The Pal/CSS freestyle tournaments are a sort of unofficial world championship of centaur chess play, although correspondence chess players who pour dozens of hours of sweat into their masterpieces may object to this description. One thing is clear - the presence of a time control in the freestyle tournaments shifts the emphasis from art to sport.
The latest installment of this tournament, the 5th Pal/CSS Freestyle, took place March 23-25 and was won by Danish centaur-play specialist Dagh Nielsen. We had a chance to catch up also with a few of the participants.
Dagh Nielsen alias Flying Saucers (Denmark)
·
Winner of 5th Freestyle
· Runner-up of 4th Freestyle
Jiri Dufek alias Xakru, Etaoin Shrdlu (Czech
Republic)
·
Winner of 4th Freestyle
·
Runner-up of 3rd Freestyle
·
3-time Freestyle finalist
·
Over-the-board national master
·
Correspondence IM
Arno
Nickel alias Ciron (Germany)
·
Correspondence GM
·
2-time Freestyle finalist
Nelson Hernandez alias Intagrand, Cato the Younger
(USA + England)
·
Runner up of 5th Freestyle
·
2-time Freestyle finalist
Eros Riccio alias Rodo (Italy)
·
Correspondence IM
·
2-time Freestyle finalist
Jochen Rindfleisch
alias Kaputzze (Germany)
·
Finalist of 5th Freestyle
Nick Carlin alias Flying_fatman (USA)
Can you tell us a bit more about yourself and your involvement in chess?
Dagh Nielsen: Currently, my main interest is freestyle chess and opening
studies with this in mind. I am also a member of a local chess club and enjoy
the socializing that takes place around low-level club league matches.
Arno
Nickel: Over the years I practiced
and loved many different kinds of chess, but got most involved in
correspondence chess and lately in freestyle chess. I am still more fascinated
than frightened by the amazing progress of computer chess. Otherwise I would
probably prefer classical over-the-board chess and especially chess960, better
known as‚ Fischer chess‘.
Jiri Dufek: I am 6 years retired from OTB. This year I will make comeback to
correspondence chess.
Nelson Hernandez: Team Cato the Younger is actually a proxy for Team
Intagrand. The team consists of team leader Anson Williams, his girlfriend
Yingheng Chen in London and myself, Nelson Hernandez, in the Washington, DC
suburbs.
None of us are rated players. I personally have not played a game
over the board in over 20 years and if I were rated could not possibly rate higher
than an 1100 ELO, no kidding. Anson is a bit better but no great shakes
as a chess player either. Yingheng is not a chess player. We have
no affiliation or connection with the chess community outside of our Freestyle
efforts.
Nolan Denson: I am married and have 2
kids ages 25 and 16. I work for Intel Corporation. I have been around computer
chess since I got involved with chess in 1977. I use to be the owner of many
dedicated units. Once PC Chess got strong enough to beat the dedicated models I
have since then only play with software playing programs.
I have been the operator for many CCT tourneys. The
last CCT trny I was the operator for Scorpio.
Jochen Rindfleisch: I am an average club player with a local (German)
chess rating of about 1720 DWZ. I play at the best chess club in the world -
Chaos Mannheim. Due to my time consuming job, I have only little time for
playing chess except for late night online bullet games on playchess.com
and participation in online freestyle tournaments. I play about 6-8 over the
board tournament games per year.
Nick Carlin: I am 43.
I’ve been playing chess since I was a kid but never studied opening
theory, on a good day I might play to 1800, on a bad 1500 :) I work for a large computer company and once
hosted a challenge of 40 school kids vs Deep Blue II Jnr (only one node of the
monster that beat Kasparov) – I snuck in a game myself and managed to not win. Last year at least that one node was still running
in a basement in a Research Lab but it’s not on the network right now as it’s
level of OS (the AIX operating system) is too backlevel. I spoke with Frederic Freidel about porting
Fritz to the PowerPC platform back in 1994 but that never got off the ground at
that time. Since then I’ve been in and
out of the comp chess “scene” on ICC with various accounts such as “tribbles”
and “BigMomma”. I’ve been spotted on
playchess more recently as “Flyingfatman”.
It was only in late 2005 that I started to really get interested again
when Rybka came out and did what we all thought no chess engine could do – that
is crush all the other chess engines.
It was, and still is, an exciting time in computer chess.
Eros Riccio: I started playing chess over the board at the age of 14 years,
thanks to a friend who told me there was a chessclub in my city (Lucca) and in
a few years I got a rating of 2100, which allowed me to take my first norm of
National Master, but even now, after so many years, (I am 29 now) I am still
trying to achieve my final norm to get the full "National Master
Title". My current fide rating is 2175. I got my first computer with
internet connection in 2000, and immediately started playing correspondence
chess, first by normal (snail) mail, then by email, and now I prefer to only
play on webserver. I have had very good results in these 6 years, in corr
chess, in fact I am currently rated over 2500 ICCF and just achieved my Senior
International Master title. About computer chess, I first discovered, in 2000,
ICC, I used to play some games with computer there as centaur, and at the end
of 2001 I got the fritz7 program that made me discover the great Playchess
Server, which I preferred by far to ICC, and so I switched to engine room
there, mostly
The first thing that jumps out when browsing through the games are the incredibly deep theoretical battles in the Najdorf. Many games follow the same first 20 moves. Are there objective reasons for this? Or is this mostly a matter of fashion?
