During the month of February I played just under 400 games on Lichess. These games were mostly played experimentally, without putting much study into the improvement of my game. I then started taking studying more seriously, and began on the chess.com platform, where I played just under 200 games so far. This has yielded interesting statistics already.
The stats from Lichess show that over the course of last month's experimenting, I was on a pretty good roll on the Lichess platform. There were peaks and valleys, but it was overall pretty steady and linear progress. While I have shifted my focus to the chess.com platform for now, I intend to check out Lichess again down the road for comparison.
The shift to chess.com marked considerably stiffer competition at my level, and it's a known fact that a rating on Lichess is often higher than a rating on chess.com for the same player. I hopped onto an account I hadn't used in several years, and it took a few days for the matchmaking system to place me. This resulted in me quickly dropping down to the 600s, and then linearly improving into the mid-700s, over the last few weeks. Study, puzzles, lessons, and a lot of practice (with books on a physical board and on the chess.com platform) have enabled my game to continue to improve.
This is illustrated by the 90-day graph from chess.com and the 7-day graph from chess.com. A naive way to interpret this data is that by increasing my rating at the same rate as I have over the last week, I can achieve a 2500+ rating within a year, but that is not the most educated way to look at the data. There are some noteworthy things to look at here.
Firstly, let's take the graph from Lichess and what it represents. I fumbled around a bit as I got my bearings (I have been a casual chess player since I was a young child, but only very recently began studying how to "get good" at it), and then when some of the important parts "clicked" I experienced explosive gains until the mid-1200s, where I ran into competition that could not be beaten without further study (comparable, I think, to the 600s or so on chess.com, for me). I didn't do much studying per se during the Lichess phase of last month's experiment, but I did analyze my games and work on the fundamentals, which resulted in the growth spurt. It will be interesting to measure Lichess gains further in light of the progress I have made on chess.com, down the road.
At that point (early March), I decided to take the experiment to chess.com (a larger platform with stiffer competition and a more expansive learning section). I had an old account I had used for playing very casually back in 2019, which was sitting at an oddly high rating (almost 1000). As I played a number of games over several days in the same style I had been on Lichess, my rating was normalized to the low 600s for a bit.
Not settling for this, I began to dig in to the (laudably encyclopedic) learning materials on chess.com. At this point I have beaten all of the beginner and intermediate training bots on the platform (rated up to 1400) and have taken almost all of the stock lessons (although I have not even scratched the surface of all the lessons available through the platform). I also purchased several books, and attended a local chess club (which offers learning material as well). This improved the fundamentals of my game to the point that I have experienced linear progress up to about 760 over the last week.
The matchmaking system, combined with the rating system, on chess.com is very effective. One could stay around the same rating by doing okay against opponents of their own rating, losing most games to players of a higher rating, and beating most players of a lower rating. To continue to improve one's rating consistently one would want to beat most players of their own rating and do okay against player's of a higher rating, while beating most players of a lower rating. To do that requires applying lessons learned, analyzing games and statistics, and making a point of shoring up weaknesses and taking advantage of strengths at every level of play. This is entirely feasible. So while the unrealistically optimistic timeline of reaching 2500+ in less than a year is likely not accurate, it is demonstrably possible, I think, to reach Grand Master within 10 years -- provided it is taken seriously! The chess.com platform provides a shocking wealth of statistics, but I am going to build my own database here and do analysis which goes beyond that, as time goes on. It's important to note that you don't have to win anywhere near all of your games to have this linear growth. You only have to win about 50 games for every hundred, actually, while losing about 40, and drawing about 10, roughly.
There are complications to consider: first of all, rating increases from a win diminish at higher ELO ratings, so that someone rated, say 2000+, will get less points for a win than someone rated 700. Also, some concepts, habits, and style improvements will take considerable time and effort to take to "the next level" at any given level, so that time should be assumed on the front end for that in a precautionary way. Nevertheless, I think ten years to Grand Master is a totally realistic goal.
At this stage of the experiment, my strengths and weaknesses are pretty obvious and it is clear from the available statistics on chess.com how I can continue to experience this linear growth and win most games against players my level. I leave too many pieces hanging, I fall for the occasional avoidable checkmate, and my mid-game strategies are not yet ideal. My endgame play is strong for my rating, and I win way more games than I lose in both the opening and end game phases. I win more games with black than white, by far, and usually by playing the Sicilian (which I don't know super well, but something about it at this rating level is very effective right now).
