$20,000 ICUC - a report on wins

[Ozymandias]

well as former ICCF world champion i presume Ron L did already
face some serious opponents you know..

as for the difference between correspondence and freestyle chess,
well, during the second half of the Infinity tourn i approached the game in a similar way as i'm doing correspondence chess on ICCF; only this freestyle games were much faster, ofcourse.

Finally, yes the drawing rate is going up, and to avoid that the top ranking is determined by luck of having face a weak opponent (or someone with a disconnect or so), it's probably better to split such a tourn in two or three pools. Something for Infinity/Arno to decide, next time. A 'themed' freestyle tourn with openings restrictions can be an idea ofcourse, as an alternative, justl ike chess960 is an alternative (chess variant).

Some ideas are in the field of scoring systems, a win with Black could/should be rewarded higher than a win with White

Also we can think of some slight endgame rule modifcations, and some, eg with different stalemate scoring can be experimented with

These are the names I recognise, from the ICCF top ratings, as Freestyle players:

Code: Select all

#  ICCF ID	Count Titl Name               Games Rating
4	 81027	GER	GM	Krabben, Matthias  210	2665
7   241018	ITA	GM	Riccio, Eros       425	2644
16   81282	GER	GM	Nickel, Arno       395	2612

The first one (a.k.a. Stabilobss) never amounted to anything, Eros had his moment of glory (but has since fallen quite behind) and only Arno, remains a competitive freestyler. In essence, I’m not so sure about Ron having faced the kind of opposition he’d find in Freestyle, specially considering that he’s not used to the format. You may’ve been able to win a couple of games, against not so well prepared opponents, but that’s it. The kind of chess entities that should take part, in a “Freestyle World League”, would lose 1 game out of 50 (give or take). BTW, Matthias Krabben was the last one to beat Ron, unless I missed something, and just glancing over the moves, I can tell you I’d sign for black to play that against me, in a Freestyle event, instead of the much proven semi-slav. Under the proper conditions, d4 is still good!

Having, at least, two separate categories, is in my opinion a no-brainer, and Arno pretty much admitted to the necessity of such a measure in the tournament chat. Whether he enforces that decision, or postpones it once again (scroll down), is yet to be seen.

I can’t accept your analogy, between chess960 and “chess minus X openings”. The latter is still chess, slightly different from a gameplay perspective (basically when it comes to preparation), but as a problem... totally unchanged. Viewed from that angle, taking the solved lines out of the equation, is the only thing that makes sense; if you want to solve chess, you don’t waste more time in what’s already been solved. We have to move on and focus on the rest of the positions. There’s quite a few of them left, once you remove that bottleneck.

A change in the scoring system won’t have much effect (if at all), as long as everything else remains the same. The stalemate was already tweaked with in the past, and other experiments which have proved to be good in OTB chess, fail when applied to Freestyle (look at the bottom of this report, for a glimpse of what the "Bilbao" scoring system would amount to).

Someone could specify a set of selection criteria ensuring ECO variety, draw-rate, human vs. engine popularity, move-length, etc. and come up with however many unique positions that met all requirements. [..] These positions could be kept secret from the players until right before the game started, or the day before. Allotting one day to prep would offer the analysis kings ample time to prepare for the game correspondence-style and possibly spell doom for those who couldn't devote sufficient time and resources to preparation.

Alternatively, the set of positions could be made public in advance, if it were large enough to prevent people from preparing them, up to the point where draws could become a problem again. Then before each round, a transparent lottery could determine the opening for the day. To minimise randomness, players would need to play both sides of the opening against the same opponent, preferably without pause, to ensure that the number of people, who would just follow on the best players’ footsteps, is kept in check.

"Many show interest in what is to expect from 8-man endings. First, take note that the longest 6-man mate took 262 moves (KRN-KNN). Moving to 7-man endings doubled this value. Second, 8-man tablebases include much more endings with both sides having relatively equal strength. All this gives us a strong hope to discover a mate in more than 1000 moves in one of 8-man endgames. Unfortunately the size of 8-man tablebases will be 100 times larger than the size of 7-man tablebases. To fully compute them, one will need about 10 PB (10,000 TB) of disk space and 50 TB of RAM. Only the top 10 supercomputers can solve the 8-man problem in 2014. The first 1000-move mate is unlikely to be found until 2020 when a part of a TOP100 supercomputer may be allowed to be used for solving this task."

