Monday, May 9, 2011

Are chess engines ruining correspondence chess?

      Aren’t chess engines so good that humans cannot win against them? The truth is, at least at higher levels of CC play, that engines still have disadvantages. And in order to overcome those disadvantages, it still requires input from humans.
      To analyze a position to full accuracy, chess engines need to check all possible moves, all their countermoves and so on. This means their ability to look ahead and see developments is limited. To compensate for that, chess engines use different criteria to select moves which they believe are worthy of further analysis.
      For example in a middle game, where each position allows for 30 different moves, the engines would need to look at 531,440,000,000,000,000 positions to reach a depth of 12 ply (6 full moves). At an average speed of 1.2 million positions per second it would take, for all practical purposes, forever to compute. By limiting the number of moves they consider and including the different search algorithms used by different engines, they do make mistakes in evaluations.
      Another argument is that the use of chess engines allows weaker players to successfully play against stronger players so the gap between low rated and high rated players has been narrowed.
      Actually the gap between strong players using chess engines and the weak players using chess engines remains the same. Look at any GM CC tournament and you will find some players winning almost all their games while some players lose almost all of them. The result is not because of engines alone. It is because of a difference in human skills when it comes to correctly evaluating positions. The chess engines neutralize each other so the difference is based on chess skills.
      From a practical stand point let’s assume your average player lets his engine run for only a minute or two then selects its first choice. Let’s assume the stronger player lets his engine run for several hours, looks at the position and then evaluates its output and then runs though its analysis and in addition considers some moves of his own based on his understanding of the position. What's going to happen is that the stronger player will probably win. 
      When I started at LSS I lost a lot of games owing to using Fritz 5.32 and an old, slow computer. I never relied entirely on the engine to select my moves. I also tried out moves that looked best to me but did not realize I was not allowing sufficient time to get anywhere near a reliable analysis. The result was, believe it or not, a few moves that actually turned out to be tactical errors, and quite a few losses. As my knowledge of how to use engines, coupled with an upgrade in equipment, I’ve had much better results and have even bettered my overall score to a plus 2!
      Did chess engines change correspondence chess? Yes, they did, significantly so. If we look at games of the times before chess engines, we frequently find bad mistakes in the play of weaker players, including gross blunders. Today, we do not see such blunders even in the games of low-level players.
      The requirements on players have changed. Besides their chess skills and knowledge, players now need to know how to use chess engines efficiently and to complement to their own skills. Players need to know the strengths and limits of their engines. Another quirk of engines is that sometimes they do not consider moves until prompted to by the human. Also, in many cases you’ll see their evaluations fluctuating as they go deeper into the position. This means they are having a hard time finding the final solution and ultimately may require some human guidance. This is what makes correspondence chess interesting for some of us although I recognize this is not the case for everyone. In fact it has taken me a few years since returning to CC in 2004 to come to this conclusion. And, as my understanding of engines has gradually increased, my results have improved and my enjoyment of CC has returned.

3 comments:

  1. I know some chess engines have attempted to vary their evaluation depending on whether the game was in the opening, middle or end. That is the fun part of creating chess "personalities" if a particular engine allows it, like Shredder or ProDeo. Prodeo can be set to search for "pins" for example. It can be set high for positional knowledge or can be set as a tactics only engine. An intelligent player will have some sort of idea on what to look for in a position and can call up a special personality just for that; instead of always using the default personality which is a sort of "mean", but will play best overall in an engine tournament. Sometimes I will run my Shredder-Positional engine and Shredder-tactics engine side by side to see if there are any differences.

    ReplyDelete
  2. I divide correspondence chess into two groups: the fun group, where engines and tablebases are allowed (the ICCF lazy approach), and the serious one, where every player is expected to choose his/her move without external assistance of other players (computers with chess software running are other players, not analytical tools). The first group I don't consider worthy to call themselves successors of ancient pre-computer correspondence players and their titles, ratings, champions are empty numbers and phrases. The second group, unfortunately, is not taken seriously yet, because of the possibility of cheating. But I hope this is going to change soon as more effective cheat detection takes place. Finally, we will get true human XV. correspondence chess world champion, next in line after Tonu Oim.

    ReplyDelete
  3. Enjoyed your site! I started playing CC in 1959 and played through the early 1990’s when I gave it up because too many people were using engines. During those years my most memorable games were one annotated by John Collins that was published in Chess Review, a win over one of the US top 10 rated CC players in the US, two losses to a former competetor in the US closed championship and a draw with Reshevsky. If I had played those games today it would be assumed that engines were involved and they would be tainted.
    After returning to CC play in 2004 I discovered things had only gotten worse and most all opponents above 2200 were using engines so that’s why I started playing only on sites where they are allowed. The funny thing is my results on those sites are not as good as in the pre-engine days! And that’s even after I bought a new computer and upgraded from Fritz 5.2! But, even on those sites I sometimes, apparently to my detriment, play moves that are not recommended by the engines. This kind of chess is not for everybody and today’s CC is waaaay different from what it used to be.

    ReplyDelete