Organizing my openings Part 3: Repertoire Reliability

Duckfest Digest 14

Repertoire reliability

Organizing my openings Part 3

Introduction
About a year ago I decided to build an opening repertoire, for reasons and to create some structure in my learning process. Improving my opening repertoire has led to some success, as I shared earlier. But I also discovered that there is more to building a repertoire than just selecting moves. The decision making process involved more considerations on a conceptual level than I anticipated beforehand. But, that was not a real problem. If anything, it made the process more interesting. I found it far more interesting to analyze positions in depth where I had to compare lines with vastly different outcomes, than looking at positions where I simply discovered there is an obvious best move that I should play next time.

However, there was an additional layer of complexity that turned out to be more impactful than I had anticipated. As my repertoire expanded, the necessary effort to keep track of it all grew as well. Soon, my plan to simply document the moves I wanted to play in key early game positions, was no longer adequate. A more nuanced approach was needed. And when I accepted the reality that selecting moves is a more complicated process, a whole new dimension was added to the repertoire. I’m referring to the issue of ‘repertoire reliability’. Though I could also have called it ‘my Confidence in my Choices’, ‘Choice Completeness’, ‘Rating my repertoire’, ‘Managing my Moves’ or any other ‘Alliterating Alternative’.

I wanted to keep growing my repertoire, but I also wanted to make sure to do it right. This meant I wanted

• to make good, well-informed decisions
• to be thorough, making sure I didn’t miss anything
• to create a playable repertoire, not just a collection of best engine moves
• to focus on the most relevant positions and lines
• to keep track of my progress, what’s done and what still needs to be done

This sounds reasonable enough. Right? However, very subtly all these factors each added a bit more complexity to managing my repertoire. In this article I will try to illustrate how.

About this series
This series might provide other players with some practical tips on what to play, but that’s not what I’m going for. There are plenty of other sources to help you with building an opening repertoire, with an emphasis on ‘opening repertoire’. Almost all of the available content is focused on the repertoire, few on actually building it. My articles will be on the process of building an opening repertoire, with an emphasis on ‘building it’.

Chapters
Organizing my Openings part 1 - Practical Progress
Organizing my Openings part 2 - Conceptual Considerations
Organizing my Openings part 3 - Repertoire Reliability (this article)
Organizing my Openings part 4 - Practical Playability
Organizing my Openings part 5 - Structural Struggles (postponed)

Choosing Candidates

Superficial Selection
As I was reviewing my games with the post game analysis, going over each move, I learned for each move whether it was the best move, or not. For each move that wasn’t, if it was almost the best move (excellent or good), if it was an inaccuracy, a mistake or even a blunder; and the engine would always tell me which move was best. This was how I began building my repertoire, even before I was calling it a repertoire.

Right from the start I realized the engine evaluation was not the whole story, especially at lower depths. I used openingree.com to see how popular (frequency played) a move is and the winrates for each move. Ideally, the combined resources would provide a rock solid answer in the form of a triple threat: The most played move, with the highest engine evaluation is also the move with the highest average winrate.
In summary, my initial move evaluation was based on 3 factors

• engine evaluation
• how popular the move is, how often other players play the move
• the average winrate (not just for Masters, but also checking a database of millions of games by the community of more average players)

Expert Opinions

A key source of information is provided by experts. In articles, books and courses. For me the content of choice was video content. Levy Rozman, Eric Rosen, Hikaru, Chess Brah, ChessVibes, Hanging Pawns and many many others. Not only did they educate me on openings and which move to play in which position, they also explained why. A move recommendation by any of them carries a lot of weight for me. Certainly more than a move that’s recommended by the post game analysis at depth 22. But these are not always available.

Creating consistency
It helped me to use a simple straightforward opening repertoire. It created consistency in my play. So often on this site, I see players, mostly lower rated players, that are struggling to get better and when I see what they are playing I’m baffled by complete randomness in their opening repertoire. If you play, e4 and e5 is played, you shouldn’t have 8 different moves you play on the second move. Don't get me wrong, I don’t think 800 rated players should prioritize studying openings, but I also think low rated players should not play random variations every single game. At least, to me it made more sense to, whenever possible, to pick a move that sounds reasonable and play it for a while. Later on, I could always switch when I find a better move to play.

Classifying Candidates
For some positions I was quickly able to find the best move, in other positions I settled for a move that was playable for the time being. What emerged was a classification system I began using where I put moves into one of these categories

Indisputable best moves
When the best move is totally obvious and I know why and how to play, I can safely put it in my repertoire. I should mark them as ‘solved’ so I can focus on other moves.

Obvious best moves
Sometimes, there is more to it. Knowing the best move is one thing, in order to play the necessary follow up, I still have to understand the move. That usually takes a while, so in the meantime I label them as the leading move so I know that I still need to look into it until it becomes an indisputable best move.

Preliminary qualification
In many cases there wasn’t a clear answer. When I have to choose between a move with an engine evaluation that was 0.3 higher than the alternative, but with a 5% lower winrate, what should I do? What I did: just pick one to play for now. With winrate as the leading criterion and create a reminder for myself to do a deeper investigation at a later time. As a preliminary classification I used

• Leading - to indicate that this move looks like the most promising move
• Contender - to indicate that this move could potentially be the best move but I haven’t had time to fully investigate
• Solid - to indicate that this move is a good move, unlikely to be the best move, but too relevant to ignore. In many cases I put the most popular move (most played) in this category. Whenever, I don’t see a move that stands out as best and I don’t see any drawbacks, I can assume this is a solid move at the very least.

To be determined
I have no Idea what the best move is, but the position is important enough to check later
Contender - In the meantime, all moves that show any promise at all get the label contender

Diving Deeper
Conceptual considerations
There is not always an obvious best move and in most positions there are multiple candidate moves that are playable each with pro’s and con’s. I’ll have to make a judgment call on what's the best move to play for me. Conceptual considerations that complicate the decision include complexity (do I want to play a move that is objectively better than the alternatives, knowing I am about to enter a minefield of complexity?) and the risk profile (do I want to play a move that has a lower engine evaluation, but that also wins most of the time because most players don’t know the refutation?).

Priority
After building a first outline of my repertoire based on superficial selection, the more positions I added, the more positions I still needed to investigate. The amount of positions and moves I wanted to further explore began to grow, forcing me to add priority as an additional qualifier.
The key factors here were
Frequency - A position you will see in 10% of games is far more relevant than positions you will play only once every 500 games. Many positions will appear far far less than that, especially at the lower ratings.
Impact - The relative difference between each candidate move also impacts its relevance. In the early game the difference between the best move and the second best move is fairly small so it’s usually fine to just pick one and don’t worry too much.
On the other hand, for every new position, it’s very beneficial to do a quick scan of the position to identify the moves NOT to play.
As a result I added another category: Discouraged moves.
By adding these to my repertoire as well, I was abusing Chess Position Trainer a bit, but it seemed to work.

Final remarks
Instead of a simple collection of positions and my preferred move, I was building a system of multiple candidate moves for each position with varying degrees of completeness or certainty. Since each candidate move should be explored for multiple possible responses by my opponent, you can imagine how the workload grew exponentially. But it’s how I liked it, slowly and methodically improving my repertoire in a structural and reliable way.
Next article, Practical Playability, will cover my experience applying my repertoire to actual games. Or at least, trying to do so.