A 1000-move mate (or more) would be of little practical use, unless FIDE changes the 50-move rule (which they won’t). It may seem contradictory that I talk about solving the problem that chess is, while ignoring the purely theoretical approach, but as I explained somewhere else, to Bob, long before irrefutable evidence were shown, about chess being a win or a draw, people in the computer chess community would've lost interest. Once engines start being released with 10 ELO improvements, and draw rates approach 100%, who’s going to follow development anymore?

[Leo]

Maybe the problem is because most of the contestants are using stockfish and its like stockfish playing itself. Of course it will be hard to get a win. The last Thoresen Computer Championship (Played on a high end Intel Xeon V4 44 core machine) result was 13 wins for Stockfish, 7 wins for houdini and 80 draws. I am personally very OK with 20 wins in 100 games.

[Ozymandias]

Maybe the problem is because most of the contestants are using stockfish and its like stockfish playing itself. Of course it will be hard to get a win. The last Thoresen Computer Championship (Played on a high end Intel Xeon V4 44 core machine) result was 13 wins for Stockfish, 7 wins for houdini and 80 draws. I am personally very OK with 20 wins in 100 games.

I see the number of draws as a result of combining deep and wide books with SF, more than HW having that much to do with it. My PC was a slightly OCed 5820K, and performed between striking distance of the many core servers.
As for an 80% draw rate being acceptable, of course it is, as long as you don't include weaker opponents in the final, something that doesn't happen at TCEC, ICCF and was promised would not happen:

The best 8 centaurs according to the final Grand Prix table will be nominated for the next official Freestyle event by InfinityChess in 2016/2017.
(Regarding engine players we will of course also consider all participants' performances in the CWT, but also their achievements in the Engine Masters tournaments.)

[Leo]

Maybe the problem is because most of the contestants are using stockfish and its like stockfish playing itself. Of course it will be hard to get a win. The last Thoresen Computer Championship (Played on a high end Intel Xeon V4 44 core machine) result was 13 wins for Stockfish, 7 wins for houdini and 80 draws. I am personally very OK with 20 wins in 100 games.

I see the number of draws as a result of combining deep and wide books with SF, more than HW having that much to do with it. My PC was a slightly OCed 5820K, and performed between striking distance of the many core servers.
As for an 80% draw rate being acceptable, of course it is, as long as you don't include weaker opponents in the final, something that doesn't happen at TCEC, ICCF and was promised would not happen:

The best 8 centaurs according to the final Grand Prix table will be nominated for the next official Freestyle event by InfinityChess in 2016/2017.
(Regarding engine players we will of course also consider all participants' performances in the CWT, but also their achievements in the Engine Masters tournaments.)

I have learned from this site that less fast cores is better than more slow cores. Its a lot less expensive also. It looks like when I someday buy my favorite rig it will cost about $2500 compared to $5,000 plus.

[Ozymandias]

The last Thoresen Computer Championship (Played on a high end Intel Xeon V4 44 core machine) result was 13 wins for Stockfish, 7 wins for houdini and 80 draws.

I have learned from this site that less fast cores is better than more slow cores. Its a lot less expensive also. It looks like when I someday buy my favourite rig it will cost about $2500 compared to $5,000 plus.

What was the reference to TCEC's HW for, if you already knew that high core count isn't a determining factor?

[Leo]

The last Thoresen Computer Championship (Played on a high end Intel Xeon V4 44 core machine) result was 13 wins for Stockfish, 7 wins for houdini and 80 draws.

I have learned from this site that less fast cores is better than more slow cores. Its a lot less expensive also. It looks like when I someday buy my favourite rig it will cost about $2500 compared to $5,000 plus.

What was the reference to TCEC's HW for, if you already knew that high core count isn't a determining factor?

I meant that a high quality tournament showed that 20/100 games were not draws. It was my way of emphasizing that they spent a lot of money to put the tournament on. At least 16 of the cores we would assume were really calculating.

[Ozymandias]

There is a forum here http://infinitychess.freeforums.org/portal.php

When I mentioned IronMan's performance, as one of the worst six in the whole tournament, in my first post, I meant to say that excluding those players wasn't an arbitrary decision. He wasn't on the initial list provided, because he only played a few games, but even with a few games, when you're so far behind the rest, it's safe to say you're comparatively weak. Imagine my surprise then, when the following conversation ensued with Kevin Plant:



It looks like my point didn't get across, and some people still think that anything that counts are results, despite the circumstances involved. Ok, if what I (and others) have been saying all along, isn't enough, let me resort to an extreme example:

-You play a tournament and you win every single game on time, because your opponent forgets to punch the clock, gets asleep, goes out smoking, dies,,, whatever. Now, you play horribly, but still you end with a perfect score, are results what matter?

