Modern Correspondence Chess from the Inside Out:

Modern Correspondence Chess from the Inside Out:

Mar 1, 2016, 4:09 AM |

What is ICCF?

The International Correspondence Chess Federation is the world governing body on officially-rated correspondence chess games. There are delegates from many countries that promote their own national or international ICCF tournaments & events. In addition to which, the ICCF itself promotes open tournaments as well as rating-based events in addition to the world championship cycle that starts ever year. Over the years, the ICCF has modernized the way correspondence chess is played. The vast majority of events are now played on a web server, where you can opt to be notified of your opponent's move via email. Then you would log into the server to play your move reply.

Now the main difference between ICCF and most other correspondence chess organizations is that the use of computer 'engine' assistance is allowed. It became apparent that cheating was going to be a major problem as chess software became more and more powerful. So rather than spend so much time and energy in a battle that wouldn't conclusively prevent all cheating, the decision was made to make engine use fully within the rules. Indeed when you review top-rated correspondence chess games and players from organizations that prohibit engine use, you will no doubt find the level of cheating is rampant with everything from deep engine analysis to the more clever sparse use of an engine only in critical moments of a game. On ICCF, the playing field becomes much more fair in this regard, and it's up to you to use whatever means you can to outsmart your opponent's engine-assisted play. As such, the quality of most games on ICCF are stronger than engines or humans by themselves. They represent the closest to perfection that can be currently achieved in chess moves on the board, but even in this arena, blunders and fundamental mistakes still happen to this day (albeit much more rarely).

Because of the incredible strength of engine-assisted correspondence chess, certain endings have become a trivial matter of just "going through the motions". Specifically 6-piece tablebase positions, where the ending is played flawlessly. The result of which eventually caused ICCF to adopt a 6-piece tablebase claim that you can make when you arrive at such an endgame. This rule was enacted to combat deliberately slow play (an opponent stalling a loss as long as they can). It became clear that the 50-move draw rule from human-only chess was obtrusive to the 6-tablebase endgame, and ICCF adopted the stance that tablebase endings overrule the 50-move draw rule. So if you reached an ending that was mate in 94, but contained 57 moves of no pawn push or material captured, you would still win using the tablebase claim. This is controversial for some in that this now differs from over-the-board rules, but in the end, it just makes finishing games more efficient and easier to adjudicate if needed.

"Correspondence chess has become just pushing buttons on a machine" 

- GM Nigel Short

Nigel's quote is the typical reaction of those looking from the outside at modern correspondence chess. Some will even react very negatively to the accomplishments of earned ICCF title holders and champions of the ICCF arena, citing a lack of over-the-board mastery of many ICCF players and believing that the level of human interaction in engine-assisted correspondence chess must be minimal as a result. The reality is nothing could be further from the truth in strongly-rated ICCF games. The skill set it takes to win against an opponent using an engine does not translate to the skills used in over-the-board chess mastery. In short, the mistake laypeople are making is comparing apples to oranges. That's not to say human chess skill is completely useless when competing on ICCF however. You in fact need a good understanding of fundamental concepts like recognizing that you might fall into drawish positions like an opposite-colored bishop ending, or spotting dynamic advantages versus being up worthless material for example. These are issues engines still do not address, and the more clever human can use to their advantage. From my experience, you need about a 1500-1700 level of chess understanding to have a chance at excelling on ICCF, but the other skills needed only come from experience on ICCF itself.