Dagh Nielsen: With the level of opening knowledge and preparation
displayed in these freestyle events, some sequences of moves probably have to
be considered more or less forced. On the other hand, earlier deviations should
definitely be possible, but that could require a lot of independent
investigations that may end up being futile anyway. In a sense, it is just
economical to follow the fashion, because you will then study exactly those
lines that you are most likely to meet.
Arno
Nickel: For one part it’s probably a
fashion, but it also reflects a strong will to go for a win, taking risks and
hoping to find killer moves in the middlegame.
Nelson Hernandez: The reason for this is very straightforward. At
this time the most frequent opening played in the Playchess.com engine
room is the Najdorf. This situation has happened because of the recursive
nature of .ctg opening books, which overweight moves with higher N. The
Najdorf is a risk management technique as there are lots of statistics that
accompany this opening. Furthermore, quite a few of them have invested a lot of
time conducting deep analysis for this opening.
Nolan Denson: I have noticed the same
trend and looks for less well know lines to deviate. But, it seems those wins
very quickly become well know lines also. I think most of the lines came from
what was setup in books that came out with the programs. It was the most
popular and many just let their programs play what ever it wanted to play. Now
since PAL has came along many are fine tuning these lines with over night
analysis.
Jochen Rindfleisch: Due to the limited time, available for
preparation, we trusted into the Najdorf main lines, we had assembled into our
opening books. The rationale was that there would be only little surprises left
in this extremely popular opening and thus the danger of running into a
prepared opening trap, unknown to us was minimized. The popularity of this opening - at least in
freestyle and engine tournaments might stem from the fact, that people
iteratively improve their books. Thus, the books tend to contain more and more
variations of the Najdorf than on other openings and thus the likelihood to
play the Najdorf in those games is higher. Me personally, I would like to see
people exploring other openings with equal intensity.
Nick Carlin: I think it’s the latter, I suspect that
whatever lines are currently being played a lot on the internet chess servers
you will see in a Freestyle.
Eros Riccio: well, Sicilian Najdorf is the
most frequently played opening in chess, especially at high level, so I was not
surprised to see so many of them in this freestyle. The theoretical battle also
was great to see, everyone was so well prepared that I think some top level gm
might steal some of the variation that were played there :-) I think the best
Novelty might be Zacks' move in Rodo - Zacks game: 18.Kb1 Rxa2! (18.Kb1 seems
to be a novelty for white too, even thought it has been played many times in
playchess engine room, but after that move, I found no engine games that
continued with 18...Rxa2!, that's why I consider the real novelty 18...Rxa2 and
not 18.Kb1)
The Najdorf itself is disproportionately popular in these events, for both colors. What is it about this opening? And why do the relatively less well-prepared teams enter these battles?
Dagh Nielsen: One advantage of the Najdorf (for both colors) is
that it is so difficult to handle, even with an engine running on steroids
beside you. With white, you usually have to come up with some precise attacking
schemes (or ingenious positional maneuvering), or you can easily find yourself
in a worse ending or under a fierce counter-attack. And if white manages to
establish a dangerous attack, a black centaur player will often see his engine
suggest inadequate responses.
As to the second question, I guess that the most common way to realize that you are relatively less well-prepared is by achieving an inferior position in a serious game. If I am not mistaken, there is a Chinese saying going something like: "No matter how big a knife you have, there will be somebody in town with a bigger one." Playing the Najdorf can be seen as a bet that you are the one bringing the biggest knife to the table. Probably it also deserves mentioning that the Najdorf is just a very rich opening (again, for both colors). One can almost always look for new or rare ideas that may surprise the opponent and give you the upper hand.
Arno
Nickel: To me it seems, that
opposite castling is attractive for many players on both sides, although some
players prefer to castle kingside with White. Such positions are considered to
be very difficult for engines. So, there might be space for human creativity on
both sides. I am not surprised to see many centaurs playing the Najdorf, but
why are also many engine-only players opting for it? May be, the climax of that
fashion was already in the 5th PAL/CSS Freestyle Tournament and we
will see a change in future freestyle events.
Jiri Dufek: For weaker team (worse HW, weaker OTB player or worse
opening „book“) is much easier follow some main „tunnel“, it’s not important,
if in the Najdorf or in different opening. There is little space for mistake
(long variation have drawish tendency :-) ), but this kill creativity and
chance for winning game.
Nelson Hernandez: The opening's popularity is explained above--it
is a quirk of fate. It could very easily have been a French or a Pirc
line that got caught in this feedback loop. As for why less
well-prepared teams enter these battles--simple enough. They have
collected quite a number of Najdorf games. They too feel safest staying
in book, so they follow the most common lines.
Nolan Denson: Its all about who did the most
work to fine tune their books. Like i mentioned earlier many unaware as to how
deep others books can be feel into the traps of this opening by others that did
fine tuning.
Jochen Rindfleisch: As I stated in my response to the previous
question, it is very tempting for less well-prepared teams to stick to the main
lines. Those teams might not have the time to prepare less popular variations.
