Saturday, February 13, 2016

Piece aura vision exercise

When you filter the M1-h position so that you keep only the positions with white pieces, you get the kind of positions as in the diagram. I think these are perfectly suited for the training of piece aura vision.

White to mate in one
I'm further breaking down the M1-h dataset in subsets like
  • White has at least one pinned piece which for that reason cannot deliver mate.
  • White has at least one piece that can't deliver mate because the checking square is protected
  • White can at least give one check that can be blocked by a black piece
  • White has at least one piece that creates an escape square for the black king when giving check
The properties of these subsets are mutually exclusive, so that you can train the distinctive subtasks in isolation.


  1. The idea what I try to accomplish is fairly simple. I try to divide the task M1-h in as many subtasks as possible, and train each subtask until I reach, say, 18 subtasks per minute. Then we will see what that does for the main task. Today I scored 18 attackers per minute and 18 defenders per minute, so I don't have to worry about those subtasks anymore. Currently I'm training M1-h subset no black pieces. It seems very well improvable. So far, I reached 7 M1-h-nbp per minute.

    1. The idea is great but there is a problem that your trainingset might not be big enough. My experiments with easy mate problems years ago did show that i memorise easy mates at chesstempo something like 8 days ( solving speed is bigger if i saw them less than 8 days before ). If you create a trainingset like say 4000 problems then you will see "many" puzzles "too soon" again, you will start to memorise them instead improving your skills/vision.
      I did extract the m1-h positions here, but this file is now much bigger

      My script has a routine to mirror positions this might help to prevent memorisation too? This way you "double" the number of exercises.

  2. Off-topic: 'find-all-escape-squares' beta =>

    I don't know if this is the right place to talk about this, but I did it to publish the tool.

    I need opinions about my filtering criteria to continue extracting positions, because I don't find some of them very "instructive"/helpful and I don't know which kind of positions would be better. Opinions about the characteristics of the tool are welcome too.

    Also, I'm looking for possible bugs and unexpected behavior before adding more positions to the database. If anyone finds something would be great.

    So feel free to share your opinions. Thanks.

    1. The positions seem a bit too simple. That might lead to RSI ;) Positions with the king in the middle of the board are probably harder.

      I found the following problems:
      position 184 looks mate to me. The black king cannot live on e8.
      position 168 the king can live where it stands, since white can take the black check giving piece with another piece than his king.
      position 195 seems to be messed up. There are two black kings, and they can live on the squares mo, ve, s.

    2. I really like these puzzles. I hit the same problem as Tempo mentioned above (position 195 - illegal - two Black Kings).

      In general - a very nice idea! :)

    3. @Lain, how did you manage to put your code on weebly?

    4. I was trying to overwrite all my weebly site content with arbitrary data, when I found the perfect way to do that.

      If you go from the page editor > Theme > Edit HTML / CSS, there is a file manager that can create folders and upload multiple files in a single shot.

  3. i found problem nr 4 with a bug: c1 is "unreachable"
    i dont think the problems are too easy, im far away from any "flow"

  4. The castling flags aren't working properly? I can't replicate the bug.

    The FEN position is "1r1q2r1/pk3p1p/b1p5/3pP1b1/1P3B2/P6Q/2PN1PPP/R3K2R w KQ - 10 20 e1d1c1b7a8c7b6" and according to it, white can castle king/queen-side.

    When I remove all the problems and keep only this one, everything seems correct in the interface.

    Btw, what do you think about handling "check-removing" moves like interferences and captures?

    Tempo said that the king can live in his current square if it's possible to stop the check with another piece. I will fix it, but I'm considering a stupid case where the check-remover piece can be captured next, resulting in check again, and then I don't know how to proceed.

    I guess I would only consider statical one-move things, but I like to be sure what I'm doing. (?)

    1. One move things is deep enough for these kind of exercises. There is no need to complicate matters.

    2. oops my fault, no #4 is ok, i did forget 000. Im so used that "my" prog is without casteling... ;)

  5. I agree with Aox. The problems SEEMS too trivial (easy). However as we have already noticed (many times) even the easy tasks may contain some difficulty! It is a shame to me, but I had problems with pointing right answers (squares) even if I knew "it is too easy to go wrong!".

