Mike Murray wrote:
"Jerzy" wrote:
Do you think that there is an exact formula for giving e.g. time
odds ? ;-)

Never heard of one, but here's what pops into my head: Split the
total time available by the inverse of the odds predicted by the
difference in Elo.

This doesn't work. The Elo rating was calculated from games at
classical time controls and it is only predictive for those games.
Knowing the two players' Elo ratings allows you to estimate the
probability of A beating B at classical time controls. Suppose A
usually beats B in long games. Nonetheless, it might be that A gets
very nervous when playing blitz so does very badly, while B is a blitz
specialist who doesn't do all that well at classical time controls.
But your scheme will give less time to A, the weaker blitz player.

(Or A might tend to use all of his two hours, while B tends to bash
out his moves quickly.)

It's also not clear to me that your time division is the right way of
doing it but that doesn't matter so much as the whole idea is, I'm
afraid, flawed.

Making some sort of time division based on Elo ratings from rapid
games would be a better approximation. It would still suffer from the
problem I described but, perhaps, to a lesser extent: for example, A
might be OK at 5-minute chess but really hate 2-minute chess, which is
what he might end up being forced to play.


Dave.

--
David Richerby Flammable Boss (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ middle manager but it burns really
easily!

On 23 Mar 2007 13:25:49 +0000 (GMT), David Richerby
wrote:

Mike Murray wrote:
"Jerzy" wrote:
Do you think that there is an exact formula for giving e.g. time
odds ? ;-)

Never heard of one, but here's what pops into my head: Split the
total time available by the inverse of the odds predicted by the
difference in Elo.

This doesn't work. The Elo rating was calculated from games at
classical time controls and it is only predictive for those games.
Knowing the two players' Elo ratings allows you to estimate the
probability of A beating B at classical time controls.

Elsewhere in the thread, we discussed the correlation between blitz
ratings, online blitz ratings, and ratings derived from classical time
controls, and the consensus was they usually matched pretty closely.
Of course, there are bound to be exceptions. We probably all know at
least one coffee-house hustler with an 1800 OTB rating who can beat
damn near everybody at one or two minute chess.

Suppose A
usually beats B in long games. Nonetheless, it might be that A gets
very nervous when playing blitz so does very badly, while B is a blitz
specialist who doesn't do all that well at classical time controls.
But your scheme will give less time to A, the weaker blitz player.

(Or A might tend to use all of his two hours, while B tends to bash
out his moves quickly.)

It's also not clear to me that your time division is the right way of
doing it but that doesn't matter so much as the whole idea is, I'm
afraid, flawed.

Making some sort of time division based on Elo ratings from rapid
games would be a better approximation.

Sure. Use 'em if you got 'em. We mentioned the old WCBA ratings.

It would still suffer from the
problem I described but, perhaps, to a lesser extent: for example, A
might be OK at 5-minute chess but really hate 2-minute chess, which is
what he might end up being forced to play.

Any form of odds runs the risk that the odds-giver will be
uncomfortable with them. In a one-on-one session, of course, the
players will tend to adjust the odds to reflect this, either during
the session or at the next one. My personal preference is for odds
that tend toward equal results but leave the stronger player with a
slight edge.

I'm guessing from your objections that your paradigm is something like
an odds tournament. I suspect any automatic system for calculating
odds would risk anomalies similar to those you outline.



Dave.

"David Richerby" wrote in message
...
Mike Murray wrote:
"Jerzy" wrote:
Do you think that there is an exact formula for giving e.g. time
odds ? ;-)

Never heard of one, but here's what pops into my head: Split the
total time available by the inverse of the odds predicted by the
difference in Elo.

This doesn't work. The Elo rating was calculated from games at
classical time controls and it is only predictive for those games.


Evidence, please. It is a gross misconception that ratings are
administered with a view towards increasing their predictive value.
That is a miniscule factor at best. The reason that organizations
have different types of ratings is that, essentially, these organizations
are selling a product and have determined that the consumers do not
*want* their blitz games to influence their classical ratings. It
does *not* follow that blitz games do *not* predict classical results.
Of course they do, at least to a degree.

What if an organization decided to offer time-of-day ratings. Games
played in the morning were given one rating. Games played in the
evening were given a different rating. Still no predictive value?

Knowing the two players' Elo ratings allows you to estimate the
probability of A beating B at classical time controls. Suppose A
usually beats B in long games. Nonetheless, it might be that A gets
very nervous when playing blitz so does very badly, while B is a blitz
specialist who doesn't do all that well at classical time controls.
But your scheme will give less time to A, the weaker blitz player.

(Or A might tend to use all of his two hours, while B tends to bash
out his moves quickly.)

