Hi guys, I'm Loki!
You might know me from my other chess thread: http://forumserver.twoplustwo.com/14...l-log-1510701/

I'm currently busy with uni and work, so I've taken a break from competitive chess (not sure how long). I have time, however, to play competitive correspondence chess. ICCF is the official chess league for correspondence and is recognised by FIDE.


About me

I'm a young student in my 20s. I played chess for a while now. I've also enjoy studying the game, whether it is for tournaments or just the fun of learning new things. I'm currently a CM (Candidate Master). I'm make my income to pay for uni etc. through chess coaching and poker.


Correspondence Chess

In correspondence chess, at least the ICCF, ANYTHING GOES! You can use engines, database, tablebases. You aren't allowed to consult other human players (this is kinda hard to enforce).

Obviously, if this was using engines is all it takes, correspondence chess would be just a battle of the computers (which some already argued that it is). But yet there are ICCF IMs and ICCF GMs, so a human skill factor is existent.

Quoting a soon-to-be GM and opening speciallist "Based on my study of correspondence games, I would say that your best winning chances will come from systems such as the Anti-Meran or Meran Semi-Slav, Main Line Slav with 6.Ne5 or Slow Slav, 3.e4 against QGA, 4.f3 against Nimzo or 4.g3 against Queen's Indian, and Classical KID combined with the Russian Variation of the Grunfeld, provided of course that you know all these lines very thoroughly

For 1.e4, I'd recommend absolute main line Open Sicilians (so 6.Bg5 against the Najdorf, English Attack against the Taimanov, etc), the Italian Game or Scotch (don't waste time trying to refute the Berlin Wall), and probably 3.Nc3 against the French and the Short Variation of the Caro-Kann"

This suggests that there are positions and lines where the computer doesn't really understand how to play and human input is required.


My plan

A correspondence game may last over a year. So it's better if I play multiple tournaments/games at the same time.
I'm playing the world cup next and see how I'll go.

I will use two engines: Stockfish and Komodo, I believe they can compliment each other, and if if one can't analyse a certain position well, the other can come in.

I have a 4 core processor, this is abit weak since I might be going up against fanatics who use 16-cores or 32-cores processors which are just alot stronger and can go deeper in depth. Might borrow a 8-core using the cloud engine system on chessbase.

I have Mega Corr 2015 database and all the most recent ICCF games in my database, which might catch the weaker people off guard (those who only use engines). Also having a good repertoire understanding will help and compliment use of database.

I have access to endgame tablebase because you need them for endgames.

I will use my own chess knowledge to bolster and suggest moves to the computer based on my understanding of the position. Just because an engine suggest a move as 1st, doesn't mean it's the best move..., once you show it other ideas or let it run at greater depth, it might change it's mind. Hence it's better to go deep into a position and see where the flaws on an engine is, and possibly exploit that. This article is one of many that examines the flaws of chess engines and why they aren't "perfect" yet: http://www.chessdom.com/im-erik-kisl...inal-in-depth/

I don't have a specific goal in mind in terms of rating or getting ICCF IM/GM (which doesn't compare to an actual OTB title).

I basically just want to play correspondence chess to have some fun and also improve my chess understanding as I go... through deep analysis and theoretical study.

Godspeed! But how much chess income do ICCF GMs (who're not OTB IMs/GMs) earn?

Congrats on becoming a CM (albeit not by means of the rating ).
Curiously, ICCF even allows to claim a win by reference to the 6-men Nalimov TB. That would be convenient on chess.com or Lichess, but they allow neither tablebases nor engine assistance (which is even harder to enforce but they somehow manage). Opening databases are however not frowned upon, and it makes CC quite distinct already.

I've almost wrapped up my CC activity because it's a bit tilting to have slowly moving positions, especially difficult ones, when everything in the real life is changing fast, poker takes time (should ideally be taking much more) and I'm not sure how busy I'll be in the next few months.

That you have the 7-men TB installed is really cool, I'm going to ask you to check a K+R+B+P vs K+R+P position from my last remaining CC game when it's finished, if you don't mind It's quite clear to me even without an engine, though
Speaking of, what do you think about 3. f3 vs 2... g6?

That however doesn't interest me that much because I normally play 1. Nf3, thus I have no Samisch KID option vs 1... d6 / 1... g6. I guess you prefer the Bayonet attack with White in the classical KID (1.c4 Nf6 2.d4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 O-O 6.Be2 e5 7.O-O Nc6 8.d5 Ne7 9.b4); if not, what's your 9th move there?

What's your opinion on the early h3 KID?

Not sure, but they might be used as seconds for GMs.

Thanks, I'm abit stronger than 2k, still need to be consistent to play at higher level (2.2k+) and learn my lines better

My friend have one atm. I still don't have mine installed. But I'll get it installed before the world cup prelims start.

