Author: Chris Whittington
Date: 11:28:27 09/27/97
Hmmm, this seems an interesting conspiracy theory ....... Amir ban wrote: >Robert Hyatt wrote: > > When they publish details, we'll find out. By the way, for Rolf, > who is always nagging about why they have published something, here > is an important question: > > "Have *you* ever published something in a significant scientific > technical journal? (of course not) Do you have *any* idea of > the lead time from writing an article until it is published in > such a journal? (of course not) Here is a clue: the typical > lag from submission to publication is over one year. One year > is considered very quick turnaround, while many journals operate > on a 2-year delay. So back off and give them time." In view of the huge amount of commentary, analysis and whatnot by anyone reomotely close to this match during the match and after it, even you Bob m ay admit that this yours is a weak statement. What kind of preparation and lead time is needed to give some meaningful comments on what everyone is asking ? If you couple this with IBM's (especially Mr. Tan's) statements of say-nothing gibberish, and IBM's very obvious policy of keep-your-mouth-shut before, and amazingly, after the match, you will understand that it is NOT the case that some annoyingly slow editors of some obsocure journals are keeping the story from the world while the the Deep-Blue team is biting its nails in frustration at the delay. Please remember we are not waiting to hear details about the SP architecture, or IBM's marketing hype on molecular biology. We are waiting to hear something on computer chess and the match. The world has noticed, and IBM are well aware, that Kasparov has raised the suspicion of cheating, with the 2nd game in mind. Rather than discuss and reveal everything they have on this remarkable game, they chose to deliver a few pages of cryptic dumps marked "IBM confidential" to Kasparov and his aides eyes only. Since Kasparov asked us to decipher those, I have seen them (I understand something was also published on the NY Times. I don't know if everything or only part of it). The interesting part of it were the printouts of the 36'th and 37'th moves of game 2 (axb5 and Be4). Garry's cheating theory was that there was a "go-for-second-best" button, and that this may have been used at the 36th move. This is quite possible to implement of course, and may be quite effective. In this regard Kasparov said something that is very significant I think: "If an expert is allowed to override the computer just once in any game, I am already in trouble". This still doesn't mean that this kind of cheating took place. I didn't find any evidence for this, but I cannot rule it out in view of what the printouts tell. The printouts show that Qb6 was the move considered all the time. DB did well to see Black's best counterplay after Qb6, throwing 2 or 3 pawns to open lines and attack the white king, as shown by the PV. The evaluation throughout was around +0.5 pawns for white and dropping slowly. Finally axb5 was accepted without a PV with a value of 0.4-0.5. There were several things the needed explanation in the deciphering: - There was no explanation why the Qb6 score dropped so low with white ahead by 2 pawns or more. I mean, it's not like you could see the material regained or a decisive king attack mounted. The only explanation was that the evaluation function gave the compensation, in contrast to all PC programs who evaluate white to be more than a pawn up there. So it seems that DB is not so much the calculating giant as the positional wizard ? - No PV for axb5 (but that can happen of course). - Time management mystery: I think I could make out DB's time management policy, but could not understand what happened in this move. The normal time allotment was only about 3 minutes, but a move was not played by that time, although in other moves, DB would always play a move at the planned time. There was no obvious reason not to play the move at the normal time, which was still Qb6. Instead DB went into a procedure it called "panic-time", where there started a countdown to about 14 minutes. Several minutes into "panic-time" DB switched to axb5. Some time later, not connected to any event shown in the log, with the panic-time countdown still with 7 or 8 minutes to go, DB announced "panic-time" and played a move. DB still had other moves to consider at that ply level. So what happened ? In the previous move, the 35th (Bxd6) DB thought for 15 minutes. It seems cleat that the "panic-time" mechanism was activated there (for some unknown reason), and the machine thought until "panic-time" interrupted it. This is significant I think. Was DB malfunctioning at this point of the game. Could it be that an operator, seeing the depressing "panic-time" messages appearing for the 2nd move in a row really panic and push the "Move Now" button ? There is no reason to rush into any conclusions here, but the DB team needs to give answers to the these questions: 1. What was the basis for preferring 36. axb5 over 36. Qb6 ? 2. Why was "panic-time" activated in the 36th move and why was a move played when it finally played ? 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.