Oh, also...should there be a user confidence threshold for this built in? Like for instance I could skip all puzzles that have less than +10 or a negative rating etc. since no one will likely want to review flawed puzzles? It might be really tricky to identify the theme in a flawed puzzle as well.
#10 I suggest CSV or some other easily parseable format. I doubt that any one person is capable of compiling a list of all the tags that could make sense. Maybe looking at other sites and programs would provide some inspiration...
#11 Why would a puzzle be limited to one theme? Maybe add a "cooked" tag to identify such puzzles, I don't know.
#11 Why would a puzzle be limited to one theme? Maybe add a "cooked" tag to identify such puzzles, I don't know.
Looks like my free time is now taken up. Ill do a bunch and send it in. Expect an email with 100 or so done. I will do thousands if it turns out to be a good idea.
9,
I like your idea in last paragraph, except, that a voting system is not really that bad if the voters are strong players perhaps with a rating of 1900 and up, bullet.
A multiple select of tactics theme can be used. Example.
Fen1
[ ] deflection
[ ] pin
[ ] discovered check
And others
The voter may select deflection and discovered check for example.
The votes in each fen per tactics theme are then analyzed. If the fen received 2 or more votes in deflection and there are 20 percent and above of voters in that tactics theme then that deflection can be one of the regular tag of that certain fen.
Of course the number 2 and 20 percent are just an initial guess.
I am willing to help in this project, perhaps a team will be created and members will be given specific tasks.
Someone will handle puzzle 1 to 10 for example to avoid work duplication if voting is not used.
I like your idea in last paragraph, except, that a voting system is not really that bad if the voters are strong players perhaps with a rating of 1900 and up, bullet.
A multiple select of tactics theme can be used. Example.
Fen1
[ ] deflection
[ ] pin
[ ] discovered check
And others
The voter may select deflection and discovered check for example.
The votes in each fen per tactics theme are then analyzed. If the fen received 2 or more votes in deflection and there are 20 percent and above of voters in that tactics theme then that deflection can be one of the regular tag of that certain fen.
Of course the number 2 and 20 percent are just an initial guess.
I am willing to help in this project, perhaps a team will be created and members will be given specific tasks.
Someone will handle puzzle 1 to 10 for example to avoid work duplication if voting is not used.
To address, there are 60120 puzzles.
To be specific, there's no point submitting small batches at all because they are not useful. Also, the tags you use in your batch would be different from the tags used in other batches, so fragmented attempts aren't compatible with others.
Someone would have to lead such a project because it's unlikely for any sole individual with the goodwill, willpower and strength in tactics to go through all of it. And it's not me.
Re: "A voting system is not really that bad if the voters are strong players perhaps with a rating of 1900 and up, bullet."
I'd like to see you get 1. players with 1900+ ratings, 2. does lots of puzzles frequently, 3. with the motivation, determination and drive to review tons of puzzles, and 4. do it all for free.
When I watch other players do tactics puzzles, they either only do a few, when they do think through the puzzle and give feedback by voting, or they belong to the others who does a lot in bulk, and doesn't care about spending their time on voting. Such a voting system would be virtually useless because only a few people would use it, which in turn makes even fewer people want to use it.
To be specific, there's no point submitting small batches at all because they are not useful. Also, the tags you use in your batch would be different from the tags used in other batches, so fragmented attempts aren't compatible with others.
Someone would have to lead such a project because it's unlikely for any sole individual with the goodwill, willpower and strength in tactics to go through all of it. And it's not me.
Re: "A voting system is not really that bad if the voters are strong players perhaps with a rating of 1900 and up, bullet."
I'd like to see you get 1. players with 1900+ ratings, 2. does lots of puzzles frequently, 3. with the motivation, determination and drive to review tons of puzzles, and 4. do it all for free.
When I watch other players do tactics puzzles, they either only do a few, when they do think through the puzzle and give feedback by voting, or they belong to the others who does a lot in bulk, and doesn't care about spending their time on voting. Such a voting system would be virtually useless because only a few people would use it, which in turn makes even fewer people want to use it.
Right. Makes a lot of sense. Also I had no aspirations of being capable of handling every single puzzle, I knew there were thousands. What I meant was I'd be willing to organize such a task.
