[triword]

It doesn't use obscure words. As far as I can see BRALL is an archaic alternative spelling of BRAWL that's only listed on a site for obscure words. Dictionary.com and Merriam-Webster both have no result for that, so...that wouldn't really be very fun if that was the solution, would it?

[smcin]

Your site only says "words" or "complete words", it doesn't reference any dictionary. BRALL is a word, albeit an obscure one [0] that is not found in most dictionaries.

[0]: https://www.wordnik.com/words/brall

[bena]

It's found in one dictionary. Last published in 1911.

Brall isn't an obscure word, it's an obsolete one. I think it's fair to say that if your word isn't in Oxford or American Heritage, you shouldn't expect it to be in the result set.

It feels as if you hit "BRASS" and "ZEBRA" then couldn't make it fit with "LL", then searched for "brall" hoping to have a hit.

But if you used WinAMP back in the day, you'd have whipped this one in the ass.

[xmprt]

One solution to this problem is to have two dictionaries. One complete dictionary and one common dictionary. You pick puzzles using the common dictionary and then check the obscure/complete dictionary to make sure there are no alternate solutions.

[triword]

If I omitted any puzzle with alternate obscure solutions, the vast majority of puzzles would be invalid. There are a lot of obscure words.

If on the other hand I set it to accept the alternate obscure solutions, that actually could ruin the puzzle in my opinion.

Take today's with BRA for example - it's a nigh-certainty that the commenter suggesting BRA did not know the archaic alternate spelling of BRAWL. What's much more likely is that they saw "BRASS", then checked ZE and saw "ZEBRA", and then thought "I wonder if BRALL is a word?" and googled it to find that it was indeed a word in a very obscure sense.

Check this google n-gram for a reference of just how obscure that word is - it hasn't registered meaningfully anytime in several centuries. I added in the word "quine," another obscure word, just for comparison, and you can see it's orders of magnitude less common even than that. https://books.google.com/ngrams/graph?content=brall%2Cbrawl%...

So what would start happening is people would find something that worked for 2 of 3 (which is easy), type it in cuz why not it might work, have it succeed because the 3rd option was written once by a Gregorian monk in a remote monastery in the Balkan Mountains, and be like "huh, well that's dumb." It's actually a lose-lose that would let people off easier because they'd feel like it was done just because they found something that matched 2 of 3 and happened to match the 3rd in a fairly ridiculous way.

Edit: I came up with a solution to this. I'll verify each puzzle against a full dictionary with obscure words included, and if a user puts in a valid obscure solution I'll display a message saying "Valid, but at least one of these words is too obscure. A better solution exists." It'll take a little time to add that functionality though.

[kylecazar]

The very first thing I did was check to see if brall was a word and then moved on when I realized it wasn't. Fun game!!!

Out of curiosity though, do you ensure there is only one solution for each challenge?

[triword]

Programmatically, no, but I've made about 1500 puzzles and been sharing them with friends for months and neither I nor anyone else have found a double solution yet.

It would happen if I picked generic clues like BR/SL/ER (not a real puzzle, don't try to solve it), but I make a point not to. As long as you pick clues that are a little more specific than that, it's actually extremely improbable, despite what you would intuitively think (I had the same thought in the beginning!). Remember that there are 4 restraints, not 3 - it has to match 3 different prefixes or suffixes, but it also has to only be 3 letters (4 letters on weekends).

Sometimes I'm suspicious and I do verify them, but it's really only a risk with the compound word versions like SELL/LAW/FOX. Those I have to be more careful with.

[Karellen]

I'll point out that 26³ is only 17,576, so brute-force checking that there is only one solution for a given problem shouldn't be computationally difficult.