As you probably know, I used to play poker for a living. Poker players like to talk about “the graph.” The graph is money won over time: time on the x-axis, profit on the y-axis.
Obviously, if you’re trying to pay your bills with poker, you want this graph to be going up and to the right, reflecting a long-term strategic advantage over your opponents. But if you zoom in far enough, the graph will always be jagged. Not even the best poker player wins every hand. There are going to be times when you get unlucky and in those instances the graph will go down.
In poker, the source of luck is obvious: the random draw of cards from the deck. You can get a string of bad hands, or you can get all your money in with the best hand and still lose to an unlucky card off the top. This is simple enough to understand and most people find it unsurprising that the graph would be jagged.
Chess, in contrast, has no such obvious mechanism for randomness. There is no hidden information and both players are in control of all their moves. Yet, if you look at chess’s equivalent to the graph - rating over time - it looks suspiciously like a poker graph. As in poker, you’d love to see an overall trend going up and to the right. In the chess graph, this reflects consistent improvement. But also like poker, the reality is that everyone’s graph is jagged.
While chess doesn’t have an obvious mechanism for luck like poker, there is still luck in the sense that you can’t predict the outcome of any single game in advance. Otherwise, we wouldn’t have to play the games, and the graph would be smooth. But without the random drawing of cards, where does this luck come from?
The truth is that even in poker the cards are not the only source of luck. There are many ways luck can manifest. For example, if you get a great hand when no one else has anything, often you will bet and everyone else will fold. You win a small pot, which barely registers in the graph. But if one of your opponents had a very good hand (but not as good as yours) you would have won a big pot. But if someone had an even better hand than yours, you might lose a huge pot. So you could see outcomes of small win, big win, or huge loss, all while playing correctly from a strategic perspective.
But probably the biggest sources of luck don’t come from the cards at all. You might sit down at your usual game and run into a particularly favorable (or unfavorable) lineup depending on who chose to play that night. You might be unusually focused, or distracted because of an unrelated life event. There are things you can do to maximize focus and energy, but you never have perfect control over everything that happens in your life.
Additionally, you and your opponents are both trying to make the best decisions with imperfect information, strengths and weaknesses in your game, and inherent biases. The way these decisions line up can be unpredictable. You might make a stupid play that works out brilliantly, or a great play that backfires for reasons you could never have predicted.
In short, the luck of poker has at least as much to do with the inherent unpredictableness of life as with the obvious luck element of a deck of shuffled cards. This helps illuminate the luck component of chess: there is no obvious random mechanism, but there doesn’t need to be. There’s luck in every part of life, including chess.
In chess there is technically no hidden information, like an opponent’s cards in poker, but the complexity of the game renders some information practically hidden. You are trying to look into the future, but the complexity of the game is far beyond anyone’s ability to calculate exhaustively, and you also don’t know your opponent’s intentions.
The reason this isn’t all just empty philosophizing is it means, sooner or later, you will encounter a stretch of bad results (in poker we’d call it a downswing) and have to decide what to do about it. It is normal at these times to question your approach:
Your training plan isn’t working
You need to change your opening repertoire
You need to change how your prepare for tournaments
These are reasonable things to think about, and there could be some truth to them. But once you understand the role of luck in chess, you understand that none of these is necessary to explain a string of bad results. Sometimes it is just pure chance with no underlying cause. In fact, this may even be the case most of the time.
Assuming you are pursuing a reasonable training plan – playing, reviewing, and solving puzzles consistently – you are more likely to make a mistake by overadjusting than underadjusting. If you actually implement a solid plan for 6 months to a year and don’t see your rating go up, then yes, you should think about making adjustments. But what’s far more common is for players to freak out and stop playing for a month, or radically overhaul their plan, at the first sign of adversity. If you do this, you never really know if the plan would have worked. You didn’t give it a fair shot.
My friend Dan Bock created a spreadsheet that calculated a random string of results for a player of a given rating. I find this helpful to show students to illustrate the role of luck. In real life, there is always the possibility that some underlying factor really has changed, and the results are not due to luck, but a genuine change in ability. But in the simulation, the player’s “true” strength stays the same the whole time – that is stipulated explicitly. So it shows very clearly how easy it is for your rating to bounce around, even if your strength stays the same.
What I’ve noticed from running a lot of these simulations is it is very easy for your rating to deviate up to 100 points in either direction from your true level, but it usually rebounds pretty quickly. Here’s what I would take away from this:
Short term deviations (in either direction) are the norm. Do not read too much into rating swings. They can easily be the result of chance.
The biggest risk of a downswing is not the points you lost, but becoming discouraged, or abandoning your plan. Improvement usually comes from following a solid plan consistently over a long period of time.
If you have actually played and practiced consistently for 6 months to a year and don’t see your rating go up, consider adjusting your plan.
When I worked, my boss and I had a discussion about employee performance. Part of his job was to periodically rank (or assign ‘goodness’ numbers) to each employee, and he was struggling to explain differences.
The trouble is that there is never just one axis (or knob…). There might be 25 or 30 different components for an employee’s contribution. 15 technical ones; 8 interpersonal ones; maybe 5 character qualities; in addition to external factors like you talked about.
Chess is hard, and I think the same applies. There is a scale for experience/expertise with how each piece maneuvers, then you can add in the expertise for two-piece interactions; as well as creativity; a brain fog scale that acts as a multiplier for everything perhaps ….
No game touches all those axes or knobs; so some jaggedness is expected.
Randomness is always in the eyes of the beholder.
Even dice thrown. is just lazyness to go measure the initial conditions to the precision needed to predicted the dynamics of the free rotating dice and its bouncing on a surface with say little friction. We just prefer to not do that, and look at the odds of its 6 stable attractors that friction even if small ends up giving.
For chess, the uncertainty comes from single mind limitations, even the best of us. Each individual fog of chess. Even if we wanted to put the energy, even the best of us, chess will have some residual fog left. (for engine we are doubly in the fog, that while it can beat any of us, we do not know that the cumulated engines in a huge pool would explore the real absolute chess, not just an engine club self deluded notion of absolute chess, they could be competing over a tiny patch, as long as no human can go through that, well we hail its oracle glory).
What is nice though is that this lazyness, can be quantified and modeling in math language, exactly. We measure and control our ignorance by assigning a world of events, and some blob probablility over it to represent it.