As for #12, while I disagree that any one person could handle the entirety of this project, I think a reasonably thorough set of tags could be compiled by a single person. There's absolutely no reason to tag every puzzle by every single possible tag that could apply. A reasonable set of categories that would benefit grouping is all that is needed. For example, the obvious two tags that could applied to every puzzle would be:
win
mate
Is the puzzle about winning material / achieving a winning position, or is the puzzle about reaching mate? This single pair of tags would allow us to filter out likely almost half of the puzzles we've ever completed. Very good tags.
A second set of tags could be "fork" and "discovery." Do you want to work on forking tactics? Great, we can now sort out a large number of puzzles you've done that involve only forking tactics or discovered attacks. Also in that same line of thought would be "skewer" and "pin." How about "smother" and now you can sort out all the smothered mate puzzles to really familiarize yourself with positions where those work. Another mentioned was "deflection" which is also a good tactic to practice, identifying when you can force a piece away from the piece or square it's guarding to win.
Eventually the benefit of a tag is outweighed by the effort of tagging puzzles with it. For example, there's no point in tagging things "rim knight" because even though sure, it might be useful to examine positions in which having a knight on the rim might be beneficial rather than bad development, there is not much benefit to sorting puzzles with such a tag. The fact is, it would likely take you going through and playing virtually all the puzzles to accumulate enough games to make that a worthwhile sorting term. Likewise with "f-pawn" or "g-pawn" or the like. Are you really going to try to sort every puzzle that involves pushing a very specific pawn? Again, this could be helpful for positions that result a lot from certain openings to get in sync with the timing, but the sheer amount of puzzles you'd have to play for this to be a relevant criterion is absurd.
I'm pretty sure the tags could be kept to a trim number of say 15-20 tags and still be highly beneficial for virtually every puzzle reviewer who has at least done a couple hundred puzzles. And by combining tags players who are looking to review a very specific sort of puzzle could easily filter out those that don't involve all the criteria.
As for #12, while I disagree that any one person could handle the entirety of this project, I think a reasonably thorough set of tags could be compiled by a single person. There's absolutely no reason to tag every puzzle by every single possible tag that could apply. A reasonable set of categories that would benefit grouping is all that is needed. For example, the obvious two tags that could applied to every puzzle would be:
win
mate
Is the puzzle about winning material / achieving a winning position, or is the puzzle about reaching mate? This single pair of tags would allow us to filter out likely almost half of the puzzles we've ever completed. Very good tags.
A second set of tags could be "fork" and "discovery." Do you want to work on forking tactics? Great, we can now sort out a large number of puzzles you've done that involve only forking tactics or discovered attacks. Also in that same line of thought would be "skewer" and "pin." How about "smother" and now you can sort out all the smothered mate puzzles to really familiarize yourself with positions where those work. Another mentioned was "deflection" which is also a good tactic to practice, identifying when you can force a piece away from the piece or square it's guarding to win.
Eventually the benefit of a tag is outweighed by the effort of tagging puzzles with it. For example, there's no point in tagging things "rim knight" because even though sure, it might be useful to examine positions in which having a knight on the rim might be beneficial rather than bad development, there is not much benefit to sorting puzzles with such a tag. The fact is, it would likely take you going through and playing virtually all the puzzles to accumulate enough games to make that a worthwhile sorting term. Likewise with "f-pawn" or "g-pawn" or the like. Are you really going to try to sort every puzzle that involves pushing a very specific pawn? Again, this could be helpful for positions that result a lot from certain openings to get in sync with the timing, but the sheer amount of puzzles you'd have to play for this to be a relevant criterion is absurd.
I'm pretty sure the tags could be kept to a trim number of say 15-20 tags and still be highly beneficial for virtually every puzzle reviewer who has at least done a couple hundred puzzles. And by combining tags players who are looking to review a very specific sort of puzzle could easily filter out those that don't involve all the criteria.
15,
"To be specific, there's no point submitting small batches at all because they are not useful."
They are useful, you may just simply underestimate the power of team work, the idea behind is doing this incrementally by volunteers. After a member has completed his 10 puzzles, he can now take the next 10 puzzles if he is still willing. It is not advisable to give a member 100 puzzles or more at once.
Here is an idea, I have compiled below some tactical motifs, the plan is to give this to every member, for every puzzle assigned, the member will just select which of the tactical motifs are applicable for each given puzzle.
