> the player who committed more blunders lost 86% of the time
In some sense this is almost tautological. While finding an exact definition for a chess blunder isn't straightforward, here is one example from the Lichess UI:
Basically, if you make a move which decreases your winning probability more than 14% over the best move, that's a blunder. But winning probability is a nonlinear function of stockfish centipawns. A drop in 100 centipawns when you're up 15 points isn't a blunder. When the game was equal, it is.
Point is, by the time you know it's a blunder you already know something about the outcome of that move, that it swung the winning probability by more than 14%. So the analysis is kind of just measuring some function of winning probability and saying that it is highly correlated with winning probability.
It's not tautological, though. Your position can gradually deteriorate until it is not salvageable anymore. In that situation we say that the opponent outplayed you.
The fact that the switches in probability occur suddenly is highly relevant. One reason for this could be, that you are able to avoid this type of mistake 90% of the time. So during 9 out of ten moves, nothing changes. Then it does. So in this model, avoiding blunders means honing your skills to be able to apply all aspects of your mistake knowledge all the time. In another game, that is less blunder driven, it might be better to focus on getting more things right most of the time, rather than getting fewer things right all of the time.
At this level, chess is a tactics game. Not a strategy game.
It’s an issue of how the games are evaluated. At most skill levels a human being slowly outplayed is trading blunders from a computer’s perspective.
For a 1400 a game where an 800 crushed a 500 is most likely a comedy of errors. For a GM that 1400 crushing a 1100 will similarly be filled with missed opportunities. And for a top chess engine on significant depth, most games are practically a slapstick comedy because the best of the best represent such a small fraction of overall players.
I would go with the definition of a blunder as a move that you instantly recognize as wrong once you see the response. So the definition is relative to your own skill level.
It's at least partly a property of the game itself. If the game is tossing a coin, first to 100 heads wins, then the win probability will not take major swings. If you have a game where the person in a worse position has many opportunities to recover, win probability does not swing so much.
I don't think that's true. Speaking as someone sub-1000 rated who plays lots of other people in the same region, the computer evals don't typically show a series of blunders. 1 or 2 blunders per game is common, but blunder-less games are also not uncommon. Just lots and lots of suboptimal but not terrible moves.
If you’re rated 800 online and making 1-2 blunders per game playing blitz, that seems low and it still adds up to 2-4 in a given game.
If you mean longer time controls then that’s definitely helping, but most games are 5 minutes or less simply because players are going to be able to fit far more such games per day. Similarly the average rating is quite low simply because everyone starts terrible and most people quit relatively quickly.
Someone dying in a car accident is frequent/common as in happening several times a day, but it’s still an extreme edge case in terms of the average trip.
If we’re talking about the average game you simply need to look at all time controls to get an accurate understanding not pick an uncommon example and arbitrarily suggest it’s representative of the general case.
I meant number of blunders per game at low ELOs. I stand by my estimation that even at 800, even a 5 minute game would have a median blunder count of 1 or 2. Which is hardly a constant stream of blunders, which is what I was trying to point out.
If you’re 800 FIDE rated over the board you’re not going to be 800 online. You’ll be way above that. I haven’t played in years but I was beating 1200 rated players online as a total beginner. The ratings are not comparable.
A sub-1000 online rated player is frankly a very poor, total beginner at chess, not an enthusiastic club player.
That’s position-dependent. In one case you might be in a totally closed position where there’s no obvious way forward and your position gradually deteriorates. In another case you might be in an ultra-sharp, open position where you miss some crazy sacrifice combination that leads to mate-in-10.
The first example is clearly a strategic defeat and the second a tactical defeat. But calling it a “blunder” to miss the sacrifice in such a sharp position feels unfair. You might have been walking a tightrope for a long time to reach that point and then made one little slip any grandmaster could be expected to make.
I once watched a volleyball game between my high school’s vaunted multi-time state champion team and a small private school. It was a perfect example of no blunders vs. gradual deterioration. Our team played an aggressive style that would test the opponent’s physical ability and mental toughness. The small private school played no-mistakes, all defense, and no offense. They would dig up everything in bounds and return the ball. They only scored when our team hit the ball out of bounds. That game took four hours, and the only reason the small team lost was depth: they only had one substitute. Eventually, they wore out and just couldn't physically perform.
These evaluation-centric definitions of blunder are a bit awkward though.
Traditionally blunders were defined in more player-centric way: player blundered, when he made a mistake obvious enough, that a player of his strength is very unlikely to make. So what is a blunder for a strong player may merely be a mistake for a weaker player.
Problem with evaluation-centric definition is that not all moves that worsen position by 14% are equally obvious - if you hang a queen in one that is certainly a blunder, if you miss a non-trivial sacrificial combination on the other hand...
Chess.com is also definitely using an evaluation-centric definition to label moves as blunders. The issue is that this definition is also some function of the change in winning probability.
