Recent Posts

Nice to see another post from you! That looks like a very solid study plan, and I'm looking forward to hearing how you go!

Nodes per second isn't an ideal metric for engine strength. For example when AlphaZero was first released, and beat the version of stockfish that was available at the time, it did so with far less nodes per second than stockfish because the superior evaluation allowed it to look at fewer nodes and still be very strong.

I would guess newer versions of stockfish that are using NNUE based eval may also be slower in NPS than older versions, but are also clearly much stronger, although I havent run an NPS test recently to test that.

In terms of our own machines, we have some faster machines than were mentioned in the previous post, although not massively faster, and we also have less of the slower machines in the mix, so on average you have access to more computing power now than in 2018. However as you say, stockfish is already insanely strong compared to human players, and it will be rare to notice the differences. The main advantage of more NPS is that you'll be able to analyse for less time for equally strong results.

I've gone into more detail here, but basically I wonder if a) my custom set set-up makes sense, and b) any suggestions for improvement. My goal, since I am---ahem---very adult, is not to improve my speed, per se (I only play rapid and classical), but to build up my pattern-recognition. So, I created a new merged set comprising 50% each of two sets, one with outcome type set to mate and the other non-mates, and the following settings:

• Blitz Rating type, 1000-1100, 1-2 moves, 100+ total attempts, 4-5 star, winning problems, any color to move, any number of pieces
• Filter: problems I have never solved. Treat loss of rating due to time as wrong. Include unsolved problems, 0-15 secs solve time
• Blitz rating adjustment, sorted (looping), sort by problem rating

(I am using the combination of number of attempts by all users + rating to limit the number of problems to app. 500 each.)

Then I work through the merged set until there are no problems left (in one of the sets, anyway).

If I understand how this is working, as I go through the set, problems are dropped if I solve them in under 15 seconds unless I my time is too slow according to the time adjustment.

Does what I have make sense (I'm usually a relatively bright person, but some of the CT custom set options are still murky)? How might it be improved?

Also, I am wondering if I should also do thematic puzzle sets and, if so, how much of my time should be spent on that, proportionally?

Thanks in advance!

I started with the plan above, but quickly realized the ratings setting was too low even for me, and having only 25% non-mates in the mix wasn't enough. So I created a new merged set comprising 50% each of two sets that only differed in their outcome type:

• Blitz Rating type, 1000-1100, 1-2 moves, 100+ total attempts, 4-5 star, winning problems, any color to move, any number of pieces
• Filter: problems I have never solved. Treat loss of rating due to time as wrong. Include unsolved problems, 0-15 secs solve time
• Blitz rating adjustment, sorted (looping), sort by problem rating

I am using the combination of number of attempts by all users + rating to limit the number of problems to app. 500 each.

This resulted in a merged set of about 1000 problems. I looped through until one of the sets ran out, and am now starting another cycle with a set like the above but with the rating level set to 1100-1150, and the number of attempts adjusted to 30+.

I feel good about this approach to non-thematic training focused on pattern recognition, though I am unsure if, or how much, I should do thematic sets.

Is this still true in 2024 or did the speed increase? My machine reaches aprox. the same speed with a 11th Gen Intel(R) Core(TM) i7-11700K @ 3.60GHz and 32Gb RAM. I think it is an advantage not having to use my own processing power if this runs at the same speed. Chessify is much faster but comes with another price tag. As a matter of fact I wonder if you can get more out of this extra Chessify processing power or if this is another marketing trick. I don't think that the extra speed gives equal better moves. Stockfish is so strong that I cannot imagine that (for the casual player not competing for the world cup) more speed ithen 10 M Nodes per sec. is necessary to play a good game.       

Hi everyone!

It's been over four years since I posted on this forum. In that time, a pandemic happened, and many other things beside.

My last OTB tournament was in the fall of 2019, just before the COVID-19 outbreak in early 2020.

In March this year, I played my first tournament in 5 years. It was a rapid tournament (15+10) in another city. I finished in 3rd place with 4.5/6 and won some money. I was the 8th seed, so a pretty good result. I manged to get a key win in the 6th round against the 2nd seed to secure 3rd place and the prize money.

