Author: Don Dailey
Date: 14:56:48 01/18/98
Go up one level in this thread
I discovered 126 "flaky" problems after correcting a bug in my flake test routine. I recommend these all be taken out because they imply ambiguous solutions. The ones you guy's find also should be taken out. Bob mentioned that he wanted to throw out multiple solutions. I would prefer this too but will go along with any other decision. Here are my observations on your problems: On January 18, 1998 at 16:41:30, Amir Ban wrote: > >I did some work on the ECM suite (please don't call it CCC, the original >text still deserves credit for it). I looked at the first 100. These >are somewhat easier than the rest. In under 20 seconds, I get 81 >correct. I am assuming (without good reason I admit) that since I solved >them these are correct, and I looked at the other 19. I ran them for >some more time, about a minute or so, and also visually analyzed them to >see if I can see what the problem is with them. Here is what I find. >Feel free to go over my analysis, and point out errors: > >No. 11 (Nxg4): > >Junior immediately grabs Bb5 and sticks with it. I analyzed Nxg4 and >there is certainly some point to it. These are two completely different >plans, and although Bb5 seems better, I'm not sure. > >Recommend: Add Bb5 as a side solution. Cilkchess also wants Bb5 but goes back and forth between this and Rhc8 Recommend: Throw out on the grounds of flakiness. > > >No. 14 (Nd5): > >Junior finds Nb5 immediately. In the deeper analysis, it ultimately >prefers Nd5 slightly. Both moves have the same tactical idea. > >Recommend: No change (but would like to hear other opinions). Cilkchess likes Nd5 but only after a long search. There is no big score jump so the problem may be "flaky" Recommend: throw it out! >No. 17 (Rxd5): > >Junior finds Nf6+. After a long think, it prefers Rxd5 slightly. > >Recommend: No change, or add Nf6+ as side solution. Cilkchess likes Rxd5 from 3rd iteration on. The score gradually climbs on each iteration. I didn't look at this one but on the 5th ply on I see Nf6+ in all the PV's as 3rd move. It is likely this is a multiple solution. Recommend : Throw out on grounds of multiple solution. >No. 19 (Rxf6+): > >Rxf6+ simplifies nicely into a won endgame. However Junior never >considers it and find Nxe6 is really crushing. > >Recommend: Change key to Nxe6. Cilkchess likes Rxf6 until the 6th ply but both moves look completely crushing. But Rxf6 is "found" on ply 1 so it's too easy or ambiguious. Recommend: too easy, throw it out. >No. 28 (Rxf4): > >A difficult problem, and somewhat vague. I am not close to finding it, >but by following the sequence Rxf4 Bxf4 Nxf6+ gxf6 Qg4+ Bg5 Bxg5 f5 it >looks like it does work in the sense that it's better than the >alternatives, but I cannot see a plus score. > >Recommend: No change, or else drop it for reason of vagueness. Cilkchess does not find in 1 minute. Recommend: Let's look closer at it later in case it's a great problem! >No. 29 (Rgxg6): > >A classical problem which is found easily. The idea is that after the >knight on g6 is gone, Qe5 wins. However, after a few seconds Junior >prefers an intermediate Rxa6 Qxa6 Rxg6 which is the same plan but more >effective. > >Recommend: Add Rxa6 as side solution. The problem is that playing Rxa6 >doesn't prove that your program understood this position. Cilkchess see's big score jump with Rgxg6 but later changes to Rxa6. Throw out for flakiness if we are not allowing side solutions. >No. 43 (Bxg2+): > >Junior takes Qxc2 fast and sticks with it, thinking that white is forced >to give up the exchange with Rxg4. Looks like this works, and is >somewhat better than the key. > >Recommend: Add Qxc2 as side solution. Cilkchess changes back and forth many times. Recommend: Throw out unless we are allowing multiple solutions. >No. 51 (Rxh2): > >Was not solved in 20 seconds, but turned out in the deeper think, but >with an unconvincing score. I didn't analyze it. > >Recommend: No change (but someone else should look at it). Cilkchess seems to think something is going on with Qc5 which is not the given solution. I haven't look at it yet though. >No. 60 (Bxh2+): > >This one looks wrong. Junior has Qh6 which seems winning (Nf3 Nd2!). >Looks much more convincing than Bxh2+, where I don't see a win. > >Recommend: Change solution to Qh6. Cilkchess agree's. >No. 63 (Ncxd5): > >Junior has Nbxd5, and this looks like a simple transpostion. > >Recommend: Add Nbxd5 as side solution. Cilkchess likes Ncxd5 but agree it's a transposition. But either way it looks too easy. Recommend: Throw it out. >No. 68 (Bd3): > >I don't understand the key move. Bd3 Qxc3 Rg4 Qxa1 Bxa1 Nxg4 Qxg4 g6 >looks like just a creative way to lose material. I don't see anything >else here. > >Recommend: Drop this problem. Cilkchess tends to agree. I'm deep-searching it now after Bd3 to see what happens. >No. 71 (Qh6): > >Qh6 leads to material gain, and Junior finds it quickly, but after a few >seconds insists on the quiet and subtle Kg2! Black loses a rook at >least. > >Recommend: Change solution to Kg2. Cilkchess goes back and forth between the two and at the end seems to agree that Kg2 is better. However Kg2 is found on 2 ply (but not understanding) so I view this one as ambiguious. Recommend: Drop this one. >No. 76 (fxg6): > >Junior switches to fxe6 and sticks with it. Analyzing both options, the >key is correct but fxe6 is at least as good. > >Recommend: Add fxe6 as side solution. Recommend: Too easy, throw it out. >No. 78 (Nxc5): > >Was not found in 20 seconds, but turns up convincingly later. > >Recommend: No change. Cilkchess takes 32.9 seconds to find key move convincingly. Recommend: Keep this one. >No. 81 (Qb1+): > >Junior takes Bxc6, with idea Bxc6 Qd8 Qb1+. The idea of the key Qb1+ is >that the white queen is lost on a8 because of mate threats. Junior's >side solution uses the same idea, showing that the queen is equally lost >on d8 ! A nice cook. > >Recommend: Add Bxc6 as side solution. Cilkchess concurs. >No. 82 (exf6): > >Not found in 20 seconds, but turns up later. > >Recommend: No change. Cilkchess finds it in 8.9 Recommend: Keep > > >No. 84 (Bg6): > >Junior plays the unspectacular cxb5. The key Bg6 looks sound, but it >only forces the exchange of heavy pieces and leads to an endgame with >advantage. > >Recommend: No change, or drop it on grounds of vagueness. We concur. Cilkchess flakes back and forth and the score does not indicate it understands anything. >No. 91 (Nc6): > >Junior switches between Nc6 and Nxe6. It finally adopts Nc6, but Nxe6 >may not be far behind. > >Recommend: No change. This one is flaky, we oscillate between 3 moves including Juniors. Recommend: kill it. >No. 96 (Qe8+): > >This one's correct and is mate in 5, but so is Nf6+ with a different >move order. > >Recommend: Add Nf6+ as side solution. Cilkchess choose Nf6 also and announces Mate in 5. > >Amir
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.