Thanks!

As elegant as the convex hull idea may sound, I think you played the more interesting round, at only the cost of a difficult-to-word guess. Your rule produces more variety in the white set. For the convex hull idea, all inner letters (= all except the first and the last) would have to be four-corner-touching letters. This will boil down to identifying 3 classes of letters: (1) the blocky letters that touch all four corners, which are the letters allowed everywhere in the koan, (2) additionally allowed first letters, (3) additionally allowed last letters.

I'll write a longer post about ambiguities, but that's for a different thread. Both of the recent rules (English nouns, and corner agreement) are interesting to study in light of ambiguities. I'm also surprised how quickly both got solved.

In the meantime ...

---

Next round

White koan: WRONG
Black koan: INCORRECT

-- Simon

Quote from: Simon on November 20, 2024, 12:11:46 AMThe rule is that neighboring letters "agree" on the corners:

AKHTBN iff, for every possible pair of neighboring letters (one letter immediately following another) inside the koan, both of the following conditions hold.

• The left neighbor touches the top right corner of its own bounding box iff the right neighbor touches the top left corner of its own bounding box.
  
• The left neighbor touches the bottom right corner of its own bounding box iff the right neighbor touches the bottom left corner of its own bounding box.
  

This is exactly correct.
I wasn't very sure on picking this rule because I did not have a good way to formulate the rule in my mind at that point. Furthermore one could argue that whether a letter touches a corner or not is not exactly standardized and dependent on the font used. Decided to roll with it anyway, because as we know, if there is ambiguity about details, the master is always right (and I chose for Q not to be touching the bottom right corner).
After seeing Simon's first rule guess I kinda wish I had chosen that one because it feels very elegant.

zendofmt 1.1.0 is out with support for footer text.

How it works: Optionally, create a third file footer.txt with free-form text that will appear below the two koan lists. Backticks ` in footer.txt will appear as [tt] and [/tt] alternately. Linebreaks will appear as-is.

Love how there are 3 tools now in active development alongside the running games.

-- Simon

Consider the rectangular bounding box of each uppercase letter. The bounding box's bottom side is flush with the horizontal floor line on which all letters sit. We'll look at which of the four corners of the bounding box are reached by the letter's shape.

Examples: The letters H, X, M all touch all four corners of their individual bounding boxes. The letter A touches only the two lower corners of its bounding box, not the two upper ones. The letter R touches three corners (all except top right). The letter J touches only the top right corner. Some particularly round letters S, C, O touch no corners at all.

Q is a fickle case. I interpret it as touching no corners. Maybe it touches the bottom right corner in some fonts; it doesn't in mine.

---

The rule is that neighboring letters "agree" on the corners:

AKHTBN iff, for every possible pair of neighboring letters (one letter immediately following another) inside the koan, both of the following conditions hold.

• The left neighbor touches the top right corner of its own bounding box iff the right neighbor touches the top left corner of its own bounding box.
  
• The left neighbor touches the bottom right corner of its own bounding box iff the right neighbor touches the bottom left corner of its own bounding box.
  

-- Simon

Good guess!
Also with this post I am heading to bed so don't wait up   

White koans:
  AAA    HHH    OQ
* DO     I      QO
  E      MEL    U
  EEE    N    * VW
  END    NED    VYT
  ENE    NEE    WV
  ENN    NEN
  EXD    NNN

Black koans:
  AE      CH                 LEM
  AJ      DDD                LEMMING
* AV      DNE                NDE
* AW      EAGLE              STA
  BEGIN * EUM                STAR
  CD    * FFFFFFFXXXOZZAAA   START
  CDE     FINISH           * WM
  CED     FJ

All single letter koans are white.
E*D is white for: E H I K M N X Z (and black for every other letter)
(**), (***) and (****) is black for: B D F J L P R (and white for every other letter)

Quote from: Ramon on November 19, 2024, 11:24:43 PMit makes for a slightly better game experience if guesses are made with a goal in mind and not just with a set of RNG letters

I totally agree.

I'd say the main problem with the letter-based Zendo is that it doesn't preclude the use of strings such as "FFFFFFFXXXOZZAAA" and other nonsensical and random strings of letters.

A numerical game might be better suited to this format, or maybe add a rule to the letter-based game that all koans must be recognisable strings of letters.

Such a rule would remove the need for guesses such as "E*D where * is each of the 26 letters of the alphabet in turn".

But, it would also massively reduce the scope of the letter-based game.

Anyway, new guess for this round:

FFFFFFFXXXOZZAAA

AV
AW
DO
EUM
VW
WM

-- Simon

I'm totally fine with no limits if nobody just abuses it (which nobody has done so far and I don't imagine anybody here would do so).
The example you mentioned is exactly what I am afraid of, because if each round of Zendo could start with such a general broad wildcard guess, why even keep the koan guessing part and not just have the Master auto-mark every koan up to a specific length, reducing the game just to guessing the rule? Purely hypothetically of course because that's way too many koans to list and read through manually and it kind of takes away the part where people "work towards" the rule by basing guesses off of existing marked koans.
When I first read WillLem's wildcard guesses, I interpreted them as Simon's example and it made me a bit uncomfortable due to the above reason, but realizing afterwards what the guess actually looked like, it definitely has justification coming from the flow of previous guesses and koans. Not that justification is required, but it makes for a slightly better game experience if guesses are made with a goal in mind and not just with a set of RNG letters.

White koans:
  AAA    N
  E      NED
  EEE    NEE
  END    NEN
  ENE    NNN
  ENN    OQ
  EXD    QO
  HHH    U
  I      VYT
* MEL    WV

Black koans:
* AE       EAGLE
* AJ       FINISH
  BEGIN  * FJ
  CD       LEM
  CDE      LEMMING
  CED      NDE
  CH       STA
  DDD      STAR
  DNE      START

All single letter koans are white.
E*D is white for: E H I K M N X Z (and black for every other letter)
(**), (***) and (****) is black for: B D F J L P R (and white for every other letter)

MEL
AE
FJ
AJ

zendofmt should support a free-form notes section that gets printed below black and white lists. namida's tool supports that, and here we need it again.

-- Simon

I'd be reluctant to introduce a hard limit per day, or a hard limit between posts.

I see why you would want that though. It would lose some of the spirit if every game begun with:
foreach (1 ≤ len ≤ 10), foreach ('A' ≤ x ≤ 'Z'), test xxx...x of length len.

People haven't done it so far, therefore I'm hesitant to enforce anything.

I've tested one batch of 26 in Proxima's round, and even made the choice of batch dependant on single instances of the batch that I predicted. If you really want to enforce a limit, such prediction seems like the better way. But then you'll be hung if you see no pattern to predict, but know that you want the batch tested. I still advise against any kind of limit.

Another argument that's weakly against introducing a limit: I conjectured that all single-letter koans were going to be white in Ramond's round here even after Ramond broke my theory about convex hulls. But I wanted to keep that for myself. I wouldn't have mass-tested the singles.

-- Simon