It's also not clear to me that your time division is the right way of
doing it but that doesn't matter so much as the whole idea is, I'm
afraid, flawed.

Making some sort of time division based on Elo ratings from rapid
games would be a better approximation. It would still suffer from the
problem I described but, perhaps, to a lesser extent: for example, A
might be OK at 5-minute chess but really hate 2-minute chess, which is
what he might end up being forced to play.


Dave.

--
David Richerby Flammable Boss (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ middle manager but it burns really
easily!

Mike Murray wrote:
David Richerby wrote:
This doesn't work. The Elo rating was calculated from games at
classical time controls and it is only predictive for those games.
Knowing the two players' Elo ratings allows you to estimate the
probability of A beating B at classical time controls.

Elsewhere in the thread, we discussed the correlation between blitz
ratings, online blitz ratings, and ratings derived from classical
time controls, and the consensus was they usually matched pretty
closely.

OK -- I've only been skimming most articles in this thread.


Any form of odds runs the risk that the odds-giver will be
uncomfortable with them. In a one-on-one session, of course, the
players will tend to adjust the odds to reflect this, either during
the session or at the next one. My personal preference is for odds
that tend toward equal results but leave the stronger player with a
slight edge.

I'm guessing from your objections that your paradigm is something
like an odds tournament. I suspect any automatic system for
calculating odds would risk anomalies similar to those you outline.

Not particularly. I'm just wondering if the suggested scheme would
really work any better than the two players just guessing and
adjusting. :-)


Dave.

--
David Richerby Enormous Strange Hi-Fi (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a music system but it's totally weird
and huge!

On 23 Mar, 01:16, "help bot" wrote:
On Mar 22, 7:54 pm, "Mark Houlsby"
wrote:

Well, it's all relative, Jerzy. A Capablanca's blitz moves might be
liable to be *relatively careless* by the standard of Capablanca, but
still much better than many of us can find in a correspondence game.

Of course, since JRCyG *died* I play better than he does, but just
barely, even so....

Skippy, you are ridiculous!

By what standard could you play better than the
dead Jose Capablanca? In the last fifty years, he
has not made a single bad move, nor lost even one
game. Moreover, his rating has not gone down a
point, so he still (retroactively) out-rates you and
always will.

The only standard whereby you beat JC now is
in activity; true enough, you have him beat here,
hands down. Congratulations on out playing a
dead man, Skip.

-- coach bot

Ummm.... that was kinda my point, Kennedy.

On 23 Mar 2007 16:10:07 +0000 (GMT), David Richerby
wrote:


I'm guessing from your objections that your paradigm is something
like an odds tournament. I suspect any automatic system for
calculating odds would risk anomalies similar to those you outline.

Not particularly. I'm just wondering if the suggested scheme would
really work any better than the two players just guessing and
adjusting. :-)

Probably not. Heh, heh.

On Mar 20, 7:46 am, "SBD" wrote:
On Mar 19, 9:13 am, "Inconnux" wrote:

Canada is in the middle of this right now. CFC has started to

AFAIK, the Rating Boon was a one-time adjustment.

increase ratings based on how many games a person has played.

CFC for decades had a stable rating system (more stable over time than
the USCF's, for example) which included participation points, that is
points added for each game played. The CFC backed away from
participation points, the rating system went askew, and as an
emergency measure they did the Rating Boon, which very roughly
approximates what the participation points used to do, but
retroactively.

The CFC seems to think that Canadian ratings should mirror
the USCF ratings and not FIDE ratings.

Anthropomorphizing the CFC? I'm not aware that the CFC has such a
policy. Occam's razor would prefer the explanation given with the
Rating Boon, which was that adult but not old players were getting
their ratings hammered (and some of them getting disheartened) against
quickly-improving but still low rated juniors. I never saw that
demonstrated statistically, but the anecdotal evidence was tempting.

As a rough approximation, for the past 30 years or so, CFC ratings
have been somewhere between USCF and FIDE. I am not aware that there
was ever a conscious policy to favour one or the other.

Ive heard many
people complain of this at the last couple of Tournaments
ive been to.

There has been much made here that Canadian ratings are somehow
deflated vis-a-vis USCF. But given some comparisons, I don't see how,
in fact looking at Pascal Charbonneau, Igor Zugic, Roman Pelts,
Jonathan Berry, and other high rated players, their ratings are the
same or lower in the USCF.

Aw, shucks. I'm not a high rated player like the other three
mentioned. But if you care to look, you'll find that my USCF rating
(2312) is higher than my CFC rating (2305), even after the CFC Rating
Boon. Not that I play in the US frequently enough to make my rating
say much about the current USCF rating methods.

My FQE rating is about 100 points higher than either. Doesn't say
anything about today's FQE system because I haven't played in Quebec
in a couple of decades.

