You may have noticed a bit of a gap in “Anagram Month”…to the extent that it no longer fits in a month. We have been waiting on Jerry to post an intro to our new anagram generator. Nothing much, just a few lines. The Editor should have known better than to trust Jerry with even a task that simple. But Jerry lost all his fingers in an accident trying to learn some of Blaise Serra’s sleight of hand techniques using a deck of cards made from thin sheet metal. After a month of poking individual letters with his nose for hours each day, he still hasn’t finished that post, so we’re going over his head and skipping it. I’ll summarize it what it was intended to contain here:

• We have a brand spanking new anagram generator. (Where by “new” I mean “have been sitting on for months”.)
• You can use it to automatically generate progressive anagrams, frontloaded progressive anagrams, two kinds of transgressive anagrams, and commutative anagrams.
• You can even create them by hand (if you have fingers and an aversion to anything resembling AI).
• For most anagram types, the longer you wait for a solution, the better the solution you’ll get.
• Sometimes, immediately running the same search again can get you different results, which you might like better!
• You can bookmark a direct link to a particular anagram you like to study it again later.
• You can practice an anagram you’ve created/generated on our anagram trainer with a single click.
• If you find any bugs or have any suggestions, shoot me an email.

You can find the generator here.

Now, on to the real post about what makes our generator so much better than everyone else’s!

---

Over the last 20 years, a number of different magicians have authored software to generate progressive anagrams, and now I join their ranks. Even now, one can find several different progressive anagram generators you can use free of charge. Friends, Oval subscribers, mentalists, lend me your ears: I come to bury these generators, not to praise them.

To begin with, it’s almost not even fair to compare ours with those others that are extant today. We aren’t even aware of any freely available generators for automatically creating Transgressive Anagrams, and we are certain there are no others for generating Commutative or Frontloaded Progressive Anagrams, but I’m willing to say that even if you don’t take these features into account at all, we’re still winning.

For example, to generate Progressive Anagrams (PAs), our generator, like all the others, searches for anagrams that minimize the maximum number of “no” answers needed to get to any term. But, as I previously mentioned, that criterion will usually leave a lot of wiggle room. There are a lot of potential solutions that minimize that quantity, so we use more criteria in addition to that to decide amongst those solutions. It’s those additional criteria that give you The Magic Oval Difference™. Here’s a short list of other things we have devised ways to select for:

• Prefer to ask about letters that appear more frequently in the English language generally
• Prefer to ask about letters that appear more frequently in the specific list of terms
• Prefer NOT to ask about letters at the beginnings of words within terms
  • If this is unavoidable, prefer to ask about them as soon as possible
    • But better as the second question than the very first question, so as not to create the expectation that you already know the answer too soon
• Minimize the total number of “no” answers across the entire anagram
• Minimize (or maximize) the number of different letters that could be asked as the nth question.

These considerations not only lead to anagrams being chosen that are more suitable for specific kinds of presentations (such as Squared Anagrams) and increase the chances of several of the common pitfalls of anagrams being avoided, they also have a tendency to result in anagrams that are easier to memorize. On top of that, the flexibility of the algorithm makes it easier to quickly iterate a variety of different criteria and assess the utility of the anagrams produced.

This is precisely what led me to the discovery of Frontloaded Progressive Anagrams (FPAs). Our basic Progressive Anagram criteria already prefer FPAs unless they require the maximum number of “no” answers to increase. Once I saw some of these results coming in, I noticed how much nicer these were to memorize and script than the PAs coming out of existing generators. It occurred to me that it might sometimes be more helpful to allow the number of initial “no” answers to be higher than necessary if this concession would allow one to immediately begin down a perfect PA that was as short as possible. In other words, you might be able to split the difference between the best of both PAs and Transgressive Anagrams. And the various results this experiment eventually yielded were so nice that both types are now included with the generator. We suggest always trying both to see which gives you a result that matches best with your own personal criteria.

And that’s just one example of the kind of results we have been able to achieve by acknowledging the vastness of anagram solution spaces and opting to wade through this space and grade what we find there on a variety of criteria. We know there is a lot left to discover here, and now we—and you—have exactly the tool we need to explore that world.

Examples

Star Sign Progressive Anagram

On the left, you can see a star sign PA generated by a popular online PA generator. On the right is one that our basic PA solver eventually recommends.

At first glance, there’s no reason to prefer one or the other. The competing anagram has some advantages and ours has others. So why would I personally prefer to use ours?

For one, the anagram on the right is slightly easier to memorize. There are only two places where the tree branches. After the first two questions, you’re on a single track perfect PA. The third letter you name is always “R” and the fourth is always “A”. Meanwhile, the competitor’s anagram has four possible side branches, and sometimes you’re naming “C” and other times it’s “A” before “C” and one time it’s “B”. It’s still not bad because star sign trees are always small, but every little bit helps, you know?

For another, I reckon it looks pretty bad for your typical PA presentation to have two misses in a row. If I’m going to miss twice in a row, I’d rather it not be after a long series of guesses—for a Capricorn, the anagram on the left has two consecutive wrong guesses after a winning streak of three correct guesses, which means I