I have also decided to study again in a more serious and organized manner. I read a small e-book by GM Noel Studer. Rarely have I seen such a no-nonsense work by a grandmaster. Basically his advice is: practice tactics, play games and analyze them. The rest should be a very small minority of your time spent on chess. Why? Because at the amateur level, opening prep or deep endgame knowledge rarely means anything - games are decided by tactical mistakes in the middle game.

This really goes along what I've said multiple times on this blog. Tactical training is the fundamental skill needed to improve for anyone below master level. Ironically enough I still bought many opening books over the years. I even tried some courses on Chessable. And you know what? I stopped all opening training at this point. What I've come to realize is that, as GM Noel Studer points out in his e-book, GMs make opening courses using their own prep and ideas. It's more efficient and profitable for them to do so. But what I was finding is that the ideas in their courses make zero sense to me.

So my study plan going forward is 75% tactics + playing & analyzing my own games, with the last 25% on various topics like strategy, defense, endgame, etc. I'm currently using a custom set on Chess Tempo, 1300-1800 rating range, slowly working my way back up to my 1500-2000 set. I try to play 2 rapid 15+10 games, 4 days a week, and analyze each one of them. I'm using "zen mode" on lichess which I really like. No ratings, no names. Just a board and the clocks. I have no idea what my rating is, and I don't want to know. My goal is not to reach a certain rating per se; I just want to improve. As per GM Noel Studer's book, I have set a slightly unrealistic goal of reaching expert in OTB play in my local chess federation within 5 years. I am currently at 1684, so I need to gain 316 rating points in 5 years.

This year, I will play in 5 OTB tournaments: 3 local ones, and 2 in other Canadian cities. This should give me 29 OTB games for the year (increasing my global OTB chess experience by 30% in one year, as I have about 100 games of experience since starting tournaments in 2013), and the opportunity to visit two very nice cities in my country, Halifax and Banff. I will be posting some of these tournament games here with my analysis.

Also, I am playing exclusively 1. b3 and 1...b6 for openings. I am learning as I go along, and I've been having some decent success (notably the rapid tournament last month). I've found that if I don't know what I'm doing in the opening, I tend to think about the position instead of playing move on auto-pilot. As an amateur, my opening knowledge and my memory are too poor to allow me to learn precise lines. Whenever I think I'm using prep, I'm actually just mixing up lines and ending up with garbage positions.  The advantage of these opening systems is also that I can take my opponents out of book pretty quickly, and just play a game of chess. The psychological effect of surprise is also a pretty good weapon. Let's throw that opening book out the window and play some chess!

Yes, it is a pity to disable this one, but given being up the exchange is almost always enough to be marked correct, not having as an alt is unfortunately a fatal flaw I think in this one, so I've disabled it now.

A zwischenzug allows white to win a full rook rather than an exchange, as my comment from 2016 showed.

At the time, I agreed with @fifty's comment from 2014. Though harsh, the fail is deserved. (This may be my boomer mentality.) @fifty's comment has 4 upvotes and the problem quality rating is 3.5 stars. I think the quality rating is people acknowledging that the problem is instructive, emphasizing always looking for zwischenzugs.

The pre-move earns black a knight advantage.

The Hanging Piece tactic 1.Rxe7 only returns material to even and fails due to a perpetual check counterattack. That is 15% of mistakes.

Impulsively grabbing a rook with 2.Qxa8 gives white a clean exchange advantage that Toga under-evaluated. But there is a second zwischenzug, using a mate threat for tempo, that wins a bishop 2.Rxe7 before winning the rook 3.Qxa8.

 The blitz rating is 2090 compared to standard rating of 1690.

There are multiple complaints, with upvotes, in the past few years, which may represent sour grapes, or a different attitude among younger blitz players than among people like @fifty and me.

I hate disabling an educational problem, but you recognize the need for amity. How do you balance it in this case?

Richard, I just noticed, the site now sorts the openings by alphabet! Great!

Or if Chesstempo could ad a similar option like " >1 candidate moves", it would also help. Or maybe an offline tool would be better suited for this. That can be use before importing the pgn into Chesstempo.
I'm sure a good programmer can easily make such a tool.
I think it would be very helpful for building repertoires.