- Blind mode tutorial
lichess.org
Donate

Does it make sense to use deep engine search?

#1

I got into engines the last days and thought a bit about it.
The question I couldnt answer yet is.

How much sense does it make for a average player like me (17xx ELO in Rapid) to use Stockfish with depth 30 or more.

I came across Komodo Dragon 3 (calls itself a more human player and is aparrently positionally better than stockfish. but weaker)

But again does it make sens to use an engine that gives me a mate in 17 or wouldnt it be better when i win a piece kinda forced in 3-5 moves. This 3-5 moves I can maybe calculate but more for sure not.

Is there an option to run Stockfish here like that? So it doest go for the super sharp stuff and rather gets sails throu it?
Or is there an other engine that does it that way? Or i can put the engine to an ELO level (maybe 400 points above mine so it shows me the more realstic continuation.

I got into engines the last days and thought a bit about it. The question I couldnt answer yet is. How much sense does it make for a average player like me (17xx ELO in Rapid) to use Stockfish with depth 30 or more. I came across Komodo Dragon 3 (calls itself a more human player and is aparrently positionally better than stockfish. but weaker) But again does it make sens to use an engine that gives me a mate in 17 or wouldnt it be better when i win a piece kinda forced in 3-5 moves. This 3-5 moves I can maybe calculate but more for sure not. Is there an option to run Stockfish here like that? So it doest go for the super sharp stuff and rather gets sails throu it? Or is there an other engine that does it that way? Or i can put the engine to an ELO level (maybe 400 points above mine so it shows me the more realstic continuation.
#3

No it doesn't make sense. It makes much more sense to analyse your own games without using an engine at all or just in key positions. However, the problem is, it's very time consuming, which is why most people avoid doing it.

No it doesn't make sense. It makes much more sense to analyse your own games without using an engine at all or just in key positions. However, the problem is, it's very time consuming, which is why most people avoid doing it.
#4

Humans don't generally see mates in 17...nor should they (although that kind of thing used to be a sort of fad way back when). Winning the piece is a sure win (or it should be), so there's generally little point in searching for more.

That does bring up an important distinction though: if you're already winning comfortably, the engine's suggestions become much more disposable than if things are more or less even.

Humans don't generally see mates in 17...nor should they (although that kind of thing used to be a sort of fad way back when). Winning the piece is a sure win (or it should be), so there's generally little point in searching for more. That does bring up an important distinction though: if you're already winning comfortably, the engine's suggestions become much more disposable than if things are more or less even.
#5

@Tschesslee said in #1:

... Or i can put the engine to an ELO level (maybe 400 points above mine so it shows me the more realstic continuation.

Just tell Stockfish to calculate and show multiple lines. Then you see different options (and their evaluation) and not only the best move.

@Tschesslee said in #1: > ... Or i can put the engine to an ELO level (maybe 400 points above mine so it shows me the more realstic continuation. Just tell Stockfish to calculate and show multiple lines. Then you see different options (and their evaluation) and not only the best move.
#6

Try the Lucas Chess. I still use it.
https://lucaschess.pythonanywhere.com/index?lang=en
@Tschesslee By the way, to discover the average rating of players on this site, look at this link.
https://lichess.org/stat/rating/distribution/classical
I picked the longest time control to get the quality move ratings.
Ultrabullet does not produce quality moves so the ratings are not of quality.
The quality of a rating is better reflected by long time controls.

Average ELO ... ?
https://squareoffnow.com/blog/chess-ranking-system/#:~:text=A%20chess%20ranking%20system%20is,either%20go%20up%20or%20down.

