- Blind mode tutorial
lichess.org
Donate

Opening Inaccuracies Tool

#2

Great work, as always. I have notes :)

  • There should be a way to toggle unrated games. I have a lot of games that are just testing LiChess Tools and I guess for most people the games they care about are just the rated ones and the others are trying out openings or playing when they are tired and don't care.
  • I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green).
  • the list of 112 moves is not really useful to me. Where do I start, when do I stop? I think a good idea is to add a slider that will try to restrict the numbers of patterns detected (kind of like with Splunk, if you were used it). So still 112 bad moves, but try to cluster them together in just N categories, thus more likely to find the underlying common issue that must be addressed. You study one issue, then you move on to the next or increase the number N to get more granular. A move played in a single game does not a pattern make.
  • and finally, I think you should leverage the computer resources of the person doing the analysis. Move some if not all of the analysis on the client. Thus people can use whatever depth or engine settings they want.
Great work, as always. I have notes :) - There should be a way to toggle unrated games. I have a lot of games that are just testing LiChess Tools and I guess for most people the games they care about are just the rated ones and the others are trying out openings or playing when they are tired and don't care. - I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green). - the list of 112 moves is not really useful to me. Where do I start, when do I stop? I think a good idea is to add a slider that will try to restrict the numbers of patterns detected (kind of like with Splunk, if you were used it). So still 112 bad moves, but try to cluster them together in just N categories, thus more likely to find the underlying common issue that must be addressed. You study one issue, then you move on to the next or increase the number N to get more granular. A move played in a single game does not a pattern make. - and finally, I think you should leverage the computer resources of the person doing the analysis. Move some if not all of the analysis on the client. Thus people can use whatever depth or engine settings they want.
#3

@TotalNoob69 said in #2:

Great work, as always. I have notes :)

  • There should be a way to toggle unrated games. I have a lot of games that are just testing LiChess Tools and I guess for most people the games they care about are just the rated ones and the others are trying out openings or playing when they are tired and don't care.
  • I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green).
  • the list of 112 moves is not really useful to me. Where do I start, when do I stop? I think a good idea is to add a slider that will try to restrict the numbers of patterns detected (kind of like with Splunk, if you were used it). So still 112 bad moves, but try to cluster them together in just N categories, thus more likely to find the underlying common issue that must be addressed. You study one issue, then you move on to the next or increase the number N to get more granular. A move played in a single game does not a pattern make.
  • and finally, I think you should leverage the computer resources of the person doing the analysis. Move some if not all of the analysis on the client. Thus people can use whatever depth or engine settings they want.

Oh that is interesting, yes, I think I will include controls for importing games i.e. the date range, time control, rated and others, it will allow the user to hone in more.

There is a filter that allows to you change the settings, but I am going to make this permanently visible, as yes. We want to narrow down based on co loss, ply and book moves (this is already there, but you need to click on the filter button.

Interesting point and yes, for my repertoire builder rinha e creates the functionality to allow for the usage of local UCI engines, so I think I will replicate this functionality and allow the user to leverage their own machines (stronger and faster than running in a web browser)

Thanks for the comments, this is exactly the feedback I was looking for.

@TotalNoob69 said in #2: > Great work, as always. I have notes :) > > - There should be a way to toggle unrated games. I have a lot of games that are just testing LiChess Tools and I guess for most people the games they care about are just the rated ones and the others are trying out openings or playing when they are tired and don't care. > - I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green). > - the list of 112 moves is not really useful to me. Where do I start, when do I stop? I think a good idea is to add a slider that will try to restrict the numbers of patterns detected (kind of like with Splunk, if you were used it). So still 112 bad moves, but try to cluster them together in just N categories, thus more likely to find the underlying common issue that must be addressed. You study one issue, then you move on to the next or increase the number N to get more granular. A move played in a single game does not a pattern make. > - and finally, I think you should leverage the computer resources of the person doing the analysis. Move some if not all of the analysis on the client. Thus people can use whatever depth or engine settings they want. Oh that is interesting, yes, I think I will include controls for importing games i.e. the date range, time control, rated and others, it will allow the user to hone in more. There is a filter that allows to you change the settings, but I am going to make this permanently visible, as yes. We want to narrow down based on co loss, ply and book moves (this is already there, but you need to click on the filter button. Interesting point and yes, for my repertoire builder rinha e creates the functionality to allow for the usage of local UCI engines, so I think I will replicate this functionality and allow the user to leverage their own machines (stronger and faster than running in a web browser) Thanks for the comments, this is exactly the feedback I was looking for.
#4

@TotalNoob69 said in #2:

I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green).

