How I set puzzles

Before putting the puzzles up on this site,I solved them myself. Since they are all more than a year old, I'd largely forgotten them and could approach them almost fresh. I was gratified to find that they mostly fulfilled my ambition for my puzzles. Which is that they should be puzzles I'd enjoy solving myself. After 40 years solving, I've got some pretty well-developed tastes in crosswords and, in the time-honoured phrase, I know what I like.

Since I know what I like, I think I'm on reasonably safe ground if I compose grids and write clues which I think I'd like myself. Trying to satisfy someone else's requirements is a lot harder. You may note that (at least at the time of writing) there aren't any barred puzzles here. The reason for that is that I largely don't enjoy solving them. It's not that I can't solve them - I reckon on taking about half an hour to do a standard Azed - but I don't find them that much fun.

I'll look at the preamble to a Listener or Inquisitor and see if I can be bothered to even start. If the instructions contain any hint that I'm going to have to rearrange the letters after I've put them in the grid, I immediately pass on: I use a pen when solving, so "rub it out and start again" isn't going to happen. Nor am I going to waste my time with the ludicrous Playfair complication. I do enjoy "each across clue contains a misprint of one letter in the definition" or "wordplay ignores 15 objects of a kind", though. However, I see no reason why the types of playing about that I like can't be accommodated in the standard 15x15 blocked puzzle - and the messed-about clues section of this site can perhaps be regarded as my argument in favour of that proposition.

Even then, I don't aim to make such puzzles brain-strainers: I'm not really in the business of tying people's minds in knots. What I want to do is to take solvers on a walk round the English language and point out some amusing curiosities, get them to search through their mental attics for those nuggets of nonsense they've acquired over the years, and hopefully intrigue them with some intellectual gymnastics which they'll find challenging but by no means impossible.

I've found that I don't have too much control over how hard a puzzle will turn out to be. In fact, my deliberate attempts to be tricksy and difficult have tended to end up embarrassingly unsuccessful: my test solvers regularly return them saying "that was nice and easy", which wasn't the point at all. I can usually tone a puzzle down a bit if it comes back from test as freakishly tough, but it's a lot harder to simply re-clue a few words to toughen it up: to toughen, I have to change the words to ones which yield more difficult clues.

I say "yield" because I'm not that good at thinking up clues in the absence of a grid. Alberich's sage advice on his site is that you should have an idea of how you're going to clue a word before you put it into a grid, but that's advice which I largely ignore. What I do in practice is to fill up a grid and then discover what clues there are in it.

Though Hemical's espncricinfo series came to an unfortunate end, composing those cricket puzzles was an incredible technical boot camp. Having to come up with a variety of intelligible clues for names like HARBHAJAN and IQBAL QASIM teaches you that there is no such thing as an uncluable word. All (!) you have to do is to walk round the word, look at it up and down very critically and see what clue is available. It's not really that different from solving: the solver gets some instructions and has to work out how they lead to a word, while the setter has a word and has to work out what the instructions to assemble it are. Some words just turn out to be bloody complicated to put together.

My hypothesis is that the chief determinant of a puzzle's toughness is what one might call its obscurity index: of course it's possible to come up with fiendishly complicated clues for very simple common words, but it generally takes more ingenuity than I have available. Megastars like John Henderson or Richard Rogan can contrive mindbending clues for the simplest of vocabulary with apparent ease, but I find it hard and don't really see the point anyway. Solvers can usually do with the odd simple clue to get them started, after all.

And I'm not all that keen on very obscure words. I'm very reluctant to put a word in if I don't know approximately what it means without looking it up. I'm not a walking Chambers, so there are plenty of words around with which I have only passing acquaintance, so I have to look them up to get an exact definition, and I'm frequently surprised by what I discover - but I guess most people have heard of words but have completely the wrong idea about what they mean, so I'm hardly alone. If I'm setting a plain puzzle, I won't use anything I'm not pretty damn sure of before I look it up; I allow myself a bit more latitude with themed puzzles, and will accept one word which normally hides under a thicket in a dark forest - but I'll make strenuous efforts to make the clue for it as near a gimme as I can (which isn't always very near, but still).

My early puzzles relied to a heavy extent on anagrams. They are by some distance the easiest clues to devise, at least in my experience. But feedback eventually convinced me that I ought to at least try and find something else before falling back to alphabet salad. The Times limits setters to no more than five anagrams in a 15x15, which seems a bit low to me as a limit: I'll allow up to seven nowadays (although it's usually under five and I don't mind if it's zero).