To be perfectly honest, I don't understand KID well enough to give a perfect assessment of the lines against it. In OTB chess, I play a very non-critical line against it in the English.

The Samisch option seems promising, although I feel like the engine might do fine finding a plan in those types of position. The h3 option was recommended to me by a coach as a practical OTB choice, but I'm not sure if it's critical.

The main line Bayonet attack definitely is a weapon I'll use. I think the engine might misplay or misunderstand the plan and the dynamics in the position as black. White seems to have a very straightforward plan on the queenside and a not well prepared black might run into trouble. This is something I really have to work further on - is how engines handle positions and why they might not understand it, even if insane deep depth in their analysis.

How have your approach ICCF in the past, other then just referring to your database, engines and tablebases.? And may I ask how well are you doing in ICCF? It's definitely a challenge for me to work out how to beat other people who are also using the same resources as us (databases, engines, tablebases) - so what do you have to do to score wins :P

Sorry, I've never played ICCF games (and I don't even have a FIDE rating, I guess it would be 1600-1700, i.e. I'm a club level patzer); what I played were chess.com games, where no engines or tablebases are allowed (but opening databases and manual analysis on a board are OK).

I'm afraid these formats are too different. The chess.com games are more about how deep you can (and have time to) calculate manually, thus choosing offbeat tactical openings and then just playing good chess and dedicating enough time to analysis works decently.

But if I were facing someone equipped with an engine, I'd avoid positions where tactical tricks are probable because engines see them perfectly. The way to outplay the engine is, like you've said with regard to Bayonet, reach positions that have longterm strategic plans that engines are unable to understand.

The advice about offbeat openings (to throw engines out of their books) still stands, but in ICCF games, they need to lead to quiet closed positions instead of sharp open ones. The Stonewall formation (with both colours) and the London are thought of as good examples. (With White, I'd prefer to develop the DSB to g5 or f4, though, before playing e3, so it would be more like Trompowsky Stonewall.)

Likewise, in the endgame, engines grossly overestimate the chances of the stronger side when the defender can build a fortress, so definitely get apter at recognising them. Engines generally fail at complex endgames (those with >7 pieces), so transitioning to the endgame early (if you end up in an open position) looks a good idea.

In general, I'd be looking for opening lines that lead to decent positions according to the manual analysis of the practical games in the line (there are few enough of them if the line is offbeat), but that are undervalued by engines.

A good thing about the way DBs work is that the total winrate or performance rating of a move that is shown to your opponent is the average over all your possible responses, but it might be the case that the most popular responses are bad for you but you have one response that others play very seldom but that is really great. In particular, you may abuse transpositions and manipulate the opponent into eventually walking into a line that's favourable to you.

I struggle to find a concrete example of such an approach that would suit your repertoire. What comes to mind, though it's not really a closed opening (but looks rather positional at first glance), is the Neo-Indian (Seirawan) attack - 1. c4 Nf6 2. d4 e6 3. Bg5. (One of the points is 3... Bb4+ 4. Nd2 - the knight isn't blocking the bishop like it would in the Bogo. Yasser Seirawan is a specialist in anti-computer play, fwiw.) Its benefit is that it often transposes to the QGD or leads to Trompowsky-like offbeat lines, which is a nuisance to a Black player who's better prepared for the Nimzo or Bogo Indian.

The game may continue 3... Be7 4. Nc3 (Black will be out of book unless he transposes to QGD) d5 5. a3!? - that looks like a solid deviation, and of course it's not a total waste of time, as it prepares b4. (The thing is that e3, a3, Rc1, Qc2, in any order, seems to be often played later anyway in the Orthodox QGD; but here I'd like to leave the opponent with an uneasy feeling of being thrown out of book if he doesn't know the ins and outs of the QGD; neither do I, it's just an example of the power of transposition).

There are a ton of sources on anti-engine play, here are two articles by Rafael Leitão (both an ICCF and FIDE GM) that I like (you've probably read them already because I mentioned them in a similar post in January that is about my approach to chess.com CC and not fully applicable to ICCF, but I want to make sure you don't miss them):

Ideas on how to prepare a good opening repertoire

Modern day analysis – working with the computer
___________________________________

I've looked up ICCF's site and was a bit shocked by the fact that the World Cup is going to take 6 years to be finished... That's a ton of time. Do you know what your life would look like in 5-6 years? I don't. I need to get resettled first to remotely consider committing to such a long tourney

As for the 7-piece tablebases, don't worry: apparently ChessOK, despite charging $20 lifetime for 7-piece TB use from those who have neither Aquarium nor Chess Assistant, has a free Android app that can query them (of course they're too big to be installed on a PC anyway - 140 terabytes). Obviously, Android can be emulated on Windows via BlueStacks (I haven't tried to run ChessOK there, though; I was able to launch the Ladbrokes poker app back in 2014, though it was painfully slow). 6-piece TBs are free to query in Windows browsers as well.