Nick Carlin: The Najdorf was good enough for Kasparov
so maybe it does have some intrinsic value for black :) Still, I think the frequency of it is
popular fashion as I mentioned. If you
aren’t doing opening prep you’ll go with the lines in your book, and if you
built the lines in your book from say playchess then you’ll have a lot of Najdorfs
in there.
Eros Riccio: besides the things I said
above, I may add that playing a main line like Najdorf guarantees first of all
a sound opening, it's well known that main lines are very hard to refute, and
it's also well known that when black plays Sicilian, Najdorf especially, he has
good winning chances and the games are not boring at all. For these reasons,
probably, even some players that usually wouldn't play Najdorf as their main
variation, might have tried it, relying also on the advantage of being able to
check databases
Can you tell us a bit more about your method of opening preparation? What fraction of your prepared moves are going to stand the test of time? How much preparation time goes into this?
Dagh Nielsen: In periods, I spend quite a bit of time on opening
studies. Usually I just let my curiosity determine the direction of
investigation, though of course an eye is kept on known developments. Then, the
last two times a freestyle tournament (+final) was on the horizon, I spent some
time drawing conclusions and making more concrete decisions about what to play
and what to scrap.
I can't recall that I have ever played moves that I
knew were unsound. On the other hand, I have no doubt played many white ideas
that will one day be or are already considered undangerous. Also, my 24...Kh8
novelty from the first round of the final already came under fierce attack in
two later games between other teams in the final. I guess the nature of the
beast is such that you can only hope that you will be one step ahead in the
theoretical discussion. Ultimately, any line in chess can either be refuted or
avoided.
Arno Nickel:
Unfortunately I have only very little spared time for opening preparation. So,
I wouldn’t say I ever worked on it systematically. Freestyle is a very special
thing. That’s why you can’t simply adopt your lines from correspondence chess
or o-t-b chess. You can do it, when playing for a draw, but in order to win you
have to create something special. Set your opponent under time-pressure, get
him into a variation, where he feels either uncomfortable or on the contrary
over-optimistic. If your opponent is an engine-only player things are also
quite different. In my view opening play always has an important psychological
impact, not to forget your own individual disposition.
Jiri Dufek: I analyze permanently different ideas, without any
influence of the Freestyle for this. Time to time i analyze ideas from
playchess, but not with main idea using this analysis during Freestyle.
Nelson Hernandez: Most of this is "classified" as the last
person to whom we want to give insights is the deadliest Freestyler in the
world, namely you. However I would mention that in my view the pre-game preparation
is more important than the in-game performance, and this consists of deeply
considering strategy ("doing the right things") and getting prepared
to execute tactics ("doing things right"). Strategy comes in
several categories, like deciding how you are going to make decisions during
the game, what repertoire of lines you will prepare, what you intend to do
against specific opponents, what behavioral disciplines you intend to follow
during the match, what opening book(s) to use during the match. In short,
you need to develop a comprehensive approach to the game from end to end ahead
of time and improvise as little as possible on fundamental issues during the
tournament.
Jochen Rindfleisch: We assembled some opening books which relied on a wealth of
correspondence chess games. The preparation time was nearly nothing, since we
constantly keep those books up to date
Nick Carlin: What is opening preparation? For the freestyle we didn’t do any! This of course could be a factor in our not
qualifying for the main event. Since
then, I’ve decided that you need to analyze your opening moves to at least
depth 20 before committing them to a book.
The advantage one has in opening preparation for a Freestyle at the
moment is that you can be almost 100% sure your opponent is using Rybka, so you
can tune your openings against Rybka.
Playing 1. e4 is not a bad idea, and responding 1. … c5 or e5 is pretty
safe too. There are some nice surprises
in the Reti for Rybka also, but you can find those yourself.
Eros Riccio: I am a firm believer of
main lines. I always play them, unless I am trying to surprise my opponent. I
repeat what I said before, there must be a reason why most of the top level
players play them, they are sound and are formed by just the best moves that
can be played! Every move of a main line follows the best principles of
developing pieces, every move has a good plan to achieve. This may be true with
sub lines also, but for example, a Nimzowitsch defense (1.e4 Nc6) will hardly
allow black, with correct play by both colors, to find a good attack or just
get a good position. This because 1...Nc6 simply doesn't follow at best the
spirit of pieces development. Said this, I created about 4 years ago, my
personal all manual opening book, I must confess that at first I inserted also
some sub-openings like queen's gambit accepted, or modern Benoni as black after
1.d4, or Alekhine's defense after 1.e4, but just because I played those as
human too and had fun in watching my engines play them too. But after some time
I decided to abandon the personal fun, and gave up all the sublines and focused
on Sicilian against 1.e4, and the Indians and semi Slav against 1.d4. As I
said, I am working every day for some years on my book, what I do now that my
main lines are well defined, is just watch my lost games (or games where I was
out of book with a bad position), and find better moves to add (or realize
which were the bad moves in book and mark them in red).
What are the most important qualities of a top centaur player?
Dagh Nielsen: It's difficult to put on a list, but I think one
important mix of qualities is:
1) Understanding the weaknesses of your particular engine of
choice. What kind of moves or positions does it tend to underestimate or
misevaluate.
2) Reacting properly on engine output based on these insights.