    This tool is a MUST if anyone thinks seriously about improvement at M1 (easy and hard). If it is "too easy" one can solve it with a perfect speed, isn't it?

  6. I managed to upload the exercise M1-h BLACK KING ONLY. Hattip to Lain!!
    You can find it in my sidebar.

    I'm curious what you guys think about it.

    1. I tested it for a while (just 50 positions) and it looks EXCELLENT! I really like it as I do not have to look for checks/mates for both Kings and if I see correct - the positions are simple ones - I mean "not (too) crowded".

  7. I added 6 new links to exercises + a link to Find All Escape Squares from Lain.
    FAC+M1e+M1h positions with only a black king or K+P (the latter to get more positions).

  8. The tool should help too.. its one simplified "subtask". Im not shure if #4000 is enough

    It would be nice if the help-tools would not use information from the original m1 problems 1 to 1 because that is, to some extend, like a repetition of the problems. I suggest to mirror these ( Queen to Kingside a-h, b-g...). Then a "recognition" is not to expect. There is some code in the script for that.

    We have now several tools to crack m1. I hope it is enough.
    I already have some more ideas for more exercises for general tactics.

    Now we need some Guinea pig's

    A Guinea pig should do
    1) 2 weeks of m1-e ( and m2-e ) maybe? Maybe 300 ( each ) per Day?
    2) Report score
    3) a) improve at some board vision exercises like Attackers,Defenders,FAC..
    3) b) if an improvment at one board vision exercise is not possible, continue with an more easy one
    4) several days m1-e
    5) report exercise and new score
    6) goto 3)

    Anyone there?

    1. If I would have a detailed list of tasks - I think I could try it on. Maybe it could be a good example to follow.

      BTW. I can do it, but I need to start this task as soon as possible (because after 2 weeks from now - I will be very busy again). Let me know what and how should be done my friend.

    2. Tomasz you are "not fresh", still it would be interesting if you can improve by using any of the new tools.

  9. Hmmm...very interesting. The M1-h are very similar to the positions from Polgar's brick which I love!

    One idea for a sub-task is King escape squares. You could take all the M1 positions and the challenge is to quickly select all the safe squares around the king (plus the one he is standing on). That is how I solve the Polgar's positions -- I see the shape the safe squares make and then find the piece that makes that shape. Make sense?

    I've actually made worksheets for my son like this and would love to have a software version! The Stepping Stones workbook has these too.

    1. As soon as I hit publish I saw the Find All Escape Squares exercise! That seems like a very critical sub-task for M1-h

  10. Yes, but it's a bit far from being finished. I need to find (more) bugs, make positions harder/simpler and extract a lot of additional positions.

    1. @Dan
      we try to measure with the m1 test how quick we are thanks to aquired skills.
      If we take 1-1 positions from any m1 exercise/test then we "learn" the position ( some memorisation ) and that will have an unwanted influence on the results. As less often we see these positions as more reliable the values of m1 are.

      there might be an easy way to find good FAES-h positions :
      I did use these compositions to generate the m1-h puzzles but i think you might use the start positions of these compositions, they should be "hot" enough and are defenitly "hard" because of the number of existing contacts
      And these positions are at least not "identical" to the m1-h positions

    2. That's amazing. Thank you!


      Maybe it have some dumb bug, I can't test properly at this moment. Also there are a lot of positions where the king has only his home square available. Should I remove these positions?

      (Btw, I didn't extracted all positions, the database is too big enough for me).

  11. I'm not too worried about memorisation. I don't know for sure if is actually a disadvantage or a benefit, or that we even can do without. Anyhow, that's why I added the K+P positions, so that the training base contains over 13.000 positions M1h. The extra pawns don't complicate the position too much.

    As far as I'm concerned, the mates with black King or K+P only are a challenge for my piece aura vision, and that is what I'm after.

