[Computer-go] John got tromped by Zen 1-3

[Don Dailey]

On Tue, Jan 17, 2012 at 6:35 PM, Michael Williams <
michaelwilliams75 at gmail.com> wrote:

> I think the computers will improve more than John during the course of the
> next year.  Progress has been rapid recently.
>

I agree with that,  but saying that is not proof that it will be better
than John in 1 year since we don't really know how strong it is compared to
John based on 3 games.

I typical run 50,000 games matches to "prove" a single change in my world
class chess program and I can tell you that 3 games is a complete JOKE in
the context of evaluating a player (or the relative strength of 2 players.)


I'm not saying it was not useful or that it was not fun,  it was.

I'm not even saying that we should not replace John,  Zen's fine result
with the 5 dan performance with fast games gives us some a priori reason to
get a stronger player for the next time.     However I think based on the
time control differences that Zen is not very close to 5 dan at this level
and it's bit unsatisfying to draw strong conclusions based on 3 games.

Don




>
>
>
> On Tue, Jan 17, 2012 at 3:56 PM, Don Dailey <dailey.don at gmail.com> wrote:
>
>> I know that everyone at this point thinks John is all washed up for this
>> kind of match,   but I don't agree.
>>
>> Basically we have concluded based on a mere 3 games that John Tromp was
>> "crushed" by the computer as if there was no hope had the match continued.
>>    If the first 2 games had been a loss but John had won the last game,
>> we probably would be viewing it slightly more rationally - as John having a
>> chance.
>>
>> It's probably not very practical for logistical reasons,  but a match
>> like this should really played with a lot more than 3 games.    This was
>> fun,  but only proved what we already knew and nothing more - that
>> computers are now dan players.
>>
>> A ten game match also would not be enough,  but it would certainly be far
>> better for a more educated guess about reasonable expectations.    Would it
>> have been 9 to 1?     Would it have been a close match?
>>
>> Imagine that both John and the computer were perfect evenly matched with
>> each having a 50/50 change of winning each game.     The likely score would
>> have been 2-1 with a 3-0 quite possible too,  despite perfect equality of
>> the players and yet we are acting almost as if this was clear proof that
>> John is washed up and can no longer compete.       Had John won 2-1 how
>> would we be talking about this?     Probably we would be saying that
>> computers still have a long way to go while criticizing or laughing at some
>> of it's "gaffes."
>>
>> So seriously,  let's not make more of this than it actually is.
>>
>> Don
>>
>>
>> On Tue, Jan 17, 2012 at 12:20 AM, Michael Williams <
>> michaelwilliams75 at gmail.com> wrote:
>>
>>> John, thank you for doing this!  But with all due resepect, you have
>>> been eclipsed and we need a new challenger.  For next year (or whenever) I
>>> hope that person is as reasonable and sane as John has been.  Not everyone
>>> is cut out for the quirks of the man-machine matches.
>>>
>>>
>>>
>>> On Tue, Jan 17, 2012 at 12:02 AM, Darren Cook <darren at dcook.org> wrote:
>>>
>>>> > (i) MC winning percentages (by Many Faces of Go)
>>>> > http://www.althofer.de/zen-tromp-04-percents.jpg
>>>> >
>>>> > (ii) Traditional territory scores (by Many Faces of Go)
>>>> > http://www.althofer.de/zen-tromp-02-territory.jpg
>>>>
>>>> I think that 2nd link should have been:
>>>>   http://www.althofer.de/zen-tromp-04-territory.jpg
>>>>
>>>> (but, in his defense, Ingo was posting at 5am :-)
>>>>
>>>> Darren
>>>> http://dcook.org/gobet/
>>>>
>>>> --
>>>> Darren Cook, Software Researcher/Developer
>>>>
>>>> http://dcook.org/work/ (About me and my work)
>>>> http://dcook.org/blogs.html (My blogs and articles)
>>>> _______________________________________________
>>>> Computer-go mailing list
>>>> Computer-go at dvandva.org
>>>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>>>>
>>>
>>>
>>> _______________________________________________
>>> Computer-go mailing list
>>> Computer-go at dvandva.org
>>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>>>
>>
>>
>> _______________________________________________
>> Computer-go mailing list
>> Computer-go at dvandva.org
>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>>
>
>
> _______________________________________________
> Computer-go mailing list
> Computer-go at dvandva.org
> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20120117/33dd7d41/attachment.html>