So what are these other skills? First and foremost is your ability to be creative in the opening. Engines are much too weak in the opening even by today's technological and software advances in the field. As such, the majority of wins on ICCF come from creating lasting advantages and/or complex middlegame positions that come about from clever opening preparation. This means researching hundreds of thousands of games from various archives (including even human over-the-board games) to prepare an overall opening gameplan against your opponent. For me, the most valuable tools for this phase of the game are Chessbase and Fritz software packages. I use Chessbase to collate, filter, and condense my game collection archives into a single database I do my research from. Then I use the Fritz user interface to create an opening book comprised of the first 60 moves from that database. Then it becomes a simple matter of reviewing the lines of the opening I want to play and spotting potential weak points or critical positions that might trip my opponent up. I love this aspect of ICCF play, because it opens my mind to openings I had never considered in my over-the-board play, and gives me a greater understanding after so many years of research. For example, ICCF play gave me a profound appreciation for the Sveshnikov Sicilian. I've played and researched it so many times from the black side that I can recite all the major variations and describe the gameplan for each side without needing to consult a book or database. This is an opening I never entertained the idea of playing, much less understanding the point of before I joined ICCF.

The next skill comes in how you handle your interaction with engine analysis. It's been said before about ICCF that you get out of it exactly what you put into it. So if all you do is just plug the game position into the engine and then play the top move it picks, you're going to hit a rating wall and wonder why other people keep climbing the ratings ladder while you seem to stagnate. There are in fact players that only let the engine do all the work, and I've played and beaten many of them. This is where the "apples" ICCF skill comes into play that doesn't translate to over-the-board's "orange" skill set. I'll describe a few of my own techniques to give a sample of what I mean:

1. I keep a running diary of every game I play. This includes notes about the opening, the general impression I get from the current position (whether or not I agree with the computer's evaluation), and my overall prospects and goals for the game. This for me is important because I typically have 30 to 40 games running at the same time, and it helps me to refresh my memory on my intentions instead of just blindly plugging the game into an engine.

The opening, impressions, and goals I set come from experience on ICCF, and those concepts can only be honed from ICCF play. This is why a 1500-rated player can still have the potential to win against a 2600 GM when they play on ICCF, because that 1500 might just have several years of experience on ICCF versus the GM trying it out for the first time. I myself have defeated FIDE masters on ICCF using this same "apples to oranges" concept of experience and skill. Often times a FIDE master will want to play their pet opening from over-the-board play, not realizing it may contain deadly traps that are 20 moves deep and sacrifice two pawns for the initiative. This is stuff the engines cannot see coming, and thus, the FIDE master doesn't see it either. However, more and more FIDE masters are taking interest in the ICCF, and some of them are very quick to adopt to the arena and realize that it's never a good idea to base your opening preparation on what you did over-the-board. You have to variate and be creative, or your opponent might just have a nasty surprise waiting for you in the middlegame.

2. Subtractive analysis. This means using the engine to explore either your own ideas or alternative ideas the engine normally might skip over due to its selective search pruning function. What you do is forcefully remove the engine's top choice and have it evaluate from the remaining set of legal moves. As long as time permits, you can keep doing this and take notes of the evaluations until you start reaching moves that are clearly weak. This gives a set of candidate moves to really start your analysis from. Remember that an engine doesn't do this on its own, so it's still up to you to create the list of potential moves to consider playing.

3. Backsliding analysis. This allows you to go well above and beyond the engine's horizon and really find out if the position is going anywhere meaningful. What you do is move forward several moves in the given line an engine predicts as best, and then work your way backwards up the line and check for alternative moves the engine may have missed as being an improvement. You really have to get in there and move pieces around. Ask yourself "why doesn't this work?" or "what about this move?". Believe it or not, you can sometimes come up with a brilliant line that the engine had no clue was even a viable option. This is an essential technique to beating opponents that simply let the engine play the top choice move. It also can save your skin in a game that you are being slowly crushed in. I had a recent game where my opponent was very creative in the opening, and I found myself in that horrible situation of getting squeezed off the board. The computer engine couldn't figure out what move to play that would save the situation, so I employed my backsliding technique on several candidate moves, and uncovered a brilliant game-saving draw that sacrificed a pawn twenty moves deep (40 ply). I input the entire line as a conditional string, and my opponent followed it all the way out to the draw because his engine couldn't resist the material evaluation making it look advantageous for him. These kinds of draws are almost as satisfying as a win for me. It demonstrates technique the computer engine could not find on its own, but the human-engine combination did.