This would include integrating a "human touch" in these games.
I guess both parts take some experience to learn and
to master.
Arno
Nickel: Apart from knowledge and skills
in all respects of computer-assisted play, it’s strong nerves, patience and
something I would like to call „professional management-control“, as you have
to manage a lot of different components under time-pressure. May be also team
ability is required, if you play together with others.
Jiri Dufek: Brain, experience with analyzing and powerful HW.
Nelson Hernandez: Without question the #1 attribute is steady
nerves and a cool temperament. Excitability under pressure, general
anxiety and manic-depression are all fatal. #2 is general ship in
the military sense--having a strategic outlook before the tournament even
begins and figuring out how to tactically gain the upper hand against opponents
with known or suspected weaknesses. #3 is technique--how to most
profitably spend and manage the time available during a match, and
how to most efficiently integrate the data streams coming in to form solid
conclusions. #4 is sheer physical and mental stamina; if you lack these
you will be prone to make game-ending mouse-slip errors. #5 is your
hardware/engine combination and overall hardware resources. #6 is a
top-notch opening book. #7 are comprehensive 6-man EGTBs. #8
is a capacious bladder and cast-iron set of buttocks. #9 is a good
night's sleep between tournament days. #10, and Anson would argue this
should go a lot higher, is fervent prayer.
Nolan Denson: Prepare with good books.
Jochen Rindfleisch: Having finished 6th with 9 consecutive
draws, I do not feel I qualify for the title of a "top centaur
player". If I were asked to give a ranking of the most important factors
for success in centaur play, it might look like this: 1. Very good hardware
with a reliable operating system and chess software. 2. Outstanding skill in
interactive engine analysis. 3. Intense and elaborate opening preparation (to
win a full point requires more than just a good selection of games for the
opening database) with a wealth of deeply prepared innovations. 4. A good
knowledge of chess strategy and the standard plans for any given position
combined with a good feeling, when it is worth while to look for exceptions to
rules. 5. A realistic estimation of the limitations of the human players.
Sometimes, the engines simply know better...
Nick Carlin: Someone like Dagh Nielsen is much better
qualified to answer that question than me. I imagine it is the ability to know
which PV to follow in multi-PV mode, and to know when your engine needs help,
and often it does not of course. This
ability comes from a playing a lot as a centaur no doubt, but also being just
plain good at chess. Opening
preparation and a desire to win are clearly very important factors also.
Eros Riccio: He needs, like in all
things, experience in the field. A Correspondence player, which is used to
analyze with engines, has probably a big advantage. I completely agree with
Ciron when he says Freestyle is like blitz Corr Chess. Of course, differently
from corr games, the centaur player needs "cold blood", in order to
avoid excitement which may lead to mouse slips with little time left on the
clock. I was very impressed by the centaur operator of Cato the Younger, in his
game vs. Flying Saucers. He demonstrated to have the "coldbloodness"
I was talking about, I saw him play some of his moves, without mistakes, in a
very difficult position, with only a few seconds left on the clock.
It's pretty evident now that a top centaur combination is stronger than an unassisted engine. How would you quantify this difference? By what margin could you win a match against the newest Rybka running on your strongest machine and using Noomen's latest RybkaII.ctg opening book?
Dagh Nielsen: I would not dare to try
and quantify the difference in playing level. Personally, if I had an engine running
on auto, I would be horrified about the risk that a type of position is entered
that the machine simply will not play very well on its own. An additional
advantage for the centaur is that he can identify critical situations. Perhaps
at some point in a sharp game, the game will essentially be won, drawn or lost
within a span of, say, 5 moves. Also, I don't think pure engines can be
expected to react properly on tricky novelties, and I am not really aware of
any method to ensure that you will be the last one leaving preparation in every
single game.
Arno
Nickel: Knowing that Rybka plays
with its own opening book might be a decisive advantage for the centaur.
Further on the time-management of unassisted engines is one of their
weaknesses. These two points should guarantee a score of about 66%. But without
knowing Rybka’s opening book the score should not be more than 60%.
Jiri Dufek: I can predict, that Xakru team under this condition
probably scored about 70-80% (match with minimum 12 games).
Nelson Hernandez: Quantifying this difference is difficult to do as a
lot depends on the centaur's technique (see #3 in previous paragraph).
This is further muddled by the relative strength of each side's opening book,
hardware, access to EGTBs etc and other factors previously
mentioned. However my general impression is that, all other things being
equal, a top centaur has a 100-150 ELO advantage. You can calculate
how that translates into a success rate. An indifferent centaur might have
no advantage at all.
Nolan Denson: Centaur don't walk blindly
into the well know traps. Most programs when following books made for it do not
always follow the best line. - 75%
Jochen Rindfleisch: As the most recent tournament has shown, only the
best - or most lucky? - centaur teams perform better than a well adjusted
engine. So - taking a wild guess now - I would hope for 53% in a Kaputtze/Rybka
match, if using identical hardware for analysis.
Nick Carlin: 60-40 is my guess, I suspect that the
centaur advantage will come especially in the end game, where human assistance
can avoid losses and win games that engines might draw. An example is in some rook endings where the
engine will push a pawn to the seventh rank and obtain a draw in a won
position. The missed win is coming
through leaving the pawn on the sixth rank and bringing the King up to support
its coronation.