    I identified four other subtasks , for which I want to write a selection program.

  12. When i read that there is only 1 variation i know what to do :/
    So either:
    erase all puzzles with 1 variation.. then the number of puzzles will shrink dramatically but there are still some compositions left to use


    Dont display the number of variation,,, but then the user might click the king first and solve the puzzle by "accident". Then we need an extra rule that the king has to be selected last, then there has to be some type of punishment ... sounds to complicated


    The user has to find all escape squares for both kings, but the white king is usually without any limitations


    Leets see, 300 puzzles a day? 1 week to get warm, 3 weeks to improve = 9000 puzzles time 2 for security ~~ = 20 000 puzzles needed. Of course as more as better...

    I think erasing the puzzles with 1 variation could do. The number of puzzles could be artificially increased by erasing one white piece. I cant judge the effect though. Or maybe a potential black piece at the side of the king could be removed, to "create" an escape square? That would be "easy" to program too? The good thing of this method could be that at least sometimes the king will be in check and his own square is NOT solution...

    1. I decided to remove all 1-variation and I have a total of ~26000 problems with 2 -min- and 3 -max- variations. I can extract more soon.

      I don't know if keep the "variation" box in the interface. I removed it.

      About mutating positions erasing pieces, I will try to do that, but for the moment I want to try simpler solutions.

  13. there might be an other possibility to solve the "1 variation problem" : No display of the number of variations and the user has to select the number of variations instead of selecting the right squares..?????

  14. @Tempo

    Escape-squares Training
    Select all squares where black OR white king can live

    I am not sure, but there is an error at these problems. If the King is in check CAN it stay at this square? I tried to solve some puzzles and it looks like - your tool allows (forces) this solution - clicking the square the King is standing (as correct) even though it is in check!? Please let me know if you noticed that or if I do something wrong :(

    1. The escape square training is made by Lain. I provide just the link. If the king is in check it can stay there whenever a piece of you can capture the chess giving piece. But maybe that should not be part of an escape training?

    2. It's exactly as Tempo says. If the king is in check, and any piece can interpose or capture the attacker, then the king square is considered as safe without moving from it.

      I find it confusing, but makes sense.

      Also seems that it most of the positions the king can stay in his initial square because of this rule, so that in almost all problems we will have to click the king square.

      The rules of the exercises are open to define. If something is wrong, I will change it.

    3. Lain said

      "It's exactly as Tempo says. If the king is in check, and any piece can interpose or capture the attacker, then the king square is considered as safe without moving from it.

      I find it confusing, but makes sense."

      I find it confusing

      we have the m1 problem, m1 is for a sub 1900 not improvable maybe even vor a sub 2100..
      We know we have to simplify m1, either by simplifying the position or by selecting relevant subtasks

      The subtasks are
      a)Find all check
      b)Find all legal moves with king in check
      b1) find all takes of king attacker
      b2) find all intersects between attacker of king and king
      b3) find all all legal moves of king itself

      FAES is ( in my eyes ) b3 ( with a nullmove of the king if the king can stay )
      To mixt b3 with b2 gives me no good feeling..

      b1) is already the attack exercise at chessgym
      b2) is still to write

      whats about google sites.. anyone with a prepaid handy?
      But we could share storage place with googledrive or dropbox..
      Then a "program" could be just a zip'ed folder?
      Or ?

    4. I agree we should not confuse matters. Even if I am the initiator of the confusion.

    5. So I keep the null move as a solution when the king isn't in check, and remove the "b2" thing (intersect-capture)? or only look for all legal king moves?

  15. if there are puzzle where the king is in check AND puzzles where the king is not in check then the null move selection should be available.
    Possible intersection.. thats an other thing, thats something for a different exercise

  16. There are a lot of approaches to the SAME problem. I want to share my vision.

    Check is the situation when the King is attacked by another piece (all except King).

