Author: Uri Blass
Date: 02:04:41 01/13/01
Go up one level in this thread
On January 13, 2001 at 03:21:06, Dann Corbit wrote: >On January 13, 2001 at 03:03:38, Uri Blass wrote: > >>On January 13, 2001 at 01:58:00, Dann Corbit wrote: >> >>>On January 13, 2001 at 01:39:49, Robin Smith wrote: >>>[snip] >>>>>All that having been said, they are very likely GM's. But it will be proven >>>>>when it has been proven. Right now it isn't. >>>> >>>>"Proven when it has been proven", that is an intersting statement. It makes it >>>>all sound so definate, precise and conclusive. But if it is so precise, what is >>>>the definition of proven? The whole thing is actually quite silly, because it >>>>takes something that is inheritantly probabilistic and tries to make it black or >>>>white .... proven or unproven. Mathematically this isn't so simple as you seem >>>>to imply. >>> >>>With the current number of games played, the error bar is hundreds of ELO wide, >>>and the center is barely on GM level. If this were the level of certainty used >>>to stop elevators or control heart machines, there would be dead people lying >>>all over. My point is that the experimental evidence does not point to a sound >>>decision. If someone tried to prove a hypothesis in a scientific journal with >>>data that shaky they would be laughed out of town. Actually, it would never >>>make it past peer review and get published. >>> >>>This is what is simple: >>>The current data does not point to a reliable conclusion. >>>With more data a reliable conclusion could be reached. >>>The hypothesis cannot be concluded on the basis of the data at hand. >>> >>>All that having been said, the hypothesis is probably correct. But the current >>>evidence is inadequate to say that it is proven. >>> >>>Well, sure, we won't ever have 100% reliable answers. But we can have *GOOD* >>>answers. We don't have that right now -- not by a longshot. >> >>I do not like all this mathematical way to prove if someone is at GM strength >>because it assume a simple model that does not exist and it ignores data. >> >>If I see a beginner's game against a GM I can by looking at the moves say that I >>am convinced that this player is not a GM because a GM even in the worst day is >>not going to do a lot of stupid tactical mistakes. >> >>Your model only see 1-0 result and will say that we need more games to prove >>that the player is not at GM strength. >> >>People can have their opinion that they are convinced that a player is or is not >>GM strength even if there is no proof by a mathematical model because the >>mathematial model ignores a lot of data. > >Then propose a better model. I'm all for it. But to suggest that we can have a >better proof by saying "Looks like a GM to me." is not something that I accept. > >Now, I want to differentiate something. There is a difference between >acceptance and proof. To say that "I accept that computers are GM's because of >the prevailing evidence." is perfectly logical and sound. To say that >"computers are proven to be GM's." is not. I can believe about something with 95% confidence even if the mathematical model does not support it. The mathematical model also does not give probability that the player is of GM strength. It only gives probability to be wrong by saying that a player is not of GM strength when the player is a GM. A confidence of 95% does not mean that in 95% of the cases that we decide that a player is of GM strength we are right. It only means that in 95% of the cases when we decide about players who are not at GM strength we are right to reject the conjecture that they are GM. If there are no players with GM strength we have 100% probability to be wrong when we decide that a player is of GM strength. Uri
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.