Eros Riccio: the advantage(s) of the combination Human(s) +
engine(s) is obvious: In general, an engine alone is "stupid". It plays
random openings, wastes precious time on forced moves, stubbornly wants to play
for the win, sometimes forcing the position and losing, when a draw would have
been enough to qualify... so, relying on an automatic engine may be quite
risky... a human instead, may control all those things which an engine can't.
The only advantage of automatic engine I can see, is with little time left in
difficult positions, as there is no risk of mouse slips and losing on time.
It's also pretty evident that many strong over-the-board players have trouble adapting to this format and play under the level of a standalone Rybka. What is the reason for this? Could one week of good training, supervised by a centaur-play specialist, overcome most of the problems? Or are there deeper issues, for example with the opening repertoires, etc?
Dagh
Nielsen: I can think of five relevant
observations:
1) As mentioned above, I think a crucial quality in freestyle chess
is to be able to interpret and react properly on engine output. Even though
strong players may use engines frequently during their daily work, they may not
necessarily focus on this particular aspect. As a speculative example, one can
point to the Kramnik-Leko Marshall from their Brissago match. Yes, the Kramnik
team was busy trying to come up with new ideas and didn't have time to check
and re-check everything, but I think it is fair to speculate that a strong
freestyle player would not have stopped analysis at the point where the Kramnik
team did. He would (or should, at least) sense that the engine evaluation was
not stable and that further analysis was required, based on earlier experience
in similar types of positions.
2) I think another place where strong OTB players may come up short in freestyle chess is in the area of opponent modeling. For one thing, they may have antiquated ideas about the main weaknesses of engines (even more so after the introduction of Rybka, which has a distinctly different style than, for example, Fritz). As an example, I once saw a strong human playing as centaur try a Scandinavian against a pure engine. The result was a sure 1-0. The pure engine had no problem playing the type of positions resulting from the opening very well on its own. Similarly, strong OTB players may have inappropriate ideas about how to assault a centaur operated by a relatively weak human, again as a result of lack of experience. Another example from an earlier tournament: A +2600 GM playing as centaur played first a Caro-Kann as black, then an English as white. In both games, he reached middlegame positions of a type where understanding is traditionally regarded as more important than calculation. Nevertheless, he got duly outplayed in both games by "no-name" centaurs. For what it's worth: I think it is pretty damn difficult to outplay any pure Rybka in an equal ending, unless it happens to be of one of the few types where Rybka (or other engines) has some (commonly known) weaknesses. On the other hand, I do think that it is very easy for a pure engine or a weak centaur to throw away a win or fail to put the opponent under pressure in a favorable ending. I guess the conclusion is that one should first and foremost aim to achieve an advantage in earlier stages of the game, and that mainly by presenting the opponent with some complicated problems that would require some strong human input to solve properly.
3) This is very speculative and based on insufficient
observation and analysis, but I sense that strong OTB players sometimes
overreact in freestyle chess once they are on the verge of getting an inferior
position. Perhaps radical measures is a justified reaction in such situations
in OTB play, but less so in freestyle chess. Also here, experience seems
necessary in order to estimate where the opponent is most likely to play
sub-optimally and let his advantage slip away.
4) The opening repertoire and preparation issue is complex.
With all due respect, I think it would be wrong not to mention the results of
the opening stage in the games of GM Juri Solodownitschenko (=Engineer) in the
final. Juri always plays the open Sicilian as both white and black in OTB play.
However, if we exclude his round 3 game of the final (where the opponent pure
engine for an unknown reason played moves not intended by the book-maker), Juri
only got a favorable Sicilian in 1 out of 5 games, and that by adopting the
line I had used against him in the previous round. If we generalize, I can
think of various explanations for this phenomenon occurring when strong OTB
players enter the freestyle tournaments. Firstly, I think it is natural to
assume that practical factors are relatively more important in OTB play than in
freestyle chess. Things like intimacy with the opening structure, the element
of surprise, and initiative may be of a relatively higher importance than the
ultimate verdict of an opening line. If we additionally combine this with an
assumed insufficient experience in modeling and predicting opponent weaknesses,
perhaps one can understand why GMs may not succeed in the opening stage in
freestyle chess. Aside from this, I think one should no longer underestimate
the opening level of many of the participating "no-name" centaur
teams. I believe many of these participants will turn out to be keen and very
experienced opening analysts. In particular, in sharp opening systems, I
believe there is little reason to assume that they will generate worse opening
analysis than even the very strongest OTB players in the world (I do have good
reason to assert this). I could continue by mentioning and discussing the
Playchess engine-room, but... let's not make things more complicated than they
already are :-)
5) Finally, there is also the possibility that the playing
level in these freestyle events is simply extremely high. This is only an
anecdote, but nevertheless: Some participant in the 5th preliminary had the
help of one or two (depending on the round) very strong and from ICC well-known
GMs. In round 7 or 8, this couple gave their fullest against, if I am not
mistaken, Kaputtze (who qualified for the final). Allegedly, they were shocked
by the level of play, and one of them exclaimed after the game: "I can't
play any better than this!".
