Puzzle Ratings Accuracy

I am wondering about the accuracy of puzzle ratings and the accompanying rating that a player receives for doing puzzles... Are these ratings accurate, ie. reflective of how a player might fare in a match? For example, If someone is rated at 2000 in the puzzles on lichess, does that mean they are relatively close to that level of ELO?

I started playing chess 4-5 months ago and for a while I was only playing blitz games. At first, I played only on chess.com. I fared poorly - slowly climbed from a 450 rating to a 600 over the course of a couple of months. In classical play, I seemed to be floating around 800 or 900.

A month or so ago I got hip to the notion that learning tactics and doing puzzles is really the only way to get better (for a complete beginner). Since then, I have committed myself to doing more puzzles and playing less blitz.

Right now my puzzle "ratiing" is floating around 1650. Its been several weeks since I played a classical game. I'm wondering if I started playing classical games again, can I expect to see a dramatic improvement in my play (jumping from a 900 rating to 1500/1600)? Admittedly, this doesn't feel right - I do not feel I have improved by 700 points in a month. Like I said, last I played, I was playing around 800/900.

Furthermore, how do they determine the rating of a puzzle? Is it the same across the different websites/apps?

Yes, it is the one and only measure of playing strength. I will contact the FIDE to convert into Elo.

Seriously: it is training, no zeitnot (except yesterday when I was solving and they announced a server maintenance ;) ), it is easy cheatable, doing it right it is just like practicing free-kicks, corners and penalties... so there’s only little correlation to playing strength overall needed for whole games.

Puzzle ratings are definitely not related to actual ratings; I’m a 1797 glicko puzzle rating, but only a 73 ecf (1247 FIDE)

I like Lichess puzzles but I wish there were more 2500+ puzzles. I get a lot of 2000 rated puzzles and it makes getting to higher ratings even harder (but I guess that's the point right? lol)

Puzzle rating is determined by the success rates of players attempting the puzzle. So if 50% of players at rating 1500 fail a puzzle, that puzzle is rated 1500 (approximately). It is a self-contained system that is independent of other ratings, and only measures the degree of difficulty of puzzles that you are capable of completing, on average. Puzzles are great for training purposes, and working on puzzles ought to improve your ratings in other categories, but it isn't a one-to-one correspondence by any stretch of the imagination.

At a rating level of 800-900 in Classical, you can probably improve most effectively by just slowing down and blunder-checking every move. Don't leave pieces hanging and don't miss mate-in-one (in either direction), and you can probably improve your rating by 300+ points.

I think the puzzle ratings are to help measure improvement but don't mean much beyond that.

There are many people with huge differences between puzzle rating and classical.
My puzzle rating is 2250, my classical 1880.
But there are also a lot of people whose puzzle rating is quite similar to their other ratings.
Try it and you'll see.

Thank you everybody for chiming in. Thanks for this information.

Yes, this is what I figured. From what I observe on other users' profiles, puzzle ratings tend to be a bit higher than classical ratings, but not always. Of course I was hopeful that somehow I dramatically improved my play in a few weeks simply from playing more puzzles, but I assumed it was too good to be true. Of course the only way to really truly measure how I am playing these days is to play some rated classical games! Something tells me I still be a bit blunder-prone haha...

One additional question I have re: tactics (or perhaps its not tactics at all, but positional play?): when I play higher rated opponents I tend to play pretty well for the first ten or so moves, but once it gets into the middle game, I feel a bit lost and uncertain about the "correct" move. I almost always give up a good position at this point in the game. I would love to practice more puzzles that are specific to this portion of the game. Are there any places that allow you to practice chess puzzles that focus on the different stages of the game (in my case, moves 10-20, or early middle game)?

I think chesstempo is a nice tool for this trainning. With a paid account you can create lists of problems with N pieces or between move X and Y for example.

But If you allow me a different opinion, I believe that for the weakness you mentioned, the best way to improve is reading some books of annotated games (e.g. Smyslov selected games). You'll improve faster because most positions dont have a tactical combination at your disposal, rather they will demand the right plan (where to go with your pieces) in order to accumulate small advantages until the tactical oportunity arises.