Author: Steven Edwards
Date: 06:35:37 12/01/03
Go up one level in this thread
Some comments to various replies: 1. The proposed fingerprinting scheme is not intended to detect cloning of code that is not immediately related to move selection (why bother?), so criticisms on that basis are irrelevent. 2. If the test suite is run through the same interface that is used to play games in the event, then intentional faking on the suite is going to be difficult or impossible to perform. 3. If there is doubt about faking the test suite positions, then an additional test suite can be made from games actually played it the event. A difference here will indicate probabe faking on the initial test suite or certain faking on the secondary test. 4. The test EPD output will include a PV several moves long for each position. The probability of non-clone matching declines as the ply increases; the matching of long chucks of the PV for many different positions would be a risk too high for a cloner to take. ---------- Writing a similarity tester using EPD input should be a simple task of mulivariate analysis correlation. Perhaps I'll do this in an upcoming article in the _NACCA Journal_.
This page took 0 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.