http://www.thedailyforehand.com/2011/3/1/2023938/computer-says-connors-was-the-best-ever
Of course this sucks but whisper should get a kick out of it with Agassi besting Sampras ;))
Imagine Vilas better than Borg !
ah ha ha ha ha ha ha ha ha ha ha ha ha ha ha ha ha ha ha ha
"Computer says noooo"
"The study also ranked the top players on each surface with Guillermo
Vilas as the best player ever in clay tournaments, while on grass and
hard surfaces the best players ever are Jimmy Connors and Andre
Agassi, respectively. It is interesting to know that Roger Federer
ranked as the #3 player on grass and Pete Sampras came in at #6."
Interesting that Borg, who supposedly was a relatively weak No. 10 of
all time overall, was 4-0 on grass against the alleged greatest grass
player ever and 12-2 (roughly) on clay against the alleged greatest
clay player ever. Wow -- I guess the computer's ALI (Artificial Lack
of Intelligence) didn't give much weight to this type of head-to-head
analysis. :)
What do they feed these computers?
Obviously NOT a complex set of 'factorings' to get 'results' like
this...
P
LOL. How they tuned that computer? Borg #8 on clay behing of Oranter,
Moya, Higueras and Dibbs????
Hardcourt specific list is also wonderful: Chang ahead of Mac. Brad
Gilbert ahead of Courier.
On hard: Becker #16 after Chang/Enqvist/Gilbert/Gottfried...
Apparently longevity counts a LOT in this analysis...but who buys this
formula where "one additional year on tour" equals about two slam
titles with this program parameters.
.mikko
I suspect data from Pakis? Look who is #2 !
More than that, this scientific approach considers 2007 belongs to
Rafa with his single slam win instead of Federer with his meagre three
slam trophies, 4 slam final appearances, YEC title and positive h2h
over Rafa for the year... It's all about unbiased 'prestige' they
insist, never defining what prestige actually is.
prestige=cheating and getting away with it perhaps?
Sounds like a Raja list. Sampras no.6 on grass & Federer no.3 (13
Wimbledon singles titles)?
Clearly on grass it's;
1 Sampras
2 Federer
3 Borg
4 Laver
On HC;
1 Federer
2 Sampras
3 Lendl
4 Agassi
Clay;
1 Borg
2 Nadal
3 daylight
4 Wilander
Looks like we have a ways to go before we get true artificial intelligence.
While pretty good for a computer, it's logic would rate it the worst
poster in rst. Given we have Haze/Raja that's quite damning.
What makes you think Wilander was better than Lendl? Lendl won 28
titles on clay more than Wilander and only slightly lesser than Borg.
And you forget to mention carpet.
I would never put Vilas over Borg. That list is shit.
Matey, giving your latest ramblings, I think it would edge you.
I have to say soon as I saw #2 thought Raja will be so happy at this
list!
it'd be a better predictor of matches than you Fedfans.
I can accept Borg or Federer above Lendl, not Connors.
I contacted author of the study couple days ago, Ph.D. in Physics,
working as a research fellow for some lab providing studies for
universities.
http://amaral.northwestern.edu/news/2010/nov/05/amaral-lab-welcomes-dr-filippo-radicchi/
We changed couple emails and he has not answered my last one yet, a long
email where I was rather critical, pointing out, statistically, some
flaws with his study...and offered a simple mathematical explanation why
the model does not work in this case.
He has not answered my email yet, I wonder if he will and what will his
answer be. There's probably some money etc involved in this. I also
wonder if I should have that discussion on a public forum instead?
Appears though that he now admits that this study is flawed. Also they
have taken the study off their front page after my last email to him
yesterday!
Well, he basically uses PageRank as the tool for his study. PageRank is
a model designed to rate web pages, and is supposed to be an actual
scientific method on networks...I personally have huge doubts on it's
validity...even on rating web pages.
Below is the most obviously flawed result he got:
**Top 30 players of the history of tennis in tournaments played on clay:
http://www.plosone.org/article/fetchSingleRepresentation.action?uri=info:doi/10.1371/journal.pone.0017249.s005
You notice that players such as Dibbs, Higueras and Moya are ranked
above Nadal and Borg... :-D
His explanation was on the case of Nadal that it takes some time for
active players to adjust 10-20 years...