Whatever the reason is, we are so far yet to
see a GM step forward and give an explanation from "their" point of
view.
Arno Nickel: It
depends on the individual, but in general I would say, there are also strong
psychological barriers, you cannot tear down in one week, may be even not in
one month. It’s mostly a matter of the right attitude.
Jiri Dufek: Probably
question based on Engineer and other GM’s results. GM’s needs more analysing
experience, different opening repertoir prepared for games against engines and
powerfull HW.
Nelson Hernandez: No question there are multiple problems with OTB players. Their biggest advantage is that they are not Playchessers and thus will be far less inclined to follow Najdorf lines. That gives them a certain edge. But that is more than offset by any combination of other problems: inferior hardware, inferior software, weak opening book, no EGTBs, inexperience as a centaur, lack of preparation, lack of technical computer skill, and perhaps a tendency to hold the computer in contempt and go with their gut in inappropriate situations. On the other hand an experienced correspondence chess player almost by definition has resolved a lot of these issues ahead of time.
Nolan Denson: Most do not spend a lot of
time playing online. They think because of their OTB rating they can come in
and wipe out others with the same program. I think the key is understanding
your opponents weakness. Many that play regular online have many games about
their opponents weakness and strengths. Its all about who prepares their
openings better. Understanding the program, which may be aiding you weakness
and strengths.
Jochen Rindfleisch: In my ranking list, given above, the
points 1, 3 and 5 might be the main areas of improvement for strong over the
board players. Point 3 is especially interesting here. My assumption is that
top over the board players have a very fine tuned opening repertoire for over
the board play, which allows to fight for a full point against equally strong
opponents. In a higher sense, a lot of the lines played over the board might be
incorrect though, which is quickly evident if applied in centaur games.
Nick Carlin: I think the issues are much deeper than
mere training, I think it’s where you come from and how good at adapting you
are. Some GM’s may have bet their
entire personality on their ability to play chess and they are not going to
hand this over to a comp without a fight!
Other GM’s, and for me the best example here is Baadur Jobava, have
learnt to trust the comp when it is right and proper to do so. Right now, I don’t think that centaur play
is for everyone is what I am saying.
Eros Riccio: Indeed, I don't understand
why still so many players let their engine run alone, in important games,
instead of playing themselves as centaur... I have no idea, maybe they just
want to enjoy the game as kibitzers... we should ask them about this :-)
How well do you think Anand or
Kramnik or Topalov would fare in the freestyle final without any engine
assistance? How about the late-90s version of Garry Kasparov?
Dagh
Nielsen: Well, we recently saw Kramnik lose
to Fritz 10. I have no doubt that the strongest freestyle centaurs are at least
300 Elo points stronger than a stand-alone Fritz. Furthermore, if we take the relatively
short time control into consideration, I have problems believing that any
"human-alone" could obtain a positive score even in the preliminary.
On the other hand, if the time control was 3-4 hours for each side, and we were
talking about a match-format, I think the odds would be more unclear.
Arno Nickel:
At these time-controls (60m/15s) no chance to qualify for a
final. May be one of them would have a slight chance to qualify at longer time
controls.
Jiri Dufek: Probably about 20-30%, based on selected opening.
Nelson Hernandez: All of them would be obliterated--even Kasparov
in his prime, who is a man I practically idolize. The games being played
in Freestyle finals have to be of a 3000+ ELO quality. Humans altogether
too often make a succession of small errors that cumulatively lead to
defeat. Engines make bad moves too, but a good centaur really minimizes
that margin of error. It seems utterly implausible to me that Kasparov
could have gone +4 against the finalists in the most recent Freestyle and won a
share of the lead without computer assistance. In fact I doubt his score
would have been positive.
Nolan Denson: Anand or Kramnik or Topalov
would do very poorly, Kasparow - not so well also. I truly believe the games on
PAL are results of the best of the best. How do you not know some of the(m) are
not assisting others?
Jochen Rindfleisch: Uh oh - I think I prefer not to give an answer to
this question :/
Nick Carlin: I’ve played a lot of engines against top 10
GM’s on ICC and (annoyingly) they can draw them more easily than some might
think. Still, the humans do blunder
occasionally (just look at Kramnik!) so I would go 65-35 in favor of the comps.
Eros Riccio: I think the freestyle
tournament is by far the highest-level tournament ever! I think for sure the
top GMs and even Kasparov in the late 90s wouldn't have been able to qualify
for the finals
Two players in the final, including GM Nickel, drew all of their games.
Browsing these games doesn't reveal any especially drawish behavior. Is there
some subtle reason for this? Or should it be viewed as a normal result?
Dagh Nielsen: Yes, it should be viewed as a normal result. It takes
two to tango, and even then the result may well be a draw.
Arno
Nickel: No special reason. I had
failed to win one or two of my four advantageous games, while I was lucky not
to blunder in one of the other games. May be the specific format of these
finals (round-robin tournament with the prospect of three money-prizes) prevents
players from taking unsound risks. Secondly, players prepare on each other and
take care.
Jiri Dufek: Player very often only follow engines without enough
own ideas. These games have „tendency“ ended in a draw.