Yeah, I will add some interactivity here, as you are right, right now it is just a tool top, but there is not reason why clicking on. Circle cannot load the position. Good idea.

@TotalNoob69 said in #2: > I love the abacus chart, although it is very confusing as well, as I thought the little circles can be clicked individually. I will try to add a similar chart to the computer analysis, see if it yields any insight. If I understand correctly, the bubbles represent the PV computer evaluations, then the colored one is yours and it can be bad (red) or good (green). Yeah, I will add some interactivity here, as you are right, right now it is just a tool top, but there is not reason why clicking on. Circle cannot load the position. Good idea.
#5

nice tool, knowing the inaccuracy out of the opening can help, but the follow up question is does it get punished hard or does it still lead to favorable results due to the nature of the move, how double edged it is, how it affects overall position etc.

nice tool, knowing the inaccuracy out of the opening can help, but the follow up question is does it get punished hard or does it still lead to favorable results due to the nature of the move, how double edged it is, how it affects overall position etc.
#6

@g6firste6second said in #5:

nice tool, knowing the inaccuracy out of the opening can help, but the follow up question is does it get punished hard or does it still lead to favorable results due to the nature of the move, how double edged it is, how it affects overall position etc.

Oh very good point, I should add the result statistics, that would be super interesting and useful. As I might have a move that stockfish deems as bad, but in practical play I am winning loads.

@g6firste6second said in #5: > nice tool, knowing the inaccuracy out of the opening can help, but the follow up question is does it get punished hard or does it still lead to favorable results due to the nature of the move, how double edged it is, how it affects overall position etc. Oh very good point, I should add the result statistics, that would be super interesting and useful. As I might have a move that stockfish deems as bad, but in practical play I am winning loads.
#7

I would be more interested in a tool which suggests moves from Lichess' Opening Explorer API, since if humans aren't finding and understanding moves I sure as heck am not going to understand the moves.

I would be more interested in a tool which suggests moves from Lichess' Opening Explorer API, since if humans aren't finding and understanding moves I sure as heck am not going to understand the moves.
#8

@Toadofsky said in #7:

I would be more interested in a tool which suggests moves from Lichess' Opening Explorer API, since if humans aren't finding and understanding moves I sure as heck am not going to understand the moves.

That is true, and I wish I would use the Lichess API, but there are rating limiting considerations. I have been thinking of leveraging Maia in some shape or form.

I think I need to build my own opening database.

@Toadofsky said in #7: > I would be more interested in a tool which suggests moves from Lichess' Opening Explorer API, since if humans aren't finding and understanding moves I sure as heck am not going to understand the moves. That is true, and I wish I would use the Lichess API, but there are rating limiting considerations. I have been thinking of leveraging Maia in some shape or form. I think I need to build my own opening database.
#9

Great Work! I love the focus on identifying recurring patterns and habits rather than just one-off blunders. It's so easy to keep playing the same okay moves without realizing they are holding us back. Though with a tool like this, I'm a bit worried we'll all start playing like Stockfish in the opening soon ;P Thanks for sharing this tool and your journey!

Great Work! I love the focus on identifying recurring patterns and habits rather than just one-off blunders. It's so easy to keep playing the same okay moves without realizing they are holding us back. Though with a tool like this, I'm a bit worried we'll all start playing like Stockfish in the opening soon ;P Thanks for sharing this tool and your journey!
#10

@desillusionist said in #9:

Great Work! I love the focus on identifying recurring patterns and habits rather than just one-off blunders. It's so easy to keep playing the same okay moves without realizing they are holding us back. Though with a tool like this, I'm a bit worried we'll all start playing like Stockfish in the opening soon ;P Thanks for sharing this tool and your journey!

Yeah, this was my goal for when I dreamt up this tool, to find patterns,

That is interesting thing. I was using this for myself and there were several inaccuracies according to stockfish, but those are lines that I like as it is easier to play from a human perspective, so one still needs to make the choice.

I am going to add WDL statistics, so you can pair that with what stockfish says.

@desillusionist said in #9: > Great Work! I love the focus on identifying recurring patterns and habits rather than just one-off blunders. It's so easy to keep playing the same okay moves without realizing they are holding us back. Though with a tool like this, I'm a bit worried we'll all start playing like Stockfish in the opening soon ;P Thanks for sharing this tool and your journey! Yeah, this was my goal for when I dreamt up this tool, to find patterns, That is interesting thing. I was using this for myself and there were several inaccuracies according to stockfish, but those are lines that I like as it is easier to play from a human perspective, so one still needs to make the choice. I am going to add WDL statistics, so you can pair that with what stockfish says.