    How to defend AGAINST check. There are only 3 ways to do it:

    1) CAPTURE the piece. The King can capture the attacking piece ONLY if it is not defended (protected).
    2) ESCAPE the King to unoccupied square. You can go to any other square (one square in any direction) only if: a) it is not taken by your own piece (you cannot capture your own pieces) or b) it is not attacked by the opponent piece (or pieces).
    3) INTERFERE at one the unocuppied square at the line of attack. You cannot interfere at two situations: a) against direct check (when the piece attacks the King standing next to it) b) against the attack of Knight of pawn (due to their specific movement)

    Now go to the TOOLS - excercises to practice the skills.

    #1) CAPTURE the piece which is attacking the KING. CLICK all the pieces which can capture the checking piece.
    #2) ESCAPE the King. CLICK to all the squares you can go (run) with your King. Take notice they (the squares) have to be unoccupied and not-attacked squares.
    #3) INTERFERE against the check. Two variations:
    a) CLICK all the pieces you can move to interfere against check
    b) CLICK all the squares you can move any of your pieces to interfere against check

    NOTICE: at #1 and #3 practicing tools you have to exclude double check positions (as you cannot capture both pieces in one move nor interefere against double check)

    Let me know if this short guide is helpful and just ask or comment if you wish.

    1. There are two other subtasks related to M1:
      4. Identify all pieces that would be able to give check if they weren't pinned to the white king.
      5. Identify all pieces that give check but create a new escape square for the black king in doing so.

  17. #1 already exist thats the ATTACK training, its not that much of a difference if the attacked piece is a king or a pawn
    #2 is the FAES
    #3 is open.. but possibly not needed if we can make m1 improvable by the other exercises???
    #4 that is a FAC with all pseudolegal moves. The FAC of Fritz is only with legal moves and if i remember correct my FAC is with pseudolegal moves so this is finished too, by the way i dont think that the difference is that important if we just want to make m1 improvable
    #5 thats a graet idea but that is a complex task, not easy to improve for a human so i would put it at the end of the progamers todo list. In a way it is a real aura task a "consequences " task

    If we can make m1 improvable then i have several further ideas how to create a good aura vision.
    If we cant do it with the already existing tools then.. Minigolf? ;)

    As soon FAES is available i will try to crack m1 again
    1)find baselines of m1-e m1-h and m2-e
    2)attackers defenders FAC
    3) = 1) ... estimatingly no effect because i did already lots of them
    5) = 1)
    6) Tools of Tempo
    5) = 1)

    I hope i can survive that salt :/

  18. oops there is a bug somewhere
    While the speed is ok at FAC-KN its not ok at FAC-QK i will have to look at it

  19. I suspect the speed isn't ok at FAC-KN either. They are derived from the same index.html as M1-h. But when I derived them from the same index.html as FAC, I had the same problems. I think that the variations aren't count for the average, just the positions. When the amount of variations is low,as with the knights, you don't notice it is off. But with FAC-KQ, there are always many variations.

  20. youp maybe its puzzles per min..but a simple bug is possible too.. i check that until monday

  21. Hi Aox,

    I've just finished copying my stuff from Munich's Diary. I've asked Richard to honor Munich's request.

    Kind regards,

    Jaap Amesz

  22. @Tempo and Lain
    replace in all chess.js the line with:

    by this line:

  23. I did a 5 minute test with FAC white queen only, using a stopwatch. I scored 139 checks in 5 minutes => 27.8 CPM avg.
    But the program scored a measly 14.06 CPM. So the score is still a factor 2 off.

  24. replace in the html-file

    var startzeit;
    var startzeit=0;

    startzeit =;
    if (startzeit==0) startzeit =;

    Replace in the file Chess.js



    score = correct clicks per minute averaged from the beginning
    count = Summ of all correct squares since start
    Startzeit= time when user hit the start button in Milliseconds time at THIS moment in MilliSeconds
    60000= 1 minute in milliseconds

  25. I changed the code and it looks good. I'm much faster now without any effort ;)

    1. I am not sure if it is my computer which slowed down or new system that counts the values. I tried to see how much it differs from previous one... and I scored just 30 MPM at M1e!

