf2-e3 Kf1-e1 Nb5-a3 Bf6-d4 Na3-c4 Ke1-d1 Rc2-d2+ Kd1-c1 Be3xd4 Ra1-a4 Bd4-b2+ Kc1-b1 Nc4-e3 Ra4-a7 Ne3-d5 Ra7-f7+ Kf3-e4 Kb1-a2 Bb2-c1+ Ka2-b3 Rd2-b2+ Kb3-a4 Bc1-d2 Ka4-a3 Rb2-b1 Ka3-a4 Rb1-b4+ Ka4-a5 Rb4-b7+ Ka5-a4 Rb7xf7 Ka4-b5 Rf7-b7+ Kb5-c6 Rb7-c7+ Kc6-b5 Rc7-c2 Kb5-a4 Rc2-b2 Ka4-a3 Rb2-b5 Ka3-a2 Nd5-c3+ Ka2-a3 Bd2-c1+
28/69 01:19 488,219,176 6,130,000 +M25 Nc3-b5 Be5-h8 Bh4-f2 Bh8-f6 Bf2-e3 Kf1-e1 Nb5-a3 Bf6-d4 Na3-c4 Ke1-d1 Rc2-d2+ Kd1-c1 Be3xd4 Ra1-a4 Bd4-b2+ Kc1-b1 Nc4-e3 Ra4-a7 Ne3-d5 Ra7-f7+ Kf3-e4 Kb1-a2 Bb2-c1+ Ka2-b3 Rd2-b2+ Kb3-a4 Bc1-d2 Ka4-a3 Rb2-b1 Ka3-a4 Rb1-b4+ Ka4-a5 Rb4-b7+ Ka5-a4 Rb7xf7 Ka4-b5 Rf7-b7+ Kb5-c6 Rb7-c7+ Kc6-b5 Rc7-c2 Kb5-a4 Rc2-b2 Ka4-a3 Rb2-b5 Ka3-a2 Nd5-c3+ Ka2-a3 Bd2-c1+
29/69 02:46 1,112,547,583 6,683,000 +M18 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Nb5-d4 Ra1-a3+ Bf2-e3 Ra3-a1 Nd4-e2 Ra1-f1+ Be3-f2 Rf1-a1 Rc5-h5 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Rh5-h4 Bh2-e5 Rh4-g4+ Kg2-h3 Ke3-f3 Be5-h2 Ne2-f4+ Bh2xf4 Rg4xf4 Kh3-h2 Rf4-h4+ Kh2-g1 Rh4-h5 Kg1-f1 Rh5-h1+
30/69 03:00 1,210,441,859 6,720,000 +M18 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Nb5-d4 Ra1-a3+ Bf2-e3 Ra3-a1 Nd4-e2 Ra1-f1+ Be3-f2 Rf1-a1 Rc5-h5 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Ne2-f4+ Kg2-g3 Rh5-g5+ Kg3-h4 Rg5-g6 Bh2-g3 Ke3-f3 Bg3xf4 Kf3xf4 Kh4-h3 Rg6-g5 Kh3-h2 Kf4-f3 Kh2-h1 Rg5-h5+ Kh1-g1 Rh5-h4 Kg1-f1 Rh4-h1+
31/69 03:43 1,522,422,827 6,815,000 +M18 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Nb5-d4 Ra1-a3+ Bf2-e3 Ra3-a1 Nd4-e2 Ra1-f1+ Be3-f2 Rf1-a1 Rc5-h5 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Rh5-h4 Bh2-e5 Rh4-g4+ Kg2-h3 Ke3-f3 Be5-h2 Ne2-f4+ Bh2xf4 Rg4xf4 Kh3-h2 Rf4-h4+ Kh2-g1 Rh4-h5 Kg1-f1 Rh5-h1+
32/69 05:29 2,285,337,609 6,934,000 +M16 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Rc5-d5 Bh2-g1 Rd5-h5+ Bg1-h2 Nb5-d6 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Rh5-g5+ Kg2-h3 Nd6-e4 Kh3-h4 Ke3-f3 Bh2-d6 Rg5-f5 Bd6-e7 Ne4-g3 Be7-g5 Rf5-e5 Kh4-h3 Re5-e1 Bg5-d8 Re1-h1+
33/69 07:10 3,025,462,769 7,034,000 +M16 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Rc5-d5 Bh2-g1 Rd5-h5+ Bg1-h2 Nb5-d6 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Rh5-g5+ Kg2-h3 Nd6-e4 Kh3-h4 Ke3-f3 Bh2-d6 Rg5-g4+ Kh4-h5 Ne4xd6 Kh5-h6 Nd6-e4 Kh6-h7 Kf3-f4 Kh7-h6 Kf4-f5 Kh6-h7 Ne4-f6+ Kh7-h6 Rg4-g6+
34/69 10:45 4,638,997,140 7,191,000 +M16 Nc3-b5 Kf1-g1 Bh4-f2+ Kg1-h1 Rc2-c5 Be5-h2 Rc5-d5 Bh2-g1 Rd5-h5+ Bg1-h2 Nb5-d6 Ra1-a3+ Bf2-e3 Ra3xe3+ Kf3xe3 Kh1-g2 Rh5-g5+ Kg2-h3 Nd6-e4 Kh3-h4 Ke3-f3 Bh2-d6 Rg5-d5 Bd6-e7 Ne4-g3 Be7-g5 Rd5-e5 Kh4-h3 Re5-e1 Bg5-d8 Re1-h1+

so is the youtube video just a fraud? seems weird to do that.

1) Houdini 1.5a is over a year and a half old, I'm sure they've released an update.
2) All computer chess algorithms have edge cases where their evaluation is not quite accurate because of the way they weight positional factors versus material factors when they can't see a forced mate past their move horizon. Personally I would be cautious about playing Nb5 and putting my Knight on a square where it can't move.
3) Tablebases are not algorithms, they are solved positions, thus they can't be wrong.