I wonder if there is some epd file of problems from tablebases

[Uri Blass]

All the correct mate in 2 problems(correct problem is a problem when white has a single solution in every move)
All the correct mate in n problems by the same definition.
What is the maximal number of n?

All the correct studies of white to start and win with n moves.
White has only a single move to win the game in the main line in every move of the study
The study does not have to be finished in mate and a position when white has more than one winning move may be the end of the study.
Whyat is the maximal number of n for different tablebases?

All the correct studies of white to start and draw with n moves when white has always a single non losing move to get the draw in the main line.
What is the maximal number of n for different tablebases?

[phhnguyen]

I believe there is not any such EPD file about EGTBs. The reason is that there are too many positions to keep in text formats (or any explicit position representation). For example, a single endgame file of 5-man KBNvKP may have over 32 million positions.

Your question looks like a request to create a new feature or program. Definitely, it is doable. The program needs to scan a given endgame file to find the data with the correct values of DTM/DTZ, create the chessboards for them and verify other conditions then write out if everything is matched.

I have tried once. However, IMO, the problem is that it is quite... boring because there are too many. For a good situation, you may have "only" a few hundred positions as a result. But typically you may be flushed by some millions of quite-similar positions. Say, for a KBNvKP, many positions differ a bit on locations of B or N. In many endgames/many positions, the results may be so biased to one side. It means the winning side has so many ways to win. There is almost no drama. Thus you have very few chances to filter out/reduce positions. Also, you may work and filter for some specific endgames but not all in general. (If someone could apply machine learning to filter positions, there is a better chance than by coding).

I used to create a good GUI/tools thus I can review quickly a lot of positions, and group them by some catalogs. It may be fine for some developers/researchers but not good for average users because of the complicatedness and huge numbers.

If you are not serious and need only a few results, you may solve it manually: consider some endgames, find some positions, and use GUIs/engines to verify solutions.

For example, you may use Banksia GUI to edit an endgame then it will auto-ask your local Syzygy or online chessdb/lichess EGTBs about moves/solutions for your new position:

[jp]

I used to create a good GUI/tools thus I can review quickly a lot of positions, and group them by some catalogs. ...

For example, you may use Banksia GUI to edit an endgame then it will auto-ask your local Syzygy or online chessdb/lichess EGTBs about moves/solutions for your new position...

Can Banksia or another GUI do analogous things to what they can do with the usual databases of games? e.g. give the win/draw/loss statistics for specified types of positions (e.g. specified material, with or without extra constraints) and list all the wins. Or list all positions of specified type with DTM or DTZ greater than some value.

That'd be interesting info, especially since I've always wondered when win/draw/loss stats are cited for a type of position whether they include (and therefore are skewed by) positions where an unprotected piece is attacked. (With the above example features, one could exclude such positions by asking the GUI for stats with DTZ > 1 or 2, for example.)

[Arpad Rusz]

I might do that for studies.
There are 6719 win "studies" in the K+P vs K endgame.
The longest have depths of 6 unique moves:

[pgn][Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "6k1/8/8/8/8/8/1P6/7K w - - 0 1"]

1. Kh1g2 Kg8f7 2. Kg2f3 Kf7e6 3. Kf3e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/8/6k1/8/8/1P4K1/8 w - - 0 1"]

1. Kg2g3 Kg5f6 2. Kg3f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/8/6k1/8/8/1P5K/8 w - - 0 1"]

1. Kh2g3 Kg5f6 2. Kg3f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/6k1/8/8/8/8/1P5K/8 w - - 0 1"]

1. Kh2g3 Kg7f6 2. Kg3f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/8/6k1/8/7K/1P6/8 w - - 0 1"]

1. Kh3g3 Kg5f6 2. Kg3f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/6k1/8/8/7K/1P6/8 w - - 0 1"]

1. Kh3g4 Kg6f6 2. Kg4f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/6k1/8/7K/8/1P6/8 w - - 0 1"]

1. Kh4g4 Kg6f6 2. Kg4f4 Kf6e6 3. Kf4e4 Ke6d6 4. Ke4d4 Kd6c6 5. Kd4c4 Kc6b6 6. Kc4b4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/8/7k/8/8/2P4K/8 w - - 0 1"]

1. Kh2h3 Kh5g6 2. Kh3g4 Kg6f6 3. Kg4f4 Kf6e6 4. Kf4e4 Ke6d6 5. Ke4d4 Kd6c6 6. Kd4c4
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/6k1/8/8/1P4K1/8/8 w - - 0 1"]

1. Kg3g4 Kg6f7 2. Kg4f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/6k1/8/8/1P5K/8/8 w - - 0 1"]

1. Kh3g4 Kg6f7 2. Kg4f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "6k1/8/8/8/8/1P5K/8/8 w - - 0 1"]

1. Kh3g4 Kg8f7 2. Kg4f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/6k1/8/7K/1P6/8/8 w - - 0 1"]

1. Kh4g4 Kg6f7 2. Kg4f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/6k1/8/8/7K/1P6/8/8 w - - 0 1"]

1. Kh4g5 Kg7f7 2. Kg5f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/6k1/8/7K/8/1P6/8/8 w - - 0 1"]

1. Kh5g5 Kg7f7 2. Kg5f5 Kf7e7 3. Kf5e5 Ke7d7 4. Ke5d5 Kd7c7 5. Kd5c5 Kc7b7 6. Kc5b5
1-0

[Event "PKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/8/7k/8/8/2P4K/8/8 w - - 0 1"]

1. Kh3h4 Kh6g7 2. Kh4g5 Kg7f7 3. Kg5f5 Kf7e7 4. Kf5e5 Ke7d7 5. Ke5d5 Kd7c7 6. Kd5c5
1-0[/pgn]

Depth 1 study:
[fen]8/8/8/8/8/K7/P7/k7 w - - 0 1[/fen]

Depth 2 study:
[fen]8/8/8/8/K7/2k5/P7/8 w - - 0 1[/fen]

Depth 3 study:
[fen]8/8/8/8/K7/8/Pk6/8 w - - 0 1[/fen]

[Arpad Rusz]

Depth 4 study:
[fen]8/8/8/8/8/8/P1k5/K7 w - - 0 1[/fen]

Depth 5 study:
[fen]8/8/8/8/8/3k4/PK6/8 w - - 0 1[/fen]

Depth 6 study:
[fen]6k1/8/8/8/8/8/1P6/7K w - - 0 1[/fen]

[Arpad Rusz]

For draw studies, I have excluded position repetition.
The longest has 16 moves:

[pgn][Event "pKk"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "8/8/8/2p4k/8/8/8/7K w - - 0 1"]

1. Kh1g1 Kh5g5 2. Kg1f1 Kg5f5 3. Kf1e1 Kf5e5 4. Ke1d1 Ke5d5 5. Kd1c1 Kd5e4 6. Kc1c2 Ke4d4 7. Kc2d2 Kd4c4 8. Kd2c2 Kc4b4 9. Kc2b2 c5c4 10. Kb2c2 c4c3 11. Kc2c1 Kb4c4 12. Kc1c2 Kc4d4 13. Kc2c1 Kd4d3 14. Kc1d1 c3c2 15. Kd1c1 Kd3e2 16. Kc1c2
1/2-1/2[/pgn]