@coon74 Yes, you're right that while anti-engine play is good approach to chess.com CC, it is probably not applicable to ICCF.

I think the recent So vs Nakamura game in Sinquefield Cup 2015 shows how engines might evaluate a position poorly due to lack of understanding in the position - especially in KID lines.

http://www.chessgames.com/perl/chessgame?gid=1800188

Quite the beautiful sequence...

Maybe in correspondence chess but certainly not over the board!

1. "Anti-Meran" is the Botvinnik/Moscow-complex I assume. There is no chance that you can solve such a complex position over the board without knowing something like 20 subvariations at least 30 moves deep. Take a look at Schandorff's books and you get the idea.

2. Mainline Slav with Ne5 can lead to a piece sacrifice for 3 pawns where you need a very high technical level to convert -> engines show 0.00.

3. The f3-Nimzo is just pure crap. In the mainline you can try to hang on to an extra pawn that can't be converted and I don't even have to mention how So gut crushed by Aronian. White simply has zero development.

4. The classical KID is either a draw in the 9. b4-line (Grischuk, Radjabov and Bacrot will agree with me on this) or you get mated in the other variations more often than not (ask So and Gelfand how it felt to play Nakamura).

5. The g3-QID may transpose to a Stonewall-structure in the line with Ba6 and f5 at a later stage. Very stable position where it is tough to press. Maybe the version with Qa4 is a better attempt, but I wouldn't be sure.

6. The Qb3-variation against the Grünfeld may or may not look attractive against the Hungarian-Variation with a6, but the real practical reason you shouldn't play it is the Prins-Variation with Na6. You have to play a position with an isolani which forces YOU to prove that it is a strength and not a weakness.

That 'engines show 0.00' and that the lines are deep and technically difficult is, on the contrary, a strong argument for playing them in ICCF games, where both you and the opponent are equipped by engines: computers suffer from the horizon effect (though it's less applicable to modern ones) and poorly understand positional factors, such as the pros and cons of an isolani, so the opponent can be misled by the engine and walk into a line where he has underestimated longterm weaknesses.

Plus, the Botvinnik and Moscow variations must be a lot of fun to play for both sides

Here is a funny pair of games that show why you don't really want to go into this line:

http://www.chessgames.com/perl/chessgame?gid=1011460
http://www.chessgames.com/perl/chessgame?gid=1292546

Even if white has a path to an advantage in theory, the position is too difficult to play over the board. It is one of these positions where your opponent deviates and you know you are supposed to punish him, but you got no clue how to continue.

The world cup starts in october and I'll be beginning to prepare for it.

In danger of getting timed out on a bunch of my games LOL. Better analyse. gg

I'm kinda regretting ever playing ICCF :P and by the looks of it this will be mostly a battle of the hardwares. I played for d4 positions in all my white games in hope of generating some strategically rich position and will go through games in greater detail tomorrow.

With engines allowed I'd be surprised if it was anything other than a "battle of the hardwares", unless a strong GM was involved. Even then I wonder how often a GM's analysis can improve on what the engines can find after churning away for hours on very fast computers. They'd have to very carefully channel the games into positions the computers can't handle all that well, and that set of positions keeps shrinking as the hardware improves.

I know that when I give Stockfish a position in the opening (with no opening book), it almost always finds either the main line or a sideline that just happens to score surprisingly well on the rare occasions when it is played, if I let it run long enough. Openings being positions that have been subjected to intense analysis by GMs of course, as well as lots of empirical testing, and Stockfish doesn't seem to miss much even without knowing any opening theory.

I have 5 games in play atm, I won 1 and timed out in the other 4 :P

Will probably quit ICCF after these games are over, just don't think it's worth my time.

I did have an interesting game in the Kozul Suicide line vs a SIM

The full game is here:http://www.chessvideos.tv/chess-game....php?id=105729

I was black

This was the position before I timed out the plans available to both side in the endgame is interesting and perhaps worth looking into...

I had a look at the latest ICCF World Championship results. 85 games still in progress, 51 finished. The draw rate for those 51 finished games is:


https://www.iccf.com/event?id=52852

Yea I quit ICCF already, its horrible. You can't win a game. I accidently timed out on a couple of my games already and it's a grind recovering from that rating wise (how many games I have to grind to make it back up). So basically not worth the time and effort.

I'm back to playing real OTB chess where I'm a titled player aiming to go futher

With 54 games in the World Championship now completed, Jacek Oskulski disagrees. He won a game.

^ All hail the new champ Jacek!

ICCF chess is a scam, half teh so called correspondence GMs are like club player strength, its pretty ridiculous.

It's just a way for rich people who can afford advanced software to add more cred to their name.