The trouble with a lot of these discussions on r.g.c.* is that results
are quoted that are out of date; that the message is distorted as soon
as it gets "repeated"; and that measures are misunderstood. Ooops.
That's three. Three troubles. Nobody expected a SI.

One lower-rated player on both sides of the
border, Fred Kleist, is 1979 USCF and 2128 CFC.

He had a pretty amazing one-game-a-day 2003 Canadian Open. And he
doesn't play in Canada often. With cross-border
comparisons, there are very few who play significantly on both sides.
That's why anecdotal evidence does not mean much in CFC vs USCF rating
comparisons.

Of course, such
comparisons are dodgy with just a few players, and their activity
levels are important; along with a ton of other things; one assumes
that the CFC has done some sort of study to show exactly how or why
this assumption was made?

That might depend upon whether you think that hand-waving can be a
"study".

I'd like to see a current 2007 comparison between CFC, USCF, FQE and
FIDE ratings. When I did that in 2003, I found that the differences
were surprisingly small.

I don't see much, if any difference between
the two, my CFC Active rating was about the same as my USCF Quick
rating when I played in CFC-rated events.

So where is the evidence that Canadian ratings somehow need to be
inflated to be on a par with the US? Ken Sloan probably understands
ratings better than anyone here, given his long experience and study
of the Elo system, what does he have to say?

J.Lohnerhttp://www.chess.ca/memberinfo.asp?CFCN=144557
Chessworld.net "Inconnux"

--
Jonathan Berry

"Jonathan Berry" wrote in message
ups.com...

Canada is in the middle of this right now. CFC has started to

AFAIK, the Rating Boon was a one-time adjustment.

Yes Ive posted the official reason for ratings increases
that was written in the latest Chess Canada magazine.

increase ratings based on how many games a person has played.

CFC for decades had a stable rating system (more stable over time than
the USCF's, for example) which included participation points, that is
points added for each game played. The CFC backed away from
participation points, the rating system went askew, and as an
emergency measure they did the Rating Boon, which very roughly
approximates what the participation points used to do, but
retroactively.


Was not aware of this, but Im pretty new to playing rated chess.

The CFC seems to think that Canadian ratings should mirror
the USCF ratings and not FIDE ratings.

Anthropomorphizing the CFC? I'm not aware that the CFC has such a
policy. Occam's razor would prefer the explanation given with the
Rating Boon, which was that adult but not old players were getting
their ratings hammered (and some of them getting disheartened) against
quickly-improving but still low rated juniors. I never saw that
demonstrated statistically, but the anecdotal evidence was tempting.

This was just one of the 'reasons' I was hearing at a couple of the
tournaments I participated in. There seems to be alot of
misunderstanding on why this is going on. Hopefully the latest
chess canada news section will help players understand the
real reasons the CFC is doing this.

Aw, shucks. I'm not a high rated player like the other three
mentioned. But if you care to look, you'll find that my USCF rating
(2312) is higher than my CFC rating (2305), even after the CFC Rating
Boon. Not that I play in the US frequently enough to make my rating
say much about the current USCF rating methods.

But you are one of the BC Champs , and you were able to hold your
own in the BAP system GM Slugfest.


The trouble with a lot of these discussions on r.g.c.* is that results
are quoted that are out of date; that the message is distorted as soon
as it gets "repeated"; and that measures are misunderstood. Ooops.
That's three. Three troubles. Nobody expected a SI.

lol welcome to Usenet

I'd like to see a current 2007 comparison between CFC, USCF, FQE and
FIDE ratings. When I did that in 2003, I found that the differences
were surprisingly small.


I did find some stats on the CFC vs FIDE ratings on the CFC website
http://www.chess.ca/CFCvsFIDE.htm

J.Lohner

I'd like to see a current 2007 comparison between CFC, USCF, FQE and
FIDE ratings. When I did that in 2003, I found that the differences
were surprisingly small.

I did find some stats on the CFC vs FIDE ratings on the CFC websitehttp://www.chess.ca/CFCvsFIDE.htm

J.Lohner

Thanks for the link. As I suspected, the summary is based on a rating
list from April 2004. The price of rating relativity is Eternal
Vigilance.

--
Jonathan Berry

I'd like to see a current 2007 comparison between CFC, USCF, FQE and
FIDE ratings. When I did that in 2003, I found that the differences
were surprisingly small.

I did find some stats on the CFC vs FIDE ratings on the CFC websitehttp://www.chess.ca/CFCvsFIDE.htm

J.Lohner

Thanks for the link. As I suspected, the summary is based on a rating
list from April 2004. The price of rating relativity is Eternal
Vigilance.

--
Jonathan Berry