      Tempo - you said: "I'm much faster now without any effort ;)". It looks like I am MUCH slower without any effort - now the balance is equal LOL

    2. So far, I only trained the diverse subtasks from FAC, except for the No Queens subtask.
      I raised the bar from 15 to 35 CPM for all these subtasks, and after a few long sessions I managed to be so fast.

      The funny thing is that the No Queens subtask is now the slowest. So the queen is no longer the most problematic piece any more. I really like the Queen Only exercise, since it feels like a very critical skill. I worked hard to improve it.

      I expect to be only a little faster at M1-e, but I haven't tried my hand yet.

    3. The old rating system was weighting the fast clicks more than the slow clicks, the new rating system is weighting the slow clicks more than the fast clicks.
      But as always:
      we may turn and twist as much as we want, our a.. is always in our back ;)
      Meaning : we need to get faster, no matter if we measure the speed by mph or km/h

    4. Yet I'm glad we no longer measure in furlongs per fortnight :D

  26. @Tomasz et al: the score is now simply the average of your whole session, for each and every exercise. This means that it has become easier to compare between different exercises. Oh, and going to the bathroom during a session is punished now.

  27. Now I can improve at these excercises without any effort! ;) :). GREAT NEWS! LOL

    And now serious. Thank you very much for making changes. Now we can see VERY real values and compare them between ourselves and the expected values (level).

    I rarely go to the bathroom, but when I make a longer break - I will notice the difference or start the puzzle once again ;) :).

  28. 4B1b1/2b5/1R6/2NPQ2p/1Pk1P3/1nP2N1n/r2P3K/8 w - - 0 1 belongs to the "pinned" subset...?

    The d2-d3 move would be checkmate, but the pawn is pinned. Then this position may belong to "pinned" subset.

    There aren't **checkmates that can be stopped**, neither by blocking it or capturing at the checking square. (These subsets are considered as different in my script, btw). This position can't belong to "check-block" nor "check-capture" subsets.

    Also, there aren't check moves that **change the quantity of black king lufts**, creating one or more new escape squares. So this position can't belong to "check-escape" subset.

    The position doesn't have more than one subset characteristics. The position belongs only to the "pinned" subset. (Problem isolation).

    Are the conditions right?

    And here is the (main) reason of my first question:

    Qd4 would be mate, only if the queen were not pinned AND if the d4 square were not protected by the black knight. Which subsets match with that condition? "pinned"? "protected"/"check-capture"?

    1. The position belongs to the pinned subset, definitely.

      I assume that the clicks must be on the pinned pieces. My only concern is that the subtask might be too trivial. But I thought at first that the FAES exercises would be too trivial too. Until you came with the FAES-h positions.

      I think we need all variations as different exercises. For "pin training" we need positions with "pin only", and we need positions with "pin" and "check-capture". We need positions with "pin" and "check-block". And we need positions with all three.

      The fact that we click to identify the pinned pieces, guarantees the isolation of the training. By adding diverse distractions, we can vary the complexity of the position. That is even better than varying complexity by leaving pieces of a certain type out of the position.

      Leaving certain pieces out can always be done later, if necessary. That takes little effort.

      I hope I answered your question.

    2. Thanks for your answer. I have another doubts.

      Which squares must be clicked in the other exercises?

      For example:

      In "check-capture", the capture squares? protector pieces?
      In "check-block", the blocking squares? blocking pieces?
      In "escape", the new lufts created? the piece that gives check? etc.

    3. There are 3 approaches to this:
      The attacker, the attacking square or the defender.

      An attacker can have other possibilities and one of them might be a valid attack. So the attackers drop off.

      I'm inclined to say: click the defenders. Since it is there where to start the next task when playing: how to remove the defender. To be consistent, it might be better to click the pinning piece in stead of the pinned piece.

      If it turns out that clicking the defenders are unimprovable subtasks, we might go for the lighter option clicking the attacking squares, but I hope we don't need that.

