# Best way to solve Codeword puzzle

I always find it hard to do codewords puzzles. Whenever I fill in all the spaces with the starting letters, it doesn't really help. I always end up looking at the answer sheet. Does anybody know a good strategy for codewords?

• Years ago I wrote a codeword generator program for a publishing house. They specifically asked for it to be configurable for any given puzzle, to maximise or minimise the number of times any given letter appeared in the grid. And that no puzzle should be "ambiguous" (as, for example, if the only instances of the letters K and L were in a position where the word could be KALE or LAKE). My father used to do the Daily Mail codewords every day, but he always started by assuming the most common number must represent E (so he hated most puzzles made by my system! :) – FumbleFingers Jan 13 at 18:15

Codeword puzzles are usually designed to have an approachable way of starting. (I am fairly sure it would be possible to make one that was solvable in principle but really hard to find one's way into, but the people who make them aren't aiming for that.) I think usually there is at least one word in the grid that (1) has one or more letters that are already known at the start and (2) has repeated instances of some other letter or letters that narrow down its options. Or else there's a word that has #1 as above and (2') at least one of those already-known letters is unusual and itself narrows things down a lot.

For instance, I just looked up a random Codeword puzzle available on the web, and it has starting letters D,L,Z and one of the words in the grid is --Z-D. There aren't a lot of posibilities for this; I see DAZED, FAZED, LAZED, RAZED; maybe FUZED in American English (but this is a UK newspaper so probably not). At any rate, that E is very likely and the A is quite likely.

• Technique: spot a word with an unusual letter or letter-combination in it, think what possibilities there are for it, and try out some that seem likely.

At this point, the thing to do is to look at the grid and see whether there are words that become either easy or impossible if these guesses are right. If you're doing it online, you can put those guesses in and easily undo them if they turn out to be bad guesses. With pencil and paper, either do it in your head or write lightly and be prepared to erase.

With the puzzle I'm using as an example (I'm deliberately not linking to it because (1) I don't know whether it will be available online for ever and (2) the approach is the important thing rather than these details), if I enter the A and E, the next interesting things I find are D----EED, which seems like it should be very informative but nothing springs instantly to mind, and D-A-. The latter is interesting because we already have half the letters, and because when we have a vowel and a consonant with only one letter between them it's often quite restrictive. Here, the letter in between is probably E (impossible because we already have E), R (possible), or I or U. (Any others? Haven't thought too hard. We're just trying things out.) Let's try R first. Ah, that gives us the R of RAZED, which also tells us that a vowel will be no good so actually R is probably the only option.

• Technique: places where it's impossible for vowels and consonants to alternate are often quite restrictive. Look at those.
• Technique: when making a guess, look to see whether it interacts usefully with other words you've already got.

In my current example puzzle, this annoyingly still doesn't make anything super-obvious, but we now have -A-L-R as a word. Between L and R there needs to be a vowel; we already have E and it seems like O is the only possibility. TAILOR, SAILOR; anything else? Dunno, let's try putting in the O and the I, and see what happens.

• Technique: between two consonants you often have to have a vowel; there aren't very many vowels so this is a useful constraint. (But be aware that consonant-triples are possible!)
• Technique: the ends of words are sometimes more stereotyped than their beginnings, and may be useful places to look.

At this point I notice that there are some very long words in the puzzle. One of them is -LO-ALI-A-IO-. There's an obvious word to put here, but let's pretend I don't see that; at any rate, it seems like there's an obvious suffix: ATION. So put that in and see whether everything still makes sense. It does, and at this point in fact there are enough fairly obvious words that I'm pretty sure my guesses so far are good and I can solve the puzzle from here.

• Technique: very long words can be quite restrictive.
• Technique: look for likely suffixes. (And prefixes, but I think suffixes are commoner.)

It could happen, instead, that you get an obviously impossible situation. That means that one of your guesses was wrong. Pick whichever guess you were least confident about, and undo it or, probably better, replace it with your next-most-likely guess and see what happens.

How well any of these techniques work for you will depend on how good your vocabulary is and how good your brain is at answering questions like "are there any words matching these letters?". I don't think there's much I can do to help improve those :-), but solving Codeword puzzles probably helps at least a little bit with both.