On Borg his answer was that last episode of Borg's career lowers his
rank...which he said was 11 years of active play but is actually less
than 20 matches.
Of course the question is what are Moya, Dibbs and Higueras etc doing
that high...
My observation is that his model correlates highly with Career Matches
Won statistic...which is wrong, tennis matches are not hits on a web
page. In other words, more matches one wins - the better he is ranked
using this model. Quantity over quality. Bad science.
PageRank? For ranking achievements?
Damn you, study author, you make me quote frigging Hazelwood:
"AH HA HA HA HA HA HA HA HA H AHA HA HA H AHA HA HA HA HA"
Read my other post in this thread. This has very little to do with
artificial intelligence...
how on earth can you use pagerank as a ranking for tennis players?
only Raja would do that!!
> My observation is that his model correlates highly with Career Matches
> Won statistic...which is wrong, tennis matches are not hits on a web
> page. In other words, more matches one wins - the better he is ranked
> using this model. Quantity over quality. Bad science.
Yes, his approach is clearly volume-dictated. Is it a coincidence that
the top three men overall -- Connors, Lendl, and McEnroe -- are also
the top three in open era tournaments won? I don't think so.
That is perhaps the dumbest tennis ranking ever created, which is a
considerable achievement in a sport with a fan/pundit base noted for
making extraordinary efforts to create dumb rankings.
Why is Borg so low then? And Vilas so high and also Nastase?
Not actually tournaments won, but matches won...which is even worse.
Obviously. Any article with the words "Computer says" in the title is
suspect in my opinion. A computer can "say" anything!
>My observation is that his model correlates highly with Career Matches
>Won statistic...which is wrong, tennis matches are not hits on a web
>page. In other words, more matches one wins - the better he is ranked
>using this model. Quantity over quality. Bad science.
Hmm, I guess WTA uses the same model in the rankings at the moment...
yeah! he should be #8!
hummmmm....
hey. maybe you should inform the guy of 7543 !!!
Uh oh -- now we know that the "computer" in this analysis was actually
a Mechanical Turk with Iceberg inside!
> Uh oh -- now we know that the "computer" in this analysis was actually
> a Mechanical Turk with Iceberg inside!
I didn't know Iceberg was so small in stature :)
: You notice that players such as Dibbs, Higueras and Moya are ranked
: above Nadal and Borg... :-D
Federer, Kafelnikov or Agassi are not on the list at all, whilst journeymen
like Mantilla and Clavet are. Latter never made it even QF at Roland Garros.
The list is just hilariously absurd.
Other funny stuff...Santoro is on the hard-court all time list - ahead of
Safin! Well, I guess that makes sense, sorta.
Hey, Jarkko is ranked #28 over last decade \o/
: My observation is that his model correlates highly with Career Matches
: Won statistic...which is wrong, tennis matches are not hits on a web
: page. In other words, more matches one wins - the better he is ranked
: using this model. Quantity over quality. Bad science.
Article says:
"Players who had numerous wins over highly ranked players received bonus points and ended up as the top players. "
In some ways it sounds bit like Elo system?
At least the guy admits himself that the system is "somewhat flawed"...
Looks pretty comfy in there:
http://anticap.files.wordpress.com/2010/10/mechanical-turk-001.jpg
Yeah, somehow I was under the impression that only a person of small stature
could fit inside.
It does take into account the quality of wins, it also just adds, adds
and adds with each win, so even if you don't have quality, you can make
up for it with quantity.
There's even a small constant term which means the longer the career,
the more it will come into play ... How do you spell Connors?
--
"Listen first! Hiss later!"
Er, strike that second paragraph. It's not correct.
I quite tall but it was worth all the crouching down to get these
results out to the world!!
This I believe is correct analysis. It adds up each win, regardless of
losses in between, and AFTERWARDS takes into account quality of wins.
This means that a player winning 2 matches and losing 1 at a minor event
gets +2 wins to raw total...aka a positive result. While in real world
that would be seen as a poor result, zero or even negative for a top
player.
Meet new science.
Complex theory, networks, graph theory.