(I made up the exact numbers, but the idea is there.)
That seems like a meaningful interpretation of "600 points" that applies to anyone -- though the difficulty of actually making this improvement definitely varies with your starting rating.
It does apply to anyone, but it is more or less meaningful depending on where you start, so the meaningfulness isn't equivalent.
It's as if you say you can double your money, but it only works once and with a value < $1.
The idea that, say, Magnus could increase his chess playing abilities in 6 months (or even 6 years) to be able to beat the current version of himself 99% of the time would be insane.
Trade offs may be something like in yours teens and early 20's your brain may have the most plasticity and ability to visualize (plan 10+ moves ahead) but you might not have accumulated enough experience.
I'm purely speculating here and just wondering aloud. (I bring it up in response to this comment because Magnus' prodigious talent is so noteworthy I wonder when Magnus will stop being able to "beat" Magnus of 1 year ago.
I think there are a lot of confounders to consider. Though GMs like Anand show a drop in standard rating (https://ratings.fide.com/profile/5000017/chart), his blitz rating is near his all-time-high (ie. is his standard rating drop due to decreased mental performance or a shift in interest/focus to blitz?). Similarly, I suspect a lot of strong players who fall in the `2000<FIDE rating<2300` realize they may not be the next magnus and shift focus when/if they make the decision to pursue a career outside of professional chess.
While this doesn't answer your question, it's interesting to note Magnus's peak rating was actually 7 and a half years ago, when he was only 23.
With the exception of Anand at 51, they're all quite young.
So yes, performance does fall off with age, though not as intensely as something like hockey.
Someone who is ranked at around 1200 and really commits to improving can likely beat the old them in a couple of months by memorizing a few common openings and practicing drills/working on fundamentals.
Someone who is a dedicated chess player and ranked above 1800 may never be able to beat the old them with 99% probability.
So if someone says they improved by 600 points, certainly that is meaningful to them as an individual and it means they can basically beat their old self, but it won't be very meaningful to me.
I don't know what it means in theory. In practice I win and then lose 200 points in lichess in a few days.
Some time ago I used to win and lose 100 around the day in a cycle. Maybe ratings aren't adjusted around the globe, so for the same points there are different skill levels as you go through time zones.
There's a fact that really annoys me: for my 2+1 bullet, it's harder to be at 1500 than at 1600. Once I'm at 1600, I can reach 1700 with a winning streak. If I fall into 1500, I tend to get stuck there.
So yes, logarithmic curve and dimensionless pedantry and all that, I'll grant you; but that's missing the trees for the squirrels in the forest, since 600 points means "I went from noob to pro." No one expects Magnus to gain 600 points, since that's quite impossible.
I guess you could argue that it's possible for someone to start at 1500, then really suck at chess and drop to 900, and then merely become average again (1500), and claim a 600 point improvement. As with the other observation, I agree, that would be impressively misleading. I'm not sure that's the claim, though.
EDIT: I retract every claim. I am in fact an idiot, since the article does say they went from 1200 to 1800, not 1500 to 2100.
This is what I get for writing something dumb. At least I admit it right away though. Sorry.
At least it's proof that the pedantic-ness wasn't so pedantic.
I also take solace in the fact that the actual title of the article was stripped: the title says 1200-1800, not merely 600 points, which in this case is crucial info. But! It was remarkably stupid not to actually click on the link before writing, and this is rather public proof that sometimes I don't. Perhaps that's a strong signal that in the future, I need to be. :)
This is showing some pretty serious misunderstandings about ELO ratings. 1,500 is absolutely not the starting point. When you first learn Chess, your rating will be low hundreds (like under 500). It takes a decent bit of practice to work your way up to 1,500, at which point you are already decent. 1,500 is roughly average among people who play Chess competitively. You certainly don't start there. 1,800 means you're good enough to beat most players at an average low level local tournament.
It turns out that watching GothamChess doesn't make me a chess player.
Thank you for that. It's mildly interesting analyzing the source of why I was so wrong:
As someone who aspired to be pro at Dota (but was never too skilled at it), my "competitive instincts" have been calibrated for games where you do indeed start at some baseline, even among competitive players, because you will quickly be balanced out to the proper ELO. For example, in HoN (precursor to Dota 2), 1700 was widely considered pro, whereas everyone started at a baseline of 1500. The noobs were quickly punted down to lower than that. (It could've even been 1300 and I'm misremembering, but the point is, the competitive scene was still balanced around 1500 as a baseline.)
Ditto for Dota 2, back when they had explicit MMRs. (MMR = ELO.) Nowadays they don't have MMR, they have ... tiers? ... since they realized that it kind of sucks having a community obsessing over what your actual number is, rather than what division/tier you're in. So they were like "Ok, congratulations, you've reached Immortal tier, you're now a pro."