The real problems ICCF faces in modern times:

As you climb the ICCF ranks, your opponents become stronger and stronger. Not because of hardware power, but because they have more experience on how to specialize in opening preparation. They got to where they were NOT by blindly letting their engines play, but by becoming experts in openings/middelgames that give the most chances for incredibly complex positions that allow for more human guidance of the engine. Now what happens in these world champion finals is everyone by that point is extremely well-versed in opening theory/research, so it becomes extremely difficult to catch them in a new creative line that offers a complex difficult position to analyze with human-engine technique. As such, the vast majority of games reach a middlegame position that black (or even some rare cases white) can hold the draw with patient and deep analysis on each move.

As I've crossed 2500, I've come to realize that my opponents are so competent in the opening now that I can no longer play double-edged openings as black in order to try and win from that side. The opponents simply have too much experience and knowledge of how to avoid falling into a trap as white. Instead, I now have to use a new repertoire based on preventing white from gaining a complex position with any meaningful advantage. An example of this would be to adopt the Berlin defense. The main line of the Berlin exchanges queens, and in the ICCF arena, it becomes virtually impossible for white to create the kind of tension needed press a win from a micro-mistake black might happen to make. The engines are strong enough now that a simple quad core machine can hold the draw in the Berlin even against a $100,000,000 supercomputer. There just isn't enough complexity to trip the engine up. As such, your stronger ICCF players know this, and will avoid the Berlin even if it means playing something like the Italian game.

This is the real problem right now that we face on ICCF. Once your opponent becomes skilled enough in opening theory to reach a playable middlegame, your chances of finding a win become extremely remote. In fact, ICCF World Champion Leonardo Ljubicic winning 4 games in the event is actually an incredibly impressive accomplishment. I can tell you without a doubt he had to work VERY hard in the opening and come up with some creative lines in order to even reach a potentially winning position in each of those games. That's how ICCF is played now.

I myself only play on ICCF because I love being able to outsmart my opponent in the opening in spite of them having likely powerful hardware. My computer itself is just a standard desktop from 8 years ago, yet I was able to make USA CC Champion, Senior IM, and 2500+ rating from it, and that's all because of the hard work approach I take to analyzing the game and opening research. This is why I feel ICCF accomplishments are unfairly judged when uninformed/inexperienced non-members read press releases about these events and then make overly negative or ignorant comments about it "just being a bunch of engine versus engine play".  It's a lot harder than you think, and even the draws that are predominant now typically had a lot of hard work put into them. Unfortunately an outsider would never know this, because all they see at a glance is 1/2 - 1/2 on the crosstable report.

With that all said, the draw issue has made advancing to the upper title ranks much more difficult than it used to be in the past. In 2015, there were only 3 people in the world that made the rank of GM on ICCF. The ICCF staff are looking into this issue to see about changing the requirements and/or awarding GM titles for winning higher-rated events instead of the title being based on purely a rating performance. I will point out that both my Senior IM norms were earned in quite the opposite situation, where the performance requirement was increased instead of decreased, making my title more a "Super Senior IM" rank as I jokingly refer to it. It's quite likely they will lower the requirement back down due to the GM issue, effectively making the window where I earned my Senior IM title a uniquely difficult time to have done so on ICCF.

With regard to the number of draws issue, you may recall ICCF GM Arno Nickel wrote an article on Chessbase detailing his proposal of changing the draw rules to award the side with more material for example. Most ICCF members disapproved of the proposed rule changes, though agreed something should be done eventually to make the game more double-edged and beyond the scope of traditional chess software. I will say that if it comes to a point where I cannot progress any further due to the draw issue, I will retire from ICCF, having already proved with my ancient computer that it isn't about spending $10,000 on a 32-core CPU beast in order to win championships and titles. It's all still about what YOU do, not about what your computer's benchmarks are.