• With --Z-D, you can't have DAZED (since the leading D would be known). This also eliminates DOZED. ADZED and OOZED can be eliminated by similar reasoning. – Toby Speight Jan 13 at 14:31
• True enough. (I was just documenting my thought processes as I went. I'm sure I missed a thing or two.) – Gareth McCaughan Jan 13 at 14:48

Once you've inserted all of the starting letters, I normally see if any of the words (or certain sub-strings of them) pop out to me. For example, given ----I-G, you can be confident that the penultimate letter is a N. Comparing words in the puzzle with common prefixes and suffixes often gives a few more letters.

Also, when checking a word, remember that double letters will be discernible since they are both clued with the same letter. If you ever find a letter, immediately fill in the other letters with the same number.

From there, your first priority is to get the common letters out of the way, in particular, the vowels. For larger grids, this can be done through frequency analysis. Count up the number of occurrences of each number. Suppose that the most prevalent numbers are 15 (24 occ.), 9 (17 occ.), and 26 (15 occ.) in a grid with just under 200 letters.

Since we expect e to be the most common letter, appearing 12% of the time, we could reasonably guess that 15=e. The next two letters, t and a, could be either way around, considering the small size of the sample, but you should be able to do it by looking at the board. Suppose one word is -IN*, where *=26. We would thus guess that 26=t.

Note that one source finds the most common letters in a codeword to be, in order, eairtonslcup.

Other than these strategies, trial and error can prove fruitful. In particular, short words can reveal possibilities, while longer words are useful for checking and eliminating the possibilities (since, while it is harder to think of words that fit, it is easier to verify that no word matches -n-at-e-n. If you find this difficult, use an online solver at first).

Try to forget about the uncommon letters until you have gotten a large chunk of the puzzle out of the way. Generally, they'll only appear in one position. Keep in mind which letters are left. When you get down to the last ten or so, I write them in a list beside the Codeword. Remember that these remaining letters are the only ones that you can insert.

If you run your mind through the options for a--t, it is useful to recall that u may be the only remaining vowel. Since three consonants in a row is unlikely in such a short word, we expect the u to be present, which almost instantaneously reveals two common options: aunt and abut.

My final piece of advice is to simply ensure that your vocabulary is up to scratch.

Good luck, and be persistent. The only way to improve at these puzzles is to practice.

My favourite CodeWord trick is to find several partial words with the same uncoded number, and this can often reveal the letter.

For example, say you have:

_EST, BE_T, _ICE

Each option has several words that fit:

• best, gest, lest, nest, pest, rest, west
• beat, beet, belt, bent, best
• dice, lice, mice, nice, rice

and only _=N and _=L fit them all, which narrows it down a lot, and other clues often reveal the answer completely.

I am not very good at word games. I regularly use a hand-held crossword solver to help me finish quick crosswords. (Forget about cryptic crosswords. Well, I did finish one, but it took me a week, and my evenings were completely free because I was on holiday [vacation] at the time.) I occasionally can use this for codewords as well, particularly if there are multiple starting letters in a single long word, which narrows down its output considerably. I still have to manually check for appropriately duplicate letters in the results though. (I also have to watch out with this particular model as it omits lots of standard endings, such as plurals, and its dictionary is not that large, and it uses -ize endings which are unhelpful when solving British codewords.)

When I get really stuck, then there's a website that can look up words based on the duplicate letters. This is really helpful not just for ensuring that letters are the same when they should be but also that they are different when they should be!

I start by looking for the numbers that appear most often. I then look to see which of those in the short list appear at the ends of words. That often (but not always) helps to identify the E. (A frequent number that does not appear often at the end of a word may be an A, or sometimes an R,S,T etc. If there's a problem distinguishing between an A and and E, look for "doubles" (EE). AA is unlikely (exception: bazaar).

If really stuck there are two escape routes. The first is to try locating the combination (QU). That is occasionally possible by simple inspection, with one or generally no more Qs, with their following U. Many possibilities for Q can be eliminated (if the letter is no more than 2 at the end of a word. The combination -QUE is quite common on the end of a word. Last solution (time-consuming) is to make a list of all the numbers and count up the number of times they appear, and then try and decide if the most common are from the shortlist "ARIESTA" or similar. It's usually possible to pinpoint the QU from such a list, by looking at numbers that appear just once or at most twice.

I hope that helps. I've done scores, possibly hundreds of codewords, and using the above tricks only had to look up the answer (where supplied) on a single occasion. (No, I'd never dream of paying to access clues over the phone...).