So, objectively, by resigning less games ( most of my losses are from resignation in the midgame), I will improve my rating by winning in the endgame or drawing games that should have been losses. Over the slightly longer term, by studying the openings I am good with, I can make them stronger and ensure they will not only win for me at this level but continue to be useful to me at higher levels. I am at the moment somewhat propped up by consistently beating people of a lower rating, but I have a number of advantages against people my own rating and slightly higher as well. By focusing on my strengths and weaknesses, I can sustain or even accelerate the growth.
Over the very long term (perhaps years from now), the means to improve my game at a given level will be less obvious, and will want to do things like data mine move databases and my own games to look for things I can improve or take advantage of by going beyond the data immediately available on the chess.com and Lichess platforms' analysis tools.
I currently lose more games than I win with white, and win significantly more games than I lose with black (owing again to the Sicilian Defense, for the most part). My overall win/loss ratio is not bad, as a result, with a lot of room for immediate and long-term improvement!
I suspect that the reason the Sicilian Defense is so good for me right now is that it gets white players out of their opening very quickly, allowing me to use fundamentals to win games that I would have lost using openings and defenses they were more familiar with. To continue using it at a higher level will require learning more about it. It would be neat to go through and track how many games I won versus lost while using the Sicilian in particular, as well as other openings, tactics, and strategies. That's on the TODO list for sure!
To take an initial crack at deeper statistical analysis, I looked at 51 games I played with black over the last week. I played the Sicilian 21 times, winning 10 of them, losing 8, and drawing 2. So it's not the source of my exceptional win rate with black at the moment. That honor actual goes to King's Pawn openings where I play the Italian Game. Of 15 such games, I won 12, lost 3, and drew 0. The rest of the games with black in that sample fell in other categories. This is just scratching the surface of the kind of analysis which might be possible by categorizing games by opening, phase, color, tactics used, sacrifices made, etc., which go beyond what is available on the (already exceptional) chess.com analysis tools. For future updates I will track everything more carefully and make graph visualizations using all the data points, by keeping track of notable features of each game. This may result in less games played, but should result in better games played as the lessons learned are applied and avenues for effective study are revealed (beyond the obvious, such as hang less pieces and fall for fewer easy checkmates).
It is interesting to examine the graph of my move accuracy over the course of a game, by color. I am markedly more accurate during key midgame phases with the black pieces.
It is also interesting to note that the classic advice "castle quickly" is borne out in the statistics over the last 100-ish games with both colors. I do demonstrably better when I castle in the opening.
For study I have mostly been using the chess.com resources, and the occasional YouTube video, in addition to going over annotated games from books on a physical board. While YouTube videos are educational, entertaining, and insightful, I have found that the other two resources are more effective in general when combined with actual practice in games. The main book I am going through at the moment is Logical Chess by Irving Chernev, which is exceptional. I am on the section which focuses strictly on Kingside Attacks, and I've used the lessons learned to win games at my level.
Another important tool for learning is the wealth of training bots on chess.com, which are categorized by rating level and also have personalities and stylistic quirks which allow you to get extensive situational practice in an extremely pragmatic way. I systematically beat all of the "beginner" and "intermediate" bots, which collectively rate up to 1400, and have made good headway against a 1500-rated "advanced bot". This is perhaps one of the most valuable resources the website offers, as there is no substitute for "sparring", and it does not hurt your rating to experiment on bots (before taking the lessons learned to rated games!). That they have humanistic quirks makes them more valuable on their face than simply tackling bot games against various levels of an engine that hasn't been tweaked to have such quirks.
An interesting statistical aside is that, in my experience so far, a chess.com Rapid (10-30 minute games) player who is rated 700-800 is about as challenging as the higher level intermediate bots (rated about 1400). This suggests that the pool of competition is quite good, and underrated in general (or that the bots are rated too high). It would be interesting to know more about the inner workings of all that.
I am not going to commit to "weekly" updates or anything. I will make updates of varying sizes here and there whenever there is interesting data or milestones to look at. The journey begins! Grand Master within ten years! Gonna do it!
back to index