Add to that, Xeonrocks' comment, and it becomes clear that there's still a significant portion of the players, who think this is business as usual. It isn't, when the IC team decides to include a weak bot in a swiss tournament, it introduces randomness. This, among other things which have been purposedly left aside by Arno, shouldn't happen. I have no problem with a bot like the one playing yesterday, entering the ICUC, but that's not what happened back then.

There was also a last comment on his performance yesterday, saying that it was "impossible". To such claim I have to say, that I inspected the games and his initial 3 out of 5 was legit. Those things can happen.

[Lyudmil Tsvetkov]

Using a blacklist wouldn't be less arbitrary than having a whitelist / set of starting positions. If you vote for the Marshall, a very next candidate is the Berlin Wall. Whats next? Petroff? French exchange? Semi-Slav-Moskau-System, ... And how to avoid move transpositions finding it way to the blacklisted positions directly or some moves afterwards? Defining openings for a whitelist seems easier, but still is hard enough. Especially when you try to find positions not only with low draw rates, but also balanced chances. Securing balanced chances would be easier by playing the position twice again each opponent with switched colours. Having a set of positions large enough to make preparation very, very effortful, and using a kind of randomness for the selection... for each game induvidually or for all games in the same round. Just brainstorming.

You raise some valid points. As long as a we don't have 27-men pieces available, there's no way to guarantee, that the position after 11... c6, is a draw, so yes, to a certain extent, saying that this position is solved, is arbitrary. But focusing on the game aspect of chess, I think it's quite safe to say so. More than 3 years ago, in the ICFB, it was already being chosen as the favourite weapon of choice for black, and I remember my last game against Alvin, when I also used it. Even though I played poorly against his a3 anti-Marshall, a draw was achieved, and that was back then, with much less theory on any of the various anti-Marshall attempts, with weaker engines and weaker HW. Believe me, if someone found a way to crack that nut, I'd be popping champagne, but I don't see anyone trying anymore. The other defences you mention, may also prove to be impregnable, but they'd need a longer track record.

If we look trough the history of engine-driven opening theory, we see a trend. Back in the golden days, everything was about the poisoned pawn. (Curiously enough, the fact that most of the games revolved around this position, was the very reason why, the chess professionals scoffed at Playchess engine games, as a serious source for their databases.) But after a long fight, 6 Bg5 was found to be a draw, so people turned their heads towards the English Attack, and that lasted quite some time, although not nearly as long. When 8 ...h5 seemed to improve black chances, people then sifted their focus towards the Queen's Gambit, (more or less around the same time that the Marshall was also "refuting" e4 Nf3). Strides were made in the Anti-Meran Gambit, specially with the Botvirnnik Variation, in a relatively short period of time. After that, the only thing worth mentioning is the Giuoco Pianissimo, which we can also agree is a draw, if played correctly.

And where does this all leave us? In a barren land where white desperately tries to find an edge where none exists. As with the Marshall, if someone finds a completely novel way of making the initial tempo worth a damn, that bottle of champagne is waiting.

Perhaps the thematic opening(s) could be randomly drawn from a pool of such pre-selected lines.

If you can propose a transparent, tamper-free random system, I'm up for it.

who says the Marshall is a draw?

SF gets big advantages in it for white, 70-80cps and more.

with perfect play might still be a draw, but in general white enjoys quite some edge.

it is not by chance Capablanca easily won the first game on the line.

[Thomas A. Anderson]

My first thought about your point with ironman was, that it probably shuffles up "only" the bottom part of the rankings. But I'm the one who should have known it better, because these results possibly have cost me some thousands of dollars, leaving me on the 3rd place only by a tie-break-system, that is influenced by any game played in the tournament. But the fact that my first-round result against a player that quits the tour immediatly afterwards has been counted for my SB-Rating with 0 points (despite my veto) would have been a knock-out for me in any case. At least one future thing became crystal clear: Nowadays tournaments become such close, that we need to pay much more attention to having resonable (and FIDE-conform) Tie-Breakers, squeezing out as much randomness as possible.

[Lyudmil Tsvetkov]

indeed, what I am seeing is at least 90% of Marshall lines are won for white(after Re5 Bd6 Re1 Qh4 g3 Qh3 Be3).

so, the Marshall might pretty well be busted.

of course, with the remaining 10% of lines and perfect play the position might still be a draw, but who knows how all 100% of chess lines end?