Nelson Hernandez: That's a very good question! Nickel's run of
draws goes back to the tiebreaks and the qualifiers too, something like 15 or
16 in a row. It seems statistically implausible that this is a
coincidence. But just the same, the higher the quality of the chess
games, the higher the draw percentage will be. As technology and software
advance further I would expect the finalist draw percentages to trend even
higher.
Incidentally I am not entirely certain that Nickel does not demonstrate drawish
behavior. You would need a statistical study comparing the positions
he gets into at certain points in the game against those of other Freestylers,
and compare their respective historical draw %s. I cannot be sure, but I
hypothesize that he may be steering the game into historically drawish
positions by a slightly higher percentage than everyone else and in
competition of this caliber that might be making a difference.
Nolan Denson: Not knowing the program
they are using to assist at times better than those that play often online. I
believe they are trusting their own skills more than program they may be using.
Jochen Rindfleisch: Being one of the two "kings of
draw", I have thought a lot about this question. We played for a win without
taking uncalculated risks. Both my team as well as Arno Nickel might have used
books, derived from correspondence chess games. My theory is that it takes a
lot of time to prepare a book, which is not only safe (granting a good play)
but which also contains new, unpreceded killer variations. I did not find the
time yet to analyse all games of the tournament with a decisive result, but I
theorize, that only rarely, a game was won out of an equal middle game - at
least in the finals.
Nick Carlin: Interesting question.
Chess should be a draw I think, although I know one titled player who
says “you’ll find out black wins when chess is solved” :) I’ve seen some really interesting (but few)
wins by Rybka when you make her play gambits (like Smith-Morra for instance) –
you won’t see these type of openings in a Freestyle. Maybe in chess, like in life, sometimes you need to speculate to
accumulate.
Eros Riccio: The big quantity of draws,
I think, are due to rybka. Rybka is just extremely good at defending even in
very difficult positions.
On the other extreme, the winner of the tournament, Dagh Nielsen, won 3 games with black, all in ultra-sharp Najdorfs. Can this trend continue?
Dagh Nielsen: I certainly hope so :-)
Arno
Nickel: No.
Jiri Dufek: :-). Probably all 3 white players known repetition in
„poisoned pawn“ variation, but all went won game. Maybe next time Dagh will be
challenged in this variation :-).
Nolan Denson: No, I think that is over
and theory has already changed. Only those that don't play often and come out
without the latest updates will fall into those sharp lines.
Nelson Hernandez: Sure it can continue, if people are stupid enough to
play Najdorf against Dagh. You will notice Dagh did not beat Cato with
black because, in part, Cato played the entirely irregular 1.h3.
This silly move resulted in Dagh spending over seven minutes thinking about a
reply!
Jochen Rindfleisch: The fact that Dagh succeeded so well in this opening
shows to me, that the value of the Najdorf is still under discussion. As it is
extremely sharp, the horse power of the used hardware might play a larger role
than in more strategical variations.
Nick Carlin: I hope so, it makes for interesting
play. Although I think one of those
Dagh wins was a book win if I recall what he said correctly. I think as the engines and hardware improve
we will see lots of new answers to old problems. 1024 cores in a desktop by
2012 will be fun (providing the chess programmers can make their babies scale,
Vas :).
Eros Riccio: Flying Saucers was really
impressive. What made the difference, in my opinion, was his excellent book
preparation (mostly) and his experience as centaur.
It gets harder and harder to win games in these tournaments. The draw rate continues to sit at around 60 to 70%. The fight to qualify to the final for the top teams is mostly a fight against draws. Will changes to the format eventually be needed to deal with this issue?
Dagh Nielsen: The problem is that match formats can also be
draw-festivals. In a sense, I think the Swiss and Round Robin formats are the
best way to give an incentive to players to play risky and bold chess.
Arno
Nickel: The draw rate in the main
tournaments is quite normal. It’s only in the final, where the draw rate
increases dramatically. That’s surely a point to think about.
Jiri Dufek: Probably not, but as not only joke – how about
football rules? 3point=point, 1 point=draw
Nelson Hernandez: No. There is nothing wrong with the
format. The problem is that you have 2900-3100 ELO chess
entities playing each other in finals. Our feeling is that chess players
of that quality are inherently hard to beat because they make very few errors
and they rarely take risks. The only way you could increase the number of
wins would be to either create larger disparities in the ELOs of the players
(an absurd idea) or to reduce ELOs generally by creating
conditions whereby everyone will make more mistakes. The only
plausible way to do this would be to reduce the time controls considerably,
which in our view would be stressful almost beyond endurance for most
players. We like the current time control, draws and all. Every
game is a tough struggle, and some of the games are epics.
Nolan Denson: I don't like the current
tie break system. I think some one must
win to move on. Ties/draws show not determine the one that advance.
Jochen Rindfleisch: There are 2 conceivable ways how the
format could be changed. Increasing time pressure would increase the
probability of dramatic errors and "operative losses" (mouse slips).
This would have a positive impact on the draw rate but not on the attractivity
of the centaur tournaments.
Reduction of time pressure, in contrast would favor creative
interactive analysis of positions and thus increase the quality of the games.
It is not easy to deduce an impact on the draw rate in this case.
Nick Carlin: Yes I think changes to the format are inevitable,
especially as more players come online, although I am not sure of what the
answer to the “draw problem” is yet, were any games “GM draws”? ie. Two teams
agreeing to draw before move 20 or so, those at least should be monitored and
avoided.