The motifs in csv format from my excel sheet. The format is Id number, Category, Tactics motifs. There are motifs that are not yet categorized, but this can be done later. This can be revised and improved of course.
Table 1.0 - Tactics Motif
IdNum,Category,TacticsMotifName
1,Pawn,Advanced pawn
2,To be Categorized,Avoiding perpetual
3,To be Categorized,Avoiding stalemate
4,Mate,Back rank mate
5,Attack,Battery
6,Defense,Blockade
7,Mobility,Clearance
8,To be Categorized,Decoy
9,To be Categorized,Deflection
10,To be Categorized,Desperado
11,Attack,Discovered attack
12,Check,Discovered check
13,Attack,Double attack
14,Check,Double check
15,Sacrifice,Exchance sacrifice
16,King attack,Exposing the king
17,Weak pieces,Hanging piece
18,To be Categorized,Interference
19,Mate,Mate in 1
20,Mate,Mate in 2
21,Mate,Mate in 3
22,Mate,Mate in 4
23,Mate,Mate in 5
24,Mate,Mate in 6 or more
25,Promotion,Minor promotion
26,To be Categorized,Overloading
27,Pawn,Passed pawn
28,Check,Perpetual check
29,Pin,Absolute pin
30,Pin,Relative pin
31,Pin,Attacking the pinned piece
32,Promotion,Queen promotion
33,Sacrifice,Queen sacrifice
34,To be Categorized,Removing the defender
35,Pin,Unpinning
36,Promotion,Rook promotion
37,Attack,Skewer
38,Mate,Smothered mate
39,Mate,Stalemate
40,Traps,Trapping a queen
41,Traps,Trapping a rook
42,Traps,Trapping a minor piece
43,Weak rank,Weak back rank
44,Attack,X-ray
45,To be Categorized,Zugzwang
Here is an example of Training number 1 in lichess, also in csv format.
Example 1.0 - Lichess Training puzzle #1
"Tactics training number", 1
"Puzzle link", "http://en.lichess.org/training/1"
"Game link", "http://en.lichess.org/hbl5osqk#90"
"FEN", "8/8/6Rp/1ppPk3/p3Pp2/2P1nP2/P6P/2K5 w - - 2 46"
"Tactics motif", "Mate in 1"
"Tactics Category", "Mate"
I am not an expert in web development (knew only some c/c++, python scripts, excel vba, sqlite). This project would be easier if selected members of Lichess are able to open the puzzles in web, cast their vote on tactical motifs (multiple select method). Then gather the data for processing.
If web is not possible then we can do this manually, assign team members specific puzzles to work out, then submit to a person in charge in data collection. Data format can be decided by the data collector to suit easy processing for lichess integration.
The team leader may create reviewers, to review submitted tagged puzzles. This is not a race, just allow a member to finished its assigned job.
"To be specific, there's no point submitting small batches at all because they are not useful."
They are useful, you may just simply underestimate the power of team work, the idea behind is doing this incrementally by volunteers. After a member has completed his 10 puzzles, he can now take the next 10 puzzles if he is still willing. It is not advisable to give a member 100 puzzles or more at once.
Here is an idea, I have compiled below some tactical motifs, the plan is to give this to every member, for every puzzle assigned, the member will just select which of the tactical motifs are applicable for each given puzzle.
The motifs in csv format from my excel sheet. The format is Id number, Category, Tactics motifs. There are motifs that are not yet categorized, but this can be done later. This can be revised and improved of course.
Table 1.0 - Tactics Motif
IdNum,Category,TacticsMotifName
1,Pawn,Advanced pawn
2,To be Categorized,Avoiding perpetual
3,To be Categorized,Avoiding stalemate
4,Mate,Back rank mate
5,Attack,Battery
6,Defense,Blockade
7,Mobility,Clearance
8,To be Categorized,Decoy
9,To be Categorized,Deflection
10,To be Categorized,Desperado
11,Attack,Discovered attack
12,Check,Discovered check
13,Attack,Double attack
14,Check,Double check
15,Sacrifice,Exchance sacrifice
16,King attack,Exposing the king
17,Weak pieces,Hanging piece
18,To be Categorized,Interference
19,Mate,Mate in 1
20,Mate,Mate in 2
21,Mate,Mate in 3
22,Mate,Mate in 4
23,Mate,Mate in 5
24,Mate,Mate in 6 or more
25,Promotion,Minor promotion
26,To be Categorized,Overloading
27,Pawn,Passed pawn
28,Check,Perpetual check
29,Pin,Absolute pin
30,Pin,Relative pin
31,Pin,Attacking the pinned piece
32,Promotion,Queen promotion
33,Sacrifice,Queen sacrifice
34,To be Categorized,Removing the defender
35,Pin,Unpinning
36,Promotion,Rook promotion
37,Attack,Skewer
38,Mate,Smothered mate
39,Mate,Stalemate
40,Traps,Trapping a queen
41,Traps,Trapping a rook
42,Traps,Trapping a minor piece
43,Weak rank,Weak back rank
44,Attack,X-ray
45,To be Categorized,Zugzwang
Here is an example of Training number 1 in lichess, also in csv format.
