# Requesting Feedback on Check Sum for Word Linking Puzzle Stage

I want the community's feedback on a puzzle I am considering creating.

Let's say the solver is given, as one of the earlier stages in a puzzle, many sets of two words that need to be linked together. One set, for example, might be

iron ______ ______ spray

and one needs to figure out the linking words as follows:

iron stomach bug spray

where "iron stomach", "stomach bug", and "bug spray" are all known things.

I'm strongly considering including a check sum for both answer confirmation and also to prevent (hopefully) the solver from attempting to move to the next stage with an answer that isn't the intended answer, but that might also fit (however unlikely).

For the example above, the check sum would look like this:

iron stomach bug spray check sums: 27, 9

where 27 is the sum of the A1Z26 values for the first and last letter of the expected answer:
For "stomach", s(19) + h(8) is 27, and for "bug", b(2) + g(7) is 9.

It's exceedingly unlikely that an unintended answer would pass the check sum also.

The question:
Is this desirable? Is it helpful? Too helpful? Annoying? Overkill? Not enough? Other suggestions or ideas are welcome.

• How about just the number of letters for each of the two unknown words? In this case (7) and (3)
– cap
Jul 17 at 23:31
• @cap i thought of that, and maybe that is good, but I was thinking maybe it gives a lot of info, making the search too easy?
– JLee
Jul 17 at 23:34
• imo this is probably useful for those puzzles where there is a lot of text but probably annoying and overkill for small to medium puzzles. I also think this wont be massively adopted (No offense just being honest) because of the number of use cases seems to little or not significant enough to make a change in a big site like puzzling.se. But again this is my opinion Jul 18 at 0:14
• @Varun W. Not wanting it to be adopted by the community, or even by me- just asking for my current puzzle. Thx for the feedback.
– JLee
Jul 18 at 1:08
• Another option is to create two such chains for the pair of words you are trying to create e.g. also have "upset ____ ____ zapper" alongside your example chain, above. This makes it far less likely that unintended answers will fit.
– Stiv
Jul 18 at 7:35

The main problem is that there are too many degrees of freedom between the two given words; there's too much "play" in the three links between words, which makes the puzzle tedious to solve: anything can fit, so everything must be checked.

This cannot really be fixed with a checksum: if you intend to disqualify otherwise perfectly cromulent answers like

iron bell - bell pepper - pepper spray

or

iron cross - cross hair - hair spray

on no other grounds than "that wasn't what I was thinking", then the entire puzzle just becomes an annoying guessing game.

I like @cap's idea of reducing the search space with given word lengths, but even then you should be prepared to accept alternate answers.

• yep, makes sense. great answer, and I'm glad I asked this question. It's not so much that I want to disqualify other correct anwers, but that I imagine the next stages of the puzzle depending on the intended answers in order to function properly, so I will need to make sure that I tighten this up so that only those can make it through.
– JLee
Jul 18 at 1:06
• Chaining puzzles to one another is probably not such a great idea in itself; a single snag can ruin the entire experience, making the latter puzzles inaccessible. That said, if you have plans for getting your solvers past such snags, you'll almost certainly want to ditch this particular puzzle type, and only use ones where there can be absolutely no question that the only valid answer is the intended one.
– Bass
Jul 18 at 17:07
• Not a fan of multi-step puzzles?
– JLee
Jul 18 at 17:14
• As a general principle, no. Having the puzzles depend on one another may (non-trivial amount of ingenuity required) be a decent way to turn a set of puzzles into, say, a story (or an escape room), but having each puzzle independently produce a clue is almost certainly always better, if for no other reason, then because you won't have to spoil any puzzle to the solvers in order to help them along. (Which is a really common problem in the more difficult escape room settings.)
– Bass
Jul 18 at 17:24
• Exactly. If you have, say, ten puzzles "in parallel", and you can't solve the first one, you can solve the other nine, and thus, very likely, the overall puzzle, and maybe even use the overall solution as a hint for that first puzzle. If the puzzles depend on each other "in series", then failing to solve the first one means the rest of the puzzles are inaccessible, and could just as well not exist at all.
– Bass
Jul 19 at 5:18

If the checksum method is described (and it needs to be, or else you can't check that you got the correct answer), that limits the search space too. With the checksum you proposed, 27 can only be made with (A,Z), (B,Y) ... (M,N). Incidentally, this number is the one with the most combinations. Smaller (or larger, for that matter) also limit which characters can make the answer. A checksum of 50 means the letters must be (X,Z) or (Y,Y); a checksum of 3 means it is (A,B).

It may be possible to create a checksum composed of A1Z26 numbers (for first or last) and length, or a straight-up sum of all of the letters, which will work better, not give away too much, and is still easy to check that the answer matches with few (if any) false positives.

In any case, such a checksum needs to be a) simple to confirm; b) reasonably resistant to false positives; c) not make the search too easy, and meeting all three might actually be a hard problem to solve in the first place.

• All excellent points. After sleeping on it, I think personally I would like the checksum if I was the solver, as it gives very strong confirmation that I'm on the right track, and would just be another constraint of the puzzle. However, I realize that everyone is different, and that Bass and many others would find it annoying.
– JLee
Jul 18 at 10:19

I wouldn't use checksums. I would compare strings, in a case-insensitive form, and script a tolerance to spelling mistakes or typos. For example, an extra letter, a missing letter, 2 inverted lett(re)s, one lette(z) not as it should be, etc.

Personally, I would ignore the checksum because I'm interested in trying to find the words, not in looking at their checksum, but I would welcome a hint possibility showing the first letter of each missing word.