'' Every player on Lichess has a separate chess rating for each format
listed above. A rating is a single number that represents a player’s
skill, with higher values indicating higher skill. On Lichess, the rating
system used is Glicko-2, which is a descendant of the ubiquitous Elo
rating system. A player’s rating goes up or down after every game,
depending on if they won or lost, and is updated intuitively: when a
player beats a much higher-rated opponent, their rating increases
more than when they beat a much lower-rated opponent. The mean
rating on Lichess is 1525, with 5% of people below 1000 and 10%
above 2000.'' Quoted from http://www.cs.toronto.edu/~ashton/pubs/maia-kdd2020.pdf

This is how normally a rating declines with age.
https://link.springer.com/article/10.3758/s13428-016-0782-5/figures/2
Since so many players progressing so quickly on lichess, I am left with the assumption that they are either young players or the players are getting assistance (like: learning regularly new things) to reach levels above average so quickly. I'm an elder and I played as far as I can remember. I consider myself an average club player. I don't think I'm a late bloomer. Unless my rating is still increases at my age. I'm obviously not an early bloomer or prodige. I think they are extremely rare. Just take a good look at a profile. ''Rating: 1664.67. You are better than 66.8% of Classical players.'' That percentage to be average would be at 50%. So even on lichess Classical average player is not 1664. It's lower than that an average player.
In your profile it said ...
Rating: 1759.06. Tschesslee is better than 75.2% of Rapid players.

So what you think is average is far, very far from average. Only 24.8% of the players you cannot beat on lichess. And lichess is all over the globle. So think about that skill level you have.

Try the Lucas Chess. I still use it. https://lucaschess.pythonanywhere.com/index?lang=en @Tschesslee By the way, to discover the average rating of players on this site, look at this link. https://lichess.org/stat/rating/distribution/classical I picked the longest time control to get the quality move ratings. Ultrabullet does not produce quality moves so the ratings are not of quality. The quality of a rating is better reflected by long time controls. Average ELO ... ? https://squareoffnow.com/blog/chess-ranking-system/#:~:text=A%20chess%20ranking%20system%20is,either%20go%20up%20or%20down. '' Every player on Lichess has a separate chess rating for each format listed above. A rating is a single number that represents a player’s skill, with higher values indicating higher skill. On Lichess, the rating system used is Glicko-2, which is a descendant of the ubiquitous Elo rating system. A player’s rating goes up or down after every game, depending on if they won or lost, and is updated intuitively: when a player beats a much higher-rated opponent, their rating increases more than when they beat a much lower-rated opponent. The mean rating on Lichess is 1525, with 5% of people below 1000 and 10% above 2000.'' Quoted from http://www.cs.toronto.edu/~ashton/pubs/maia-kdd2020.pdf This is how normally a rating declines with age. https://link.springer.com/article/10.3758/s13428-016-0782-5/figures/2 Since so many players progressing so quickly on lichess, I am left with the assumption that they are either young players or the players are getting assistance (like: learning regularly new things) to reach levels above average so quickly. I'm an elder and I played as far as I can remember. I consider myself an average club player. I don't think I'm a late bloomer. Unless my rating is still increases at my age. I'm obviously not an early bloomer or prodige. I think they are extremely rare. Just take a good look at a profile. ''Rating: 1664.67. You are better than 66.8% of Classical players.'' That percentage to be average would be at 50%. So even on lichess Classical average player is not 1664. It's lower than that an average player. In your profile it said ... Rating: 1759.06. Tschesslee is better than 75.2% of Rapid players. So what you think is average is far, very far from average. Only 24.8% of the players you cannot beat on lichess. And lichess is all over the globle. So think about that skill level you have.
#7

"How much sense does it make for a average player like me (17xx ELO in Rapid) to use Stockfish with depth 30 or more."

No sense at all. A Stockfish level 30 is unbeatable. Level 20 is GM strength. Level 10 will beat all this forum. So yeah, look for another engine.

"How much sense does it make for a average player like me (17xx ELO in Rapid) to use Stockfish with depth 30 or more." No sense at all. A Stockfish level 30 is unbeatable. Level 20 is GM strength. Level 10 will beat all this forum. So yeah, look for another engine.
#8

It might make sense if you can get to see the position in the search tree that SF found on the main PV branch with the score difference that makes it the winner branch among all other branches of the legal subtree it searched.

but the tournament specifications for engine do not specify much in terms of engine usability as chess analytical tools for humans**. So it is up to each engine development group to divert dev-hours for that on their own agenda.. But if in a hurry, and not many dev-hours, the development talent will have a tendency to drop such ability to interpret and use the tool for human analysis..