I do have to impose rigid limits on myself about some clue types. I love what the Reverend Spooner can do in a clue, but I'll only allow him one appearance. I also love hiddens, but they're capped at three per puzzle now, and not all in the same direction.

On the other hand, there are clue types which I dislike and employ very rarely, if at all.

Ximenes said "a definition, an indication, and nothing else", and I thoroughly agree. I don't like clues which don't follow that pattern. The cryptic definition clue only allows one way in for the solver, which I think is rather unfair. So I almost never use CDs. I have, though, found the odd word which has to be in the puzzle because nothing else will fit and the theme prevents me replacing some of the words, and have used CD clues because everything else I've tried has ended up unintelligible or fifteen lines long. So there are two CD clues here.

Conversely, I dislike clues which have more than two components. I recently saw a septuple definition clue: several of those who commented on it on 225 thought it fantastic - I hated it. I have, however, perpetrated two tripartite clues: one with a definition and two alternatives for wordplay, and one which has two definitions and a bit of wordplay. In both cases, I didn't think that the clue was reasonably solvable with only one of the bits which I "duplicated".

Double defs can be wonderfully elegant, but I get very sick of them very quickly. (It will be apparent that I'm no fan of Rufus/Dante's style: I no longer bother attempting his puzzles because I will be bored and irritated before I'm halfway through. It's just too much of a not very good thing to do a whole puzzle, much like listening to more than three songs by Status Quo - however much I like "Caroline", "Down Down" and the rest on an individual basis.) I'm not going to call myself the Antirufus, but those who find his puzzles irritating for the same reasons as I do can try any puzzle here without fear.

I'd also rather that people solved the clue the way it was written. Which means I don't do those multi-entry clues for great long quotations and phrases. Any more, anyway. I did a 28-letter anagram very early on, and there's a 40-odd letter saying - of which about 13 letters in the middle were anagrammed - in another puzzle, but there's an extent in which I did those to prove that I could. Generally speaking, I know I solve clues for seven-word entries by getting crossers and then working it out from the enumeration, and I have no reason to suppose that most solvers won't treat clues of mine the same way, and I get no satisfaction from spending hours working out some complex anagram just so that solvers with nothing more interesting to do can check that you can recombine 32 letters that way once they've already filled the solution in. Apart from the two I've mentioned, there's an anagram for BRIDESHEAD REVISITED, but that was right on my mental limit for what is worth doing.

I don't much like clues written in crosswordese. I guess this is where I have real difficulty with the True Believers in the cult of Ximenes. I reject the view that nouns can't have articles in clues unless the article is also in the solution. English demands articles, as it also demands the extraneous "to" in infinitive verbs, and that's the language I'm writing in. English trumps Ximenean rigour every time in my book. (If an editor wants to be a stickler, that's a different matter, but it often means that I have to find another clue.) I also try and avoid using bits of what is now crossword code rather than normal usage: no-one has referred to a Royal Marine as a "jolly" since the 18th century, so it's just stupid to use it in a 21st century clue. (Obviously one must make allowances for setters who were born in the 18th century and still speak Georgian English.) I'm not going to catalogue every hackneyed device I get sniffy about, but you know what I mean.

And what I absolutely won't do is use "boy/girl/man/woman" to mean "string of letters which someone might possibly use as a name" as a subsidiary indication. I don't doubt that one can construct very nice clues using it - I've certainly seen a few - but I find most such clues annoying to downright unfair. If I want to use TIM as a name, then there are plenty of real people I can allude to, and since I'm not setting for The Times I'm entirely at liberty to do so. I'm OK with defining the entry SUSAN as "woman", but I'm not putting her into "Girl trapped by cubes" as wordplay for DISUSANCE. I may well break this rule in 2022: Alberich bet me that I couldn't go ten years without it, and that's when that bet expires, so I might then construct a puzzle which has some random name in every single one of the 30 or so clues, hopefully in such a frustrating manner that I spark off a protest movement demanding that the device be banned forthwith.

(For the avoidance of doubt, I regard all of the above strictures as matters of taste. If people want to listen to James Blunt, they're entirely free to do so as long as they don't expect me to like him myself and don't whinge that I'm not him.)

I hope I've given you a flavour of how I approach setting. When it comes right down to it, what I'm doing is amusing myself. I can only hope that you find enough of the same things amusing and enjoy the results.