Chessbase Let's Check

[jshriver]

I bought Fritz 17 a few years back and still enjoy it a lot. What initially caught my eye was the Let's Check feature and the ability to contribute your own CPU time to their system which is wonderful.

Tonight while doing some analysis thought I'd give it a try again and really turned off by the data it has now. I like that it allows viewpoints from different engines. But many times tonight while contributing SF 17.1 I would see other entries from just today with people using SF 11, Komodo 18, much older version of Fritz. Not only are engines horribly outdated, a lot of the entries I see are shallow depths like 18.

Not sure what the point is anymore. If a user can generate an analysis with a modern engine in less time and deeper than it takes to even ask Let's Check. The data is simply pointless.

Sorry for the rant, just curious what others think about it.
Thanks for listening.

[jkominek]

It was a great idea that failed. Frederic Friedel at one point admitted that the critical weakness was that grandmasters won't share their analyses, for want of not losing competitive edge. The problems you mention are additional weaknesses.

The idea done right is Chess Cloud Database Query Interface https://www.chessdb.cn/queryc_en/?rnbqk ... -%200%201

Using a current version of stockfish keeps it relevant. I believe the maintainer updates the engine whenever there is a new major release. Because stockfish now calibrates a centipawn to probability of winning, it is possible to maintain consistency. The opening tree is also larger than that offered by Let's Check.

My top complaint is that the author does not automatically feed in recent top-level games (up to say move 40) to keep the database reflective of current theory. I wish he would do that.

[Leto]

I bought Fritz 17 a few years back and still enjoy it a lot. What initially caught my eye was the Let's Check feature and the ability to contribute your own CPU time to their system which is wonderful.

Tonight while doing some analysis thought I'd give it a try again and really turned off by the data it has now. I like that it allows viewpoints from different engines. But many times tonight while contributing SF 17.1 I would see other entries from just today with people using SF 11, Komodo 18, much older version of Fritz. Not only are engines horribly outdated, a lot of the entries I see are shallow depths like 18.

Not sure what the point is anymore. If a user can generate an analysis with a modern engine in less time and deeper than it takes to even ask Let's Check. The data is simply pointless.

Sorry for the rant, just curious what others think about it.
Thanks for listening.

One thing that is disheartening about Fritz's Let's Check is that an 80 ply Stockfish 17.1 analysis can be replaced by a Fritz 18 depth 25 analysis. That 80 ply analysis? Gone like if it never existed, and worse replaced by a much weaker analysis. Maybe I'm exaggerating a little but this is the impression I have as a previous user of Let's Check.

[jshriver]

One thing that is disheartening about Fritz's Let's Check is that an 80 ply Stockfish 17.1 analysis can be replaced by a Fritz 18 depth 25 analysis. That 80 ply analysis? Gone like if it never existed, and worse replaced by a much weaker analysis. Maybe I'm exaggerating a little but this is the impression I have as a previous user of Let's Check.

That brings up an interesting point, even for my own project. I think Let's Check list the top 3 highest depths, so not seen your example specifically. But if a much older version hit a higher depth it would replace it. Depths in themselves aren't necessarily the best indicator either. What about LC0? In my own testing I had to make a special allowance for LC0 because while SF might hit depth 38, to get similiar results or more equal LC0 might only hit depth 12 or 14. It's an interesting problem to put into perspective.

[Leto]

One thing that is disheartening about Fritz's Let's Check is that an 80 ply Stockfish 17.1 analysis can be replaced by a Fritz 18 depth 25 analysis. That 80 ply analysis? Gone like if it never existed, and worse replaced by a much weaker analysis. Maybe I'm exaggerating a little but this is the impression I have as a previous user of Let's Check.

That brings up an interesting point, even for my own project. I think Let's Check list the top 3 highest depths, so not seen your example specifically. But if a much older version hit a higher depth it would replace it. Depths in themselves aren't necessarily the best indicator either. What about LC0? In my own testing I had to make a special allowance for LC0 because while SF might hit depth 38, to get similiar results or more equal LC0 might only hit depth 12 or 14. It's an interesting problem to put into perspective.

No it's more complicated than that. A newer analysis can replace an older but much deeper analysis and you can use pretty much any engine to replace the much stronger engine analysis. So people can use Fritz 1 to replace Stockfish 17 analysis just to mess with the system if they wanted to. With three old weak engines you can easily replace three strong engine analysis.