@mkubecek said in #40:
> > How much more, a month, a week?
>
> Depends on the system. But with a system that would have no dynamic reliability evaluation and would provide reasonably stable ratings for established players it could easily be 50-100 games. That's 50-100 mismatched opponents, frustrated by having to play a severely misrated opponent and those games fully affecting their rating. With Glicko-2, it's way fewer games and also playing an opponent with unstable rating affects one's rating less than the same result against an opponent with the same rating but stable.
>
> > It might take a little longer, but at least it eliminates chance and manipulation.
>
> Correction: it would take significantly longer, it would not eliminate manipulation and instead of *chance* of temporarily having inappropriate rating it would *make sure* the rating would stay inappropriate for longer.
You're worried again about opponents who will play against someone uneven. And again, on a site that allows multiple accounts, which inevitably leads to the same player repeatedly beating everyone they encounter along the way. As far as I understand, no one cares.
@Italiya said in #32:
> Again, it's no one's business if someone didn't have time for tournaments.
Your opponent’s argument was that achieving a goal does not always require taking the longest possible path. Different goals call for different methods, and the most efficient path is usually preferred.
When the goal is rating adjustment, the Glicko-2 system is the efficient path.
The path to the world championship title is structured the way it is because it was historically the most appropriate process for determining a champion — not because someone wanted players to waste time just to “earn” the right through suffering.
>"@Italiya said in #32:
> No, I mean that reaching 1900 will be a cakewalk if your potential is 2100. In online chess, you don't need to plan or allocate time. You can play simply on your way to school or work, and at any convenient time."
You will still need more steps than necessary to make your rating match your level.
Why should someone spend 30 games instead of 10–15 to achieve the same result?
@Italiya said in #32:
> You keep talking about your personal time constraints, but if you're using professional sports as an example, it's the same. People there are forced to spend a lot of time on it.
People spend time to achieve a goal, not for the sake of spending it.
If a more efficient method exists, it is reasonable to use it.
@Italiya said in #32:
> Of course, the rating is a measure of playing strength. I don't understand how the absence of a question mark could worsen the system? The player will move forward or backward, not getting stuck at one level for long if it's already inappropriate. However, there are many nuances associated with these sudden jumps. For example, as I already described, if someone decides to play on Berserker mode or is simply in a bad mood and loses 15 games in a row, their rating will be fixed at an inappropriately low level. This question mark is not only unnecessary, but also a deterrent to using some of the site's features. Or they could simply win 15 times. Such streaks aren't common, but they're also not uncommon.
A system without fast rating adaptation creates several problems:
1. A player needs more steps to reach a rating that reflects their strength (e.g., 30 games instead of 10–15).
2. Because normalization takes longer, more players will have underrated ratings. Strong players will misjudge them (playing for a win instead of a draw, for example).
3. Regarding players who lose many games due to tilt, berserk play, etc.:
(1) There is no evidence that playing in an impaired state during uncalibrated rating periods occurs often or has any meaningful impact on the system as a whole.
(2) In the long run, a system with fast adaptation still reaches rating accuracy faster for the majority.
Here is a small thought experiment:
We compare two systems — one with fast adaptation and one without.
Suppose we have 30 players: 10 at 1000, 10 at 1500, 10 at 2000.
Each group has 2 players who had either +15 or –15 streaks. Initial rating is 1500.
System with fast adaptation:
6 players will need ~50 games to stabilize.
The other 24 players will normalize in ~15 games.
System without fast adaptation:
The streaks don’t matter much, and 10 players (the original 1500 group) normalize quickly.
But the remaining 20 players will need 50+ games.
Thus, in one case we have 10 players with a quickly normalized rating, in another case we have 24 players with a quickly normalized rating.
24 > 10. The optimal system is obvious.
@Italiya said in #32:
> It's absolutely certain that you can reach level 2600 in the puzzles, since that's your fixed level. But try solving just one puzzle correctly on the medium difficulty level—you can do it, that's a fact. You'll immediately increase your record to 2800. Then switch to the easiest mode so you don't lose too much if you lose, and most importantly, so you're given puzzles that have a very low chance of making a mistake. Once the question mark is removed, you might be able to maintain 2800. Then, deliberately lose to your original 2700 and try to reach 2800 the normal way. I don't argue that you can do it, but how much time and effort will it take?
Why bother with elaborate manipulation?
If someone wants an inflated puzzle rating, they can simply use an engine. Puzzle cheating is not prohibited on Lichess because puzzles are meant for training, not competition. The rating exists only for selecting puzzles that match your level.
You don't even need an engine — just disconnect from the internet, view the solution, reconnect, and repeat it.
So puzzles clearly do not aim for competitive integrity.
@Italiya said in #32:
> This question mark only confuses and creates conditions for manipulation. It serves no purpose. And if you're worried about someone having to beat a bunch of clearly inferior players on their way to reaching their real stats, then you definitely shouldn't do it on a site that allows at least three accounts. Where you'll be playing with 1500 starting points for each account. And, roughly speaking, you'll beat the same player three times, even with a lower rating.
What do multiple accounts have to do with this?
The rules of Lichess explicitly state when multiple accounts are allowed. Sandbagging is prohibited.
So even with three accounts, each one must be played within its intended context (blindfold, berserk-only, gambit practice, etc.).
The question mark exists solely to indicate that a rating is uncalibrated and likely inaccurate. Nothing more.
@Italiya said in #32:
> And if you're stingy with the time and effort, that doesn't mean you should give up all those results you might never have achieved. But that's how sports work—results are hard-won.
That does not mean one should take the longest or most difficult path if it serves no purpose.
In sports, the optimal path is chosen unless the difficulty is part of the competitive nature of the sport — which the rating system is not.
>"@Italiya said in #34:
> Because you explained it from the perspective of someone who doesn't want to spend a lot of time and effort on chess. However, I'll note another amusing point (from a series of three accounts): this site emphasizes extremely long tournaments."
Now you're going off-topic. You bring up unrelated issues, even though your questions about multiple accounts were fully answered here:
https://lichess.org/forum/lichess-feedback/clarification-on-sandbagging?page=27#269
and questions about tournaments were answered here:
https://lichess.org/forum/lichess-feedback/is-the-lichess-marathon-more-dangerous-than-the-olympics?page=3#24
The purpose of chess is not to “spend as much time as possible,” but to play well and win.
Rating serves to reflect skill. Your opponent explained how the system works in reality.
You, on the other hand, build arguments on the assumption that rating is a reward — which is simply wrong and based entirely on subjective perception.
@Italiya said in #37:
> There's no problem here. You can't score 895 goals at once, even if you're the best hockey player; you'll have to score one at a time.
This is not a correct analogy.
If hockey allowed scoring multiple goals at once, players would do it.
Chess rating systems can adjust by large amounts at once — so they do.
The entire point is to use the system that calibrates rating as fast as possible.
If we could instantly assign a rating that matches a player’s level, that would be ideal.
@Italiya said in #39:
> How much longer, a month, a week? It might take a little longer, but at least it eliminates chance and manipulation.
A system should be optimized for the majority.
Eliminating a small inconvenience by creating a much larger one is irrational.
The question mark is a punctuation symbol (?) used at the end of a sentence to indicate a direct question. It signals that the sentence is interrogative, meaning it asks for information or a response.