I you had access to that information on routine basis, yes, very deep searches might make sense... At least you would have all the keys not just the scores.. you could see the position features, that SF detected as advantageous, that we not yet visible to it before in the search (i mean not as shallow).

Now for some editorial comments (n-th iteration for those having read some of my stuff, before).

** it seems like engine tournament have created their own universe of purpose, happy to beat the crap out of each other for some ELO from their anemic gene pools, whatever high ELO would mean in terms of chess (that is a despaired rant about the state of engine competition as human helper for human chess).

Maybe there could be a competitive engine tournament branch where the focus in not just ELO, but ELO with full output for human analysis... (and other criteria). That way ELO in engine pool with evolving specifications to have more than one use for the engine, might become imaginable.

Why do we keep such engine race.... really.. since we have no means to know if they are not fighting in their own corner of chess space, us, not having any measure for their coverage of the complete tree... or other true accuracy, that is not self-referring only.

ELO is pool dependent, and represents competition within the specifications.. there is nothing preventing a tightening of the specification, or other changes if that is what we would like to see engine get better at..... just some lack of imagination perhaps? i ask.

It might make sense if you can get to see the position in the search tree that SF found on the main PV branch with the score difference that makes it the winner branch among all other branches of the legal subtree it searched. but the tournament specifications for engine do not specify much in terms of engine usability as chess analytical tools for humans**. So it is up to each engine development group to divert dev-hours for that on their own agenda.. But if in a hurry, and not many dev-hours, the development talent will have a tendency to drop such ability to interpret and use the tool for human analysis.. I you had access to that information on routine basis, yes, very deep searches might make sense... At least you would have all the keys not just the scores.. you could see the position features, that SF detected as advantageous, that we not yet visible to it before in the search (i mean not as shallow). Now for some editorial comments (n-th iteration for those having read some of my stuff, before). ** it seems like engine tournament have created their own universe of purpose, happy to beat the crap out of each other for some ELO from their anemic gene pools, whatever high ELO would mean in terms of chess (that is a despaired rant about the state of engine competition as human helper for human chess). Maybe there could be a competitive engine tournament branch where the focus in not just ELO, but ELO with full output for human analysis... (and other criteria). That way ELO in engine pool with evolving specifications to have more than one use for the engine, might become imaginable. Why do we keep such engine race.... really.. since we have no means to know if they are not fighting in their own corner of chess space, us, not having any measure for their coverage of the complete tree... or other true accuracy, that is not self-referring only. ELO is pool dependent, and represents competition within the specifications.. there is nothing preventing a tightening of the specification, or other changes if that is what we would like to see engine get better at..... just some lack of imagination perhaps? i ask.
#9

It probably doesn't matter much if you use depth 20 or 30. But an engine is a good tool to help you with your analysis and find moves you may have missed.

But again does it make sens to use an engine that gives me a mate in 17 or wouldnt it be better when i win a piece kinda forced in 3-5 moves

The engine can tell you if that piece winning combo really works or not. Sometimes the engine will find some moves you'll understand. Sometimes, the moves will be incomprehensible. You also can't let the engine do all the work for you. You have to use common sense.

It probably doesn't matter much if you use depth 20 or 30. But an engine is a good tool to help you with your analysis and find moves you may have missed. >But again does it make sens to use an engine that gives me a mate in 17 or wouldnt it be better when i win a piece kinda forced in 3-5 moves The engine can tell you if that piece winning combo really works or not. Sometimes the engine will find some moves you'll understand. Sometimes, the moves will be incomprehensible. You also can't let the engine do all the work for you. You have to use common sense.
#10

@Frogster64 said in #9:

. Sometimes, the moves will be incomprehensible.

The mysteries of engines.... well, with full output as a norm... We would be surprised that it might not be so.

@Frogster64 said in #9: > . Sometimes, the moves will be incomprehensible. The mysteries of engines.... well, with full output as a norm... We would be surprised that it might not be so.

This topic has been archived and can no longer be replied to.