For those that don’t know Edward Duliba has attained the title of CC Grandmaster. In US CC competition he won the 1992 Golden Knights and finished fourth in the 1994 event, has played in and won two Absolutes, and played in five U.S. correspondence championships, winning one of them.
In the 1990’s he turned to international play and in the 27th World Correspondence Chess Championship series he made the GM norm by scoring +5 -0 =6.
In an article in Chess Life he describes how play at that level. First he maintains that a really good opening database is an absolute must. Next you must play all your games using anti-computer strategy. Your aim is to survive the opening with a reasonably equal position and reach a middlegame where you just try to maintain equilibrium; don't try anything fancy.
You make your move to win in the ending because, as Duliba points out, that is where many players slack off. Correspondence success at the top level means you have to be just as eager to continue to fight after five years as you were on the first day. Thus when an opponent offers an early draw, never give it to him in the hopes that he will lose interest and eventually lose.
You should play as slowly as legally allowable in the opening and first half of the middlegame so that when players find that they are not in competition their quality of play diminishes. Players who started poorly no longer exert their best effort and points are easier to acquire from them. I might also add that in another article Duliba pointed out that this strategy will also annoy a lot of opponents and that's good. You want your opponent upset with you when he's deciding on a move because it increases the chance that like all of us, he will do something stupid when he's annoyed or angry.
As for the “anti-computer strategy” Duliba points out their strength is found in sharp well-researched openings, middlegames with aggressive, tactical play and they play perfectly in endgames of seven pieces or less due to tablebases. So you should play closed games where you don’t look for sharp lines but play to hold the position with a minimal advantage or disadvantage. Either one; it doesn't matter. Caro-Kann, Closed Ruy Lopez, and Queen’s Gambit Declined...those are good openign choices. In the middlegame, king safety is paramount. You’re not interested in positions where a single tactical slip could be fatal. Try to keep the game level and equal exchanges. Rarely should you castle on opposite sides. Computers can play some endgames perfectly but endings of more than seven pieces are the weakest element of the computer programs and this is where you should try to win.
All this requires up-to-date opening databases, the strongest engines (examining the position with two engines with different styles is recommended), the most powerful computer you can afford, and patience…lots of it because when you are playing a game where you may not hear from an opponent for a month and the game can last five years, you are going to need it. But wait! You need more. You need to be a very strong chess player yourself. You can’t just let an engine choose your moves because guys like Duliba will pick you apart.
Even if I was good enough to complete with the likes of Duliba, I couldn’t do it. Just give me a free turn-based server and a few minutes to stare at the position, or if it’s really got me flummoxed, set up a board and shift some plastic and decide on a move. A month, or maybe six weeks, for a game is plenty long enough.
I used to play postal games that often lasted a year or a year and a half (more if it was international play), but those days are gone and I just don't have the patience these days to analyze a position for hours over a period of several days.