    4. In the case of creating new escape squares, clicking the attacker that causes them is the most logical choice for me.

    5. To identify "new" escape squares, I should calculate the difference between free squares before and after the check is done? (if it is at least 1) or I should compare the squares looking if they are at least the same amount but they are different?

    6. What we need for M1-h is:
      Before the check there are zero escape squares, after the check there is/are one or more escape square(s).

  29. IMO pins are not of main/first interest. Pins are an important element of m1-h and not of m1-e and i think to crack m1-h we need to crack m1-e first ( as beeing the easier task). m1-e is without a lot of pins.

    I suggest something as "Find all ((pseudo)legal) checks with queens" first

    1. I don't consider M1-e to be a logical subtask from M1h. It is just a subset of all mate in one positions, based on their occurrence in real life games. Tomasz has already improved M1-e very close to the max. I'm not inclined to develop exercises specific for M1-e. Even though I don't like the method used by Tomasz (working his *ss off). I use M1-h to identify the impediments for improvement, and occasionally that will lead to exercises that are beneficial for M1-e too.

    2. Tomasz was able to improve in m1-e because m1-e was "easy" for him because of his superior strength. For his powers, m1-h is now the/a right goal. For me its m1-e and its subtasks, thats already hard enough for me. Its the "easiest" real tactic which is to hard for me to improve.

  30. I have a question. What do you think about writing down the results (values) for some period (f.e. in one month time)? This way we could compare the data (values) between us - or try to see which subtasks (tools) gives the best (or worst) results?

    @Tempo - let me know which subtasks/tool are you currently using (especially list the tools which are available at your blog as links).

    1. great idea
      I just started m1-e wirth something around 15, try to find my baseline in the next 7 days. I think there should be a done>300 to read the value. After these 7 Days i think i will try FAES first, but continue with m1-e to see any effect right away..

  31. I chose the subtasks "pinned attacker", "protected attacking square", "blocked line of attack" not only because of they are critical for m1-h, but also because it are quite common subtasks of every tactical combination.

    1. I have no doubts that they are usefull for tactics and m1-h. I just think m1-e is the critical goal.
      I try to find a method to improve deciciv in chess.. so i look for a method to improve deciciv in tactics so i look for a method to improve m1-e.
      I detect that critical subtasks of m1-e are FAC and FAES and if i see that i get problems even there, i try to solve these by further disection.
      To me, pins are at the present moment not interesting because if i cant crack m1-e.. the whole method is junk anyway, pin exercises will be without "real" use too.

      I try now to put all of my very limitied salt tollerance at one single "weak" point: m1-e.

      IF we can crack m1-e THEN pin exercises will help us to crack tactics, no doubt

      I just started m1-e again after a loong time without any salt. If i will see that subtask abc is a problem for me i will search for methods to make abc improvable ( by disection of subtasks / uniformisation= allways the same / simplification = less "elements" )

    2. We need these different angles of attack of the problem. You focus on m1-e, I focus on exercises that go beyond m1-h. If I can improve my "pin-vision" a factor 2, I'm happy. But if I am not able to improve my pin-vision, I will dissect it and might well end up finding the same kind of subtasks you find while dissecting m1-e for subtasks. In that case, we have simply more proof that we are on the right track.

      Since we know little to nothing, it is impossible to predict the right method beforehand.

  32. MATE in 1 EASY

    DAY 1: 1600 puzzles - Mates / min: 31.55 (average score between 30-32, no better scores)

    It looks like the new system of counting is more harsh ;) :). Or maybe it shows some breakthrough of mine - insted of shooting (clicking as fast as possible) I analyse the position. I try to do it really fast, but sometimes I simply "freeze". The analysis of the position shows the answer (which squares have to be attacked), but I cannot see the correct move (drawback at FAC is clearly visible).

    I disagree with your statement Tempo ("Since we know little to nothing..."). I think we know quite much, but we have not found the most important things yet! Or to put it in simpler words - we have trouble with discovering relationship and the strenght of influence to each other subtasks :).