Example 1.0 - Lichess Training puzzle #1
"Tactics training number", 1
"Puzzle link", "http://en.lichess.org/training/1"
"Game link", "http://en.lichess.org/hbl5osqk#90"
"FEN", "8/8/6Rp/1ppPk3/p3Pp2/2P1nP2/P6P/2K5 w - - 2 46"
"Tactics motif", "Mate in 1"
"Tactics Category", "Mate"
I am not an expert in web development (knew only some c/c++, python scripts, excel vba, sqlite). This project would be easier if selected members of Lichess are able to open the puzzles in web, cast their vote on tactical motifs (multiple select method). Then gather the data for processing.
If web is not possible then we can do this manually, assign team members specific puzzles to work out, then submit to a person in charge in data collection. Data format can be decided by the data collector to suit easy processing for lichess integration.
The team leader may create reviewers, to review submitted tagged puzzles. This is not a race, just allow a member to finished its assigned job.
If thibault would agree I think we can start this job.
5,
I agree useful stats can be generated out of the results from puzzle with motif.
I agree useful stats can be generated out of the results from puzzle with motif.
I agree, small batches by a multitude of qualified members would indeed be just as useful as only a few members doing it...and it would get the job done a lot faster.
The only issue is that I continue to think that the tags must be from an established list so that there aren't people making up all their own tags. If possible, a checklist would be ideal, where basically someone just goes to their puzzle block, reviews each of their puzzles by checking boxes on a form, and the form generates the spreadsheet data accordingly. I think this could be done using Google Forms to make some sort of a survey.
As long as the choices for the checkboxes are thorough enough to satisfy what people would want to filter out in a puzzle review, but not so overly thorough as to include so many tags they become useless as filters. (e.g. If only 100 out of 60,000 puzzles had a certain theme the filter would be stupid even if you had completed 5000+ puzzles because you maybe have only 3 or 4 puzzles of that theme if any of them.)
I think the first step in this would be to create a team and compile a list of useful tags that could be applied. There's no sense waiting for a green light from thibault, basically if we compile a large chunk of data and present it to the devs, they'll find a way to use it I'm sure. I don't think anyone is going to say "no, we don't want to have the ability to do a puzzle review" considering it's a very useful feature, therefore since this is likely a feature that is going to be implemented at some point in the future, the more legwork that is already completed would simply speed up the process of that implementation, no?
The only issue is that I continue to think that the tags must be from an established list so that there aren't people making up all their own tags. If possible, a checklist would be ideal, where basically someone just goes to their puzzle block, reviews each of their puzzles by checking boxes on a form, and the form generates the spreadsheet data accordingly. I think this could be done using Google Forms to make some sort of a survey.
As long as the choices for the checkboxes are thorough enough to satisfy what people would want to filter out in a puzzle review, but not so overly thorough as to include so many tags they become useless as filters. (e.g. If only 100 out of 60,000 puzzles had a certain theme the filter would be stupid even if you had completed 5000+ puzzles because you maybe have only 3 or 4 puzzles of that theme if any of them.)
I think the first step in this would be to create a team and compile a list of useful tags that could be applied. There's no sense waiting for a green light from thibault, basically if we compile a large chunk of data and present it to the devs, they'll find a way to use it I'm sure. I don't think anyone is going to say "no, we don't want to have the ability to do a puzzle review" considering it's a very useful feature, therefore since this is likely a feature that is going to be implemented at some point in the future, the more legwork that is already completed would simply speed up the process of that implementation, no?
This topic has been archived and can no longer be replied to.