Eros Riccio: I don't think so. Now with
Rybka around, most players will continue to only draw, the strongest players
will win one or two games more, and they will qualify. The standings will be
very short, but I like it that way too. What I mean is the best players will
still win, with more difficulties, but they still will. Draws may be seen as
boring, but I think it's an unavoidable thing with general level of play
improvement.
Are there any changes to the tournaments that you'd like to see? Feel free to mention anything - format, rules, tournament software, participant composition, etc.
Dagh Nielsen: I would like to see more titled players
participating, and I would like to see more recognition of these events (like,
the games are mentioned and discussed in chess publications). I do think that
freestyle chess has the potential to become what human-engine chess was 10
years ago, but some more intensive and timely promotion would be required.
Related to this, I would like participants to be more open about their methods
and team composition, but I don't know how to ensure this (since secrecy will
always be possible and advantageous from a short-sighted point of view). In
general, I see no reason why the next freestyle tournament should not be able
to attract 500 participants instead of the ~130 who participated this time.
I do have
one specific request to make: I think it should no longer be possible to
connect an engine directly to the tournament so that it plays "on
auto". I also don't think that the "centaur" mode should be
supported. For several reasons:
1) We have seen time and time again that pure engines
encounter software problems (either because of the GUI or because of the
engine) and that games even have to be started all over for this reason. I
think this is very unfortunate and somewhat unacceptable.
2) We have also seen repeatedly (but luckily only as
exceptions) that pure engine players playing vs. centaurs decide to play on in
absurd situations where the position is a dead draw, perhaps hoping for a
mouse-slip from the human player. I think this hurts the credibility of these
tournaments.
3) It is clear that pure engines are not as strong as centaur
entities, so nothing is really lost by discouraging this kind of play.
Furthermore, it would still be possible to play as pure engine, only one would
have to simulate the game in a second interface.
4) Especially in the last couple of editions of these
tournament, we saw a tendency of teams entering several nicks in the
preliminary, letting some of them play as pure engines in the first couple of
games. Then they would pick the nick with the best score so far and focus on
that nick with their full combined forces, and then let the remaining nicks run
on ridiculously outdated hardware, or maybe even let them drop out of the event
entirely.
I think it would be far more preferable to level the
playing field by simply using the principle that moves need to be entered in
the interface by hand movements just like they are always done in normal server
play. It's simple, it would enforce the principle "one human - one
nick", it would save us from many of the problematic issues that we have
seen hitherto, and it would ultimately help give these events a more
professional and clean status.
[Obviously, the tournament rules need to be put into stone, and software
bugs should be eradicated so that they can no longer influence the course of
events.]
Arno
Nickel: There are many things still
to be improved, as freestyle events are rather new. It would take too long to
go into detail by now. The planning and organization of freestyle tournaments
is almost as complicated as freestyle chess itself.
Jiri Dufek: Freestyle need strictly clear rules about qualifying and
finale – every from last three tiebreaks had different rules on the edge of the
correctness. In rules was too proposed, that finale will be played by 8
players, but finale was played by 10 players (as previous). There is problem
with disconnecting during game – it’s hardly acceptable having problem with
resending interrupted game to players (based on presumption, that player have
some unique technical problem, not that player interrupt every game)
Nolan Denson: Better tiebreak system, at
least going to a quicker time control until some one wins. Advancing on draws
doesn't prove who's better.
Jochen Rindfleisch: To me, the tournament was big fun, even though it had
its unpleasant moments. It would be an improvement for the tournament if the
rules were more precisely defined, since all rule discussions reduce the
positive public reception of such an event. Chessbase might be interested in
providing a tournament management system for this style of multi- day events
and remove at least some of the most annoying bugs. Last not least, I am always
surprised, how little the events are advertised before the tournament starts.
Had I not received an invitation per email, it might have easily slipped my
attention.
Nick Carlin: For me, the Freestyle event is just too
long, let’s do it in one day, I probably won’t participate again due to the
time commitment.
Also, there were a couple of teams missing from the
Freestyle 5 final, notably Rajlich’s and Jobava’s, who I think would have been
there had the tournament been managed in a different fashion. One thing that is
clear is that the ability to manage mouse slips should be added to the GUI
being used. I’d like to see Freestyle
6 (if there is one) hosted by ICC just for a change, maybe alternate them
between ICC and Playchess. The other
problem with the current format is when two teams qualify for the final and
they have a connection, such as a common member playing in two teams. Such a connection could be used to
manipulate the results.
Eros Riccio: The latest freestyle was
excellently organized in my opinion, the only important thing the server should
fix is, when one reconnects let's say after 15 minutes, the game should be
resumed with the correct time on the clock.
Thank you for your time. Wish you
good luck in the next Freestyle event!
– Vasik Rajlich
Links:
·
Dagh Nielsen comments on „Flying
Saucers” games from 5th Freestyle - view
·
Dagh Nielsen comments on „Flying
Saucers” games from 5th Freestyle - download (zipped PGN)
·
Arno Nickel
report on 5th PAL/CSS Freestyle tournament: „Teaching an engine how to fly”.