    Let me know what order and which ones of the subtasks do you recommend. I can do a few thousand of any of the recommeneded ones in a short period (to the end of the week) and we can see which ones have been the harderest to score high. Just the idea - if you think different, please share your views!

  33. @Tempo:

    I uploaded a sample FEN list, with the type of problems and the solutions. But I don't tested bugs rigorously and my script doesn't handle promotions moves yet (I will add that soon).

    1. Wow! Great effort!!

      I checked the following:
      8/8/5R2/4k3/2P5/2K5/1B3P2/8 w - - 0 1 #escape f6 OK

      8/2p3Q1/p1RNKP2/n2p1p2/1P1k1p1R/1p3Prb/3PPP1b/2N5 w - - 0 1 #protected d5a5h3a6g3 WHY g3?

      2b4Q/4p1K1/4pr2/2P1k3/2Pn1R2/3P3p/7B/8 w - - 0 1 #blocked f6d4 #escape f4 OK

      b2R1B2/3p3r/nN6/1R6/rn1kp3/P3pB2/1PQ5/K3N3 w - - 0 1 #protected h7a6e3e4b4a8 #blocked a8b4 OK

      6Q1/2p1kr2/3R4/8/1R2K3/B3p2B/8/8 w - - 0 1 #escape g8 MISSING d6?

      B4n2/5n2/1p4N1/1p1K1N2/2pP2p1/5k2/1R3b1Q/8 w - - 0 1 #protected f2f7g4 #blocked g4f2 OK

      1b1R1N2/6p1/1Np2pP1/2p1k3/1pp1BnPK/6n1/q2QRrPB/8 w - - 0 1 #protected c6f4f2g3c5b8a2b4 #blocked f4g3c4 OK

      2R3N1/3prn2/NP2qQ2/3k1B2/R4PP1/BnP2P2/1K2p3/1b6 w - - 0 1 #protected e6d7e7b1f7b3 #blocked b3 OK

      3Q4/b3pr2/Kp3N2/1P2k1P1/1p1q2B1/4B3/1N4p1/4R3 w - - 0 1 #protected d4a7f7 #blocked d4 #escape f6 OK

      6Q1/1K1p1R2/3P3p/3k1N1p/3P3P/3p1p2/3B4/7B w - - 0 1 #escape f5 OK

      3b2K1/4p3/4N1k1/5rP1/4r2P/3B2N1/8/8 w - - 0 1 #protected f5e4 OK

      8/1pNN4/2k3p1/5pK1/1P2pB2/7B/8/3R4 w - - 0 1 #escape d7 OK

      8/4p3/1pP5/1K1kb1p1/R5P1/3N4/P3NP2/3Q4 w - - 0 1 #blocked e5 OK

      So you are very close!

  34. Thanks for review the positions.

    8/2p3Q1/p1RNKP2/n2p1p2/1P1k1p1R/1p3Prb/3PPP1b/2N5 w - - 0 1 #protected d5a5h3a6g3 WHY g3?

    Because of the f6-f7 discovered check. Rg3 controls the checking square.

    6Q1/2p1kr2/3R4/8/1R2K3/B3p2B/8/8 w - - 0 1 #escape g8 MISSING d6?

    Previous to any check, the king has the e7 and d6 squares.

    Rd6 has two checks: Rd7 and Re6.

    After the d6-d7 check, the king only can live in f6.
    After the d6-e6 check, the king only can live in d7.

    In both cases the king is more restricted than before, has less number of escape squares.

    1. Ok, then technically all positions look correct. Yet the escape exercise will be too difficult.

      We might invigorate the condition:
      Before the check, the king should have only 1 square ( the square it is standing on), and after the check it must still have 1 square (or more), but not the one he is standing on (since he is in check).

  35. So now "6Q1/2p1kr2/3R4/8/1R2K3/B3p2B/8/8 w - - 0 1 #escape g8" doesn't belong to "escape"?

    And "3Q4/b3pr2/Kp3N2/1P2k1P1/1p1q2B1/4B3/1N4p1/4R3 w - - 0 1 #protected d4a7f7 #blocked d4 #escape f6" yes?