> So what is a blunder for a strong player may merely be a mistake for a weaker player.
I agree - all attempts at automatic classification of blunder have same problem. This is why analysing games without engine still matters and is going to matter for foreseable future.
Don't forget also impact of time control - shorter games lead to more mutual mistakes. While in 90+30 first big blunder should decide the game, in blitz it's just the beginning.
Maybe a better analogy would be card counting in Blackjack.
To be profitable (if it can be), card counting works with extremely tight margins, like a fraction of a percent per hand. It only turns a profit averaged out over many hands.
But if you make a basic strategy blunder, you can lose the statistical benefits of maybe hundreds of perfectly played hands. That's why it may be better to play a simple strategy perfectly, than a more advanced but error prone strategy.
That's also the reason why casinos love wannabe card counters. Their strategy may work in theory, but because of mistakes, the end result is worse for the player than playing basic strategy.
Note about basic strategy: it is the optimal way of playing blackjack assuming each card draw is independent (so, no card counting), it is simple, and widely available and accepted in casinos. The player is at a loss (of course), but reasonably so. If you can play basic strategy consistently, you are better off than the vast majority of players.
I feel ignorant, I thought with card counting you were still pretty much playing the simple strategy but altering bet sizes as your expectations of winning increased?
Your understanding is correct, more or less[0], but there are two parts to strategy: an inexpert counter is likely to be distracted, and to make errors from perfect play. They’re also likely to lose the count, and make errors in bet sizing. The net of those is worse than a non-counter playing perfectly, whose edge is slightly negative but who still stands a decent chance of making money on a given day.
But note that’s a reason for casinos not to overtly discourage counting; they’ll still happily ban a player who is apparently counting well rather than roll the dice on whether they’re counting “well enough”.
[0] Sibling points out that counters will make specific deviations from “naive” perfect play depending on the count, but that’s to push earnings up a bit on an already positive edge. There’s also the element of camouflage, where a really strong counter might deviate in ways that don’t hurt their earnings but make their play look less “counter-y”.
Chess.com is more sophisticated than this in the treatment of blunders. They are divided into “misses” and “blunders”.
In my experience, it appears that the difference between the two is that a “miss” is something the computer evaluates as unreasonable or difficult for a human to find. If you had found it, it would have been deemed a “brilliant” move, which is another analysis move type that chess.com has doesn’t have. Either that or a miss is failing to capitalize on an opponent’s blunder.
It makes sense to chess players, since we consider missing an opportunity to capitalize on an opponent’s mistake to be distinct from unilaterally making one’s own position worse, even though to lichess those are going to both look like drops in the evaluation score.
Chess.com isn't really more sophisticated than lichess - it's only trying to appear so.
It's definition of blunder etc is still based only on engine evaluation. For example it marks as briliant all sound sacrifices, even the most routine ones. This is good marketing, but I doubt it's good analysis.
It's just marketing to make players feel better about themselves.
For example when i tried chess.com out they marked a simple queen sacrifice to deliver back rank-checkmate as brilliant even though it's an obvious move for any intermediate player with half a year of experience.
Lichess does nothing of that marketing bs.
The "don't blunder" advice applies to sports as well, where you hear it phrased in non-tautological ways frequently, typically something like "you need strong foundations."
Even at the peak of college level, I remember winning and losing games because someone did something that you mostly stopped doing soon after learning the sport, like pass to a teammate who wasn't looking, or fumble the ball in a preventable way, or things like that. Some teams focused on fancy plays and corner cases while their fundamentals still failed frequently, leading to more losses due to bad fundamentals than than wins due to good advanced plays.
That's how I read the "blunders" stuff. There's some small set of a priori known foundations you need which are fairly simple to keep from going wrong, which nonetheless lead to a large portion of failures/points given up/whatever.
It might be tautological, but it also happens to be correct! The Pareto rule applies: A beginner progressing to intermediate might quickly iron out the biggest blunders and by the time they're advanced get 80% there, but mastering that last 20% requires decades of practice.
Also, getting better changes what a blunder is. When I began, hanging my queen was a blunder. Then allowing a discovered check was a blunder. Then allowing the threat of a future discovered check affect my move is a blunder etc.
In some sense this is almost tautological. While finding an exact definition for a chess blunder isn't straightforward, here is one example from the Lichess UI:
https://github.com/lichess-org/lila/blob/b527746b179cdde6438...
Basically, if you make a move which decreases your winning probability more than 14% over the best move, that's a blunder. But winning probability is a nonlinear function of stockfish centipawns. A drop in 100 centipawns when you're up 15 points isn't a blunder. When the game was equal, it is.
Point is, by the time you know it's a blunder you already know something about the outcome of that move, that it swung the winning probability by more than 14%. So the analysis is kind of just measuring some function of winning probability and saying that it is highly correlated with winning probability.