Anyway, when there was MMR, it still started at some baseline. Because again, the noobs would quickly be punted down to where they belong. 5k was widely considered pro back in those days, back when 5k meant something. But MMR inflation meant that the benchmark then became 6k = pro, and eventually 7k was top tier (I think?), so this was already a de facto tier system.
Point is, saying "1,500 isn't the starting point" for chess, but yes of course it's roughly average among people who play Chess competitively. The competitive scene is all that matters. Me blatantly not reading the article was based around the assumption of "Of course this is referring to the competitive scene."
As I said, it's interesting just how wildly wrong those assumptions were. :)
It's basically impossible to discuss chess ratings or changes in them without "type" information :)
I sit around 1450-1500. Myself and others I play against regularly throw pieces away for free. Not fall in to traps or anything like that, or even when rushing on low time, I mean just straight up plonk a Queen down in the path of a bishop during the early/mid game for no reason at all. And then sometimes our opponent doesn’t even realise we did it! It’s certainly not competitive tier play.
600 points, in and of itself, is almost meaningless.
It's a good climb but nothing particularly mind blowing.
First of all they went from 1200 -> 1800, not 1500 -> 2100.
Second, just because you are given a preliminary rating of 1500 doesn't mean that you are a 1500. I don't know where you would get that idea from. If my 10 year old nephew signs up for lichess and never plays a game, by your logic he's a 1500 rated player.
Third, an elo of 2100 is definitely not "pro" in chess. Especially not an unofficial lichess rating of 2100. A FIDE rating of 2500 is the minimum to be considered a Grandmaster, which is the beginning of anything resembling pro.
But yeah 2100 lichess is not even close to pro.
1800 is still in the "learning how to play well" stage but getting there from a standing start in 6 months is indeed nice progress. He doesn't seem to claim it's anything more than that, which I appreciate.
My system is very simple: Before I play a game I must complete at least five consecutive puzzles. Yes, this might mean I play 12 or 20 puzzles before I get five in a row. What's interesting about this is that if I get to a dozen or more puzzles and did not solve five in a row, I take it as an indication that my brain isn't in "chess mode" and go do something else. Every time I ignore this indicator I lose games.
Just work on puzzles and keep it simple. My kids have a great time with this simple rule. They don't have to memorize anything and they progressively get better and better. Above all, they don't get worked-up about losing at all. Keep it simple and fun.
Though I also think the real value in puzzles is not that they help you win more, but lose less.
I often play a puzzle that the advantage gains a piece over one or two lost pawns, and when I play these I love to go then and try to beat stockfish 9, undoing moves freely. I often still lose, but what I learn is how to defend after you blunder and how to avoid blowing an advantage.
I think just solving the puzzle as if you are the stronger player taking advantage of a blunder is less beneficial than solving the puzzle as if you were the one who made the blunder and what do you hope they don’t see is better.
I prefer lichess, and sometimes go to chesstempo for more focused work, but reasonable people can disagree.
Also there are many books of puzzles.
https://www.amazon.com/Woodpecker-Method-Axel-Smith-Tikkanen...
?
This is why HN is great -- appears to be an amazing book.
https://www.chessable.com/the-woodpecker-method/course/10582...
And most of the Chessable courses can be "woodpeckered" through the interface.
I am 2100 in lichess 5/0. Most of my games are still resolved by blunders on either side. It's rare to get a positional advantage and then grind away; instead someone didn't see that backwards knight or lateral queen move.
Odds ratios being constant with difference is what you’d expect with a logarithmic scale, with a linear scale you’d expect odds ratios to be constant with the ratio of the ratings.
So, you’ve just explained the way in which elo is logarithmic as your evidence that it is not.
https://math.stackexchange.com/questions/1731991/why-does-th...
What's a good book for this?
e.g.
https://www.google.com/search?q=site%3Alichess.org%2Fstudy+s...
Memorizing the 20-30 traps and mistakes in those studies is sufficient for beginners; you can use spaced repetition (I use my own private Lichess study to collect positions) and cover the 20 or so most popular openings in a few months.
Again, not enough to win but it helps you punish bad play and learn how to handle aggressive players (since many of the traps are in fact bad play but only with perfect counterplay)
These are good tips for beginner to intermediate growth. The things that definitely help the most are:
* Pattern recognition - the best courses for this level are things like "Common traps in <some random opening>", applied with Woodpecker method. Once you've memorized all the mistakes in the Scandi or London system, you can really crush a lot of people who play haphazardly.
* Study your own games and games of people at or just above your level. Four simple methods:
1) during the game, write down (Lichess has a notes section on the left) 3 candidate moves for every move in the middle and endgame, why you're making a particular move, and what you think the opponent's response will be
2) use the "Learn from your mistakes" button after each game during analysis
3) check the most common moves in the opening that are different than yours, play through a couple of masters' games to see why those positions are preferred.
And my last tip which helped me a lot just with the "meta" of playing chess ...
* Use more time. Be okay with losing games because you run out of time thinking. Always, always, always try to play the best move, even if it means spending a lot of time.