Serious Chess Improvement Vol. 2

How I improved my online bullet rating from 1184 to 1975 from December 15th, 2020 to March 2nd, 2021.

Start: www.chess.com/member/yakunhu

End: lichess.org/@/AmazingFilms1971/perf/bullet

Some programming notes:

– This article is not going to be particularly lengthy. I did not improve my Elo by 791 points in 2.5 months by reading 300 page chess books, so I’m not going to write one either.

– I will not include images or videos as they may be triggering for people in the mental health community, of which I am a part.

– Before reading this article, you must accept the premise that LiChess.org = FIDE. There’s no difference between online rating and otb rating, because all rating is based on the relative skill gap among players and, in the future, there will be as many if not more opportunities to play for lucrative prizes online.

– I’m writing this “off the top of my head” to give you an accurate path through the structure of my thinking at this present time. This is simultaneously the most truthful way to write and the easiest way for you to read.

Theory of Improvement: How to Improve at Anything.

Chess is a skill, and like any other skill, the raw components of “being good” can be extrapolated. You are required to:

1. Assess an existing situation and make observations.
2. Evaluate the position.
3. Come up with a hypothesis with respect to the best move.
4. Execute the move such that the existing situation changes to a new situation.
5. Repeat the 1. – 4. loop from 1.

Anyone who has studied science knows that this process maps well to the scientific method. Thus, the raw unit of skill improvement is optimizing the rate with which you can repeatedly execute the 1. – 4. loop using situations, in this case board states, that are relevant to the domain in which you would like to gain expertise. There’s another argument to be made that insofar as chess is a skill that requires visual processing, a mental faculty that occupies 50% of your brain’s processing capacity, that domain expertise in chess is itself extrapolatable to other skills that require visual differentiation.

Spaced Repetition Gives You Authority Over Your Own Learning.

A key frustration that students have in skill accumulation is the sense that while the act of spending time studying is fully within one’s own control, the requisite wiring of new synaptic connections and associated biochemical changes is not. There is an underlying fear that I might invest 10, 100, or even 1000 hours on a particular improvement paradigm and, despite encountering information that is objectively useful, fail to recall that information during a critical moment such as a GM Norm Invitational tournament. This irrational fear is rooted in a lack of understanding of how information is stored in the brain and the retrieval pathways that must be established in order to retrieve it. Fortunately, the present author is a student of biology at the University of Buffalo.

Note to self: the best way to overcome writer’s block is to play a round of LiChess Puzzle storm. Top score as of 3/5/21 = 49.

In other words, the formation of a long-term memory as well as a lighting-fast synapse – neuron – synapse pathway to retrieve it requires long-term biochemical changes that are preserved on a long-term basis. I’m repeating words because scientists have reached consensus that the number of recalls required to store a quanta of information in long-term memory is 5. The five recalls must be made using a spaced repetition calendar in which recalls are practiced with intervals of n * 24 hours from the previous recall, where n is the retrieval number and the first recall is practiced 24 hours after initially learning the solution. Alternatively, you can use a spaced repetition software that automatically tracks your recall schedule.

Formula for Identifying Critical Positions Worth Studying.

Once you have established in yourself that learning is fully within your control and not a magical ability that only children have, as many in the chess community are prone to believe, the next step is identifying an improvement paradigm in which you apply the effort spend towards learning towards the highest yield positions within the chess domain. Once again, chess is a skill like any other skill, and I am not aware of a decline in ability with age, for instance, to learn science and take multiple choice exams. My father earned a third doctorate at the University of Buffalo in his late 40s as part of a mid-life career transition from surgeon to associate professor, and I am 28 years old in the midst of my most stellar academic year in undergrad. How is making a move decision in chess out of the finite set of available legal moves different from selecting the correct answer in a 5-answer multiple choice exam? It isn’t. Nevertheless, if chess improvement capacity does decline with age, then my record of 791 points in 2.5 months is a testament to the efficacy of the proposed methodology being as it has allowed me to defy the limitations of age. Previously, my record for improvement in chess was 564 points in USCF tournament play over the 9 month period of 2007-02 to 2007-11, when I was 14 years old, using traditional chess pedagogical methods. Thus, regardless of your baseline level of improvement as established during the teenage years, you should expect a (791/2.5) / (564/9) = 5.0x improvement over your prior rate of improvement, with an upwards adjustment if you are still in your teenage years at the time of reading.

Now, how do we pick the correct board positions? I applied a formula to identify board positions in which the statistically significant confidence for winning for either side is >90%. Using statistical modeling provided by chessprogramming.org, a >90% win percentage corresponds to a pawn advantage of 5 points or higher. For the purposes of this specific Elo improvement interval, ~1200 to ~2000 in 1+0 Bullet play with single-step premove enabled, +/- 5 was sufficient. However, you might find the need to tighten the interval for higher levels of play, especially in longer time controls, and I will be exploring this further as I push into the upper 2000s.

My methodology for selecting board positions to study in the latter part of the improvement sprint, in which my Elo increased from 1721 to 1975, was to use the in-browser Stockfish in LiChess to identify board positions in which I made a move that changed the evaluation of the board state from a W for me to a D, a W for me to a L, or a D to a L for me. This corresponds to an evaluation change from > 5 in the board state prior to the mistake to < |5| after for the W for me to D condition, assuming that I am white, >5 to <-5, again assuming that I am white, for a W for me to a L, and <|5| to <-5, if I am white. Flip the numbers around if you are black.

After selecting the board position, I took a screenshot of the board state prior to the mistake using the Windows + Shift + S shortcut, which captures a screenshot and adds it to the clipboard without generating an image file on your desktop, pasted the screenshot into a Cloze deletion card in Anki, and wrote a caption beneath the screenshot in the following format.

-1.0 {{c1::Nc6 Rd1 h4 Qf3 or f4}} -2.4

The first number corresponds to the board position evaluation corresponding to the top line in the Stockfish evaluation table while the second number corresponds to the evaluation of the bottom line.