Tuesday, 05 June 2012

What is the best way to build a chess opening repertoire?

There are many ways to build a chess opening repertoire. I want to share with you my personal approach and I’m looking forward to hear from you how you build your opening repertoire.

In a nutshell my approach looks like this:

  1. Finding some openings for white and black I like in general (= fits my playing style). I’m doing this either by searching opening databases online, going to a bookstore and browsing through some opening books or asking some other chess players.
  2. Looking for chess books covering these openings and buying them (making sure I cover all potential systems which my opponent could play).
  3. Reading through these books and figuring out which middle game positions I like.
  4. Checking which move orders enforce these middle games. This is actually the most time consuming and complicated part for me, but it is also quite interesting and satisfying.
  5. Entering all candidate moves into Chess Position Trainer (that means I only enter one candidate move for my side and all potential replies covered by the book for the opponent side).
  6. Start training my opening(s) which helps to remember what I just learned.
  7. Going online and playing some blitz games to practice once I feel confident.
  8. Running these blitz games against my repertoire and checking where I or my opponent deviated from my repertoire.
  9. Unless I just forgot my candidate move I will add the opponents move and research for a good reply. At this point I start using a chess engine for the first time in the process.
  10. Playing my new opening over the board.
  11. Analyzing my game by running it against my repertoire and using a chess engine.

That’s more or less how I built my own repertoire (and still extend it). As you can see I’m hardly using any chess engines for a long time in the process and actually no game database at all. I must be an exotic! However, this approach helped me to build a repertoire which let me gain an ELO of >2,000 (transforming my German DWZ into ELO) and it would be even higher if I would practice and learn more chess than developing chess software (not to mention my non-existing endgame knowledge). I can hear you saying selflessly that the current rating is OK!  :)

Game databases matter?

Like I said I’m not really using game databases to build my repertoire today. I don’t claim they are not important. Indeed I use game databases for game preparation and there are many scenarios where I can see a clear benefit (e.g. creating a repertoire of a GM playing my opening or getting ideas for middle game positions etc.). However, I wonder how much they matter for creating an opening repertoire for the average chess player with limited time.

Speeding up the process

You can save some time and money if you buy repertoire books. In this case the author made most of the work for you and selected the most promising lines instead of covering all moves for both sides for each position. That’s a pretty reasonable approach for non-professional chess players. You can substitute specific opening systems by something else later anyway, but you quickly build a solid repertoire.

The process of entering all candidate moves is quite time consuming I have to admit, but it also ensures you really go through all variations. However, since chess publishers started offering eBooks which come with a PGN file you can also just buy these and save time.

It would be even better if chess publishers would offer eBooks in a CPT native database format. Let them know to increase the probability!

Another approach: importing games

Recently I have received several emails from chess players who wanted to import something like 10,000 games of a specific opening, but they had problems as the performance was going down during the import. What’s your guess how many unique positions can be reached by “just” 10,000 games? Well, I couldn’t believe it myself, but we are talking about 1,000,000 unique (!) positions. And that is just for one opening (e.g. French with White)!

The first thing I recommend is to limit the import to just the first 15 to 20 moves. However, we are still talking about several hundred thousand unique positions then.

Now, I could improve the import performance, but then again I wonder how this approach can be efficient anyway. I mean, somehow you have to go through all positions and decide for each position which move should be your candidate move (and delete the others). On which information are you going to make this decision anyway? Number of played occurrences? Chess engine analysis?

I would prefer the analysis and evaluation by a GM.

However, importing games can be a practical approach if you do it in a more sensible way. In the past I’m across a nice video by Katar demonstrating this process with an older version of Chess Position Trainer. Although Chess Position Trainer 4 changed quite a bit, the same principles can be applied:

Chess Video by Katar about Chess Position Trainer 3.3

I really like this video and I’ve learned quite a few things myself. I plan to create some videos about Chess Position Trainer 4 too. So, any suggestions for topics are welcome!

To get back to the original topic: how do you create your own chess opening repertoire?

Social Bookmarks

Comments (2)

  • Johan
    06 June 2012 at 17:48 |

    Yes, you do need to make a video about this software PLEASE. I can not believe that when I go to youtube, there is no video on your product. Please take the users at least through the first few steps of using your product, teach them how they can make different repertoires, how to train, etc. Please go through the most commonly used functions in your program. I know there is a manual, but it helps people like myself to watch a video first, see how someone else does it. Thank you so much.

  • Dave
    24 June 2012 at 17:59 |

    I am probably better qualified, if anything, to say what not to do. I found the article very good and have learned some useful ideas, particularly points 3) and 4); however this is a little ahead of my capabilities at the moment! I have used the options of importing from e-books and selected players' games. This serves a good purpose and you do learn from it, however, at my level (Blitz 1200) I find that the larger the repertoire files the less productive and more frustrated you become particularly as you get trained in lines you do not really understand; nor want to. It is great for learning your favoured line from the Caro Kann, but alas when at the chessboard your opponent plays anything but...

    So I have now created much smaller repertoire with simpler lines (with some expert help). This is very much work in progress but my idea is to know my smaller repertoire very well and know that I will reach a comfortable middlegame. I have tried to get a broad range of openings with less depth, rather than a narrow range of openings with more depth.

    Creating the smaller repertoire was hard. My starting point was my original large repertoire(?) and my own games database set as the reference database on Chessbase. Now I could see how much of my large repertoire was never occurring in my games and thus having it there was counter productive. The painful bit was removing the lines from the repertoire, but I knew it made sense. Additionally where I had a number of opponent replies I used my reference database to delete the less common moves. This was an iterative process until I felt I had a manageable repertoire. By exporting the games to PGN files so that each line was a unique game I could see how many games there were and decided to made sure that each opening consisted of no more that 80 unique games; in some cases an opening consisted of just three unique games! At most I have three opponent replies and always only one move for myself in any given position. So as an example my White d4 repertoire consists of 14 named openings with anywhere from three to eighty unique game lines, with 15 to 20 moves each.

    Now I train on this smaller repertoire and can see my progress, or otherwise! It could be some time. There is even a chance that with the smaller repertoire that the "photo reading" training will make more sense and I might even recognise one of the positions!

    Hooking up a DGT Board in the training mode makes you feel like you a playing a real game and often more thoughtful in your response. I can train on any one of the 14 openings. I have in the past experimented with merging all the openings into one, but again with a large repertoire this was counter productive, so I will hold off on attempting this, if at all, until I have learned the individual smaller openings.

    I am going to avoid the temptation of adding new lines until I have mastered my existing repertoire, or at least got proficient in it. My intention will be to add any new lines very selectively, based on large occurrences during my actual games; otherwise I may find myself back to a large unwieldy and unknown repertoire!

    There is probably no right or wrong way of creating your repertoire; hopefully I have learned something from my past mistakes.


Leave a comment

Please login to leave a comment. Optional login below.

CPT on Facebook


Stay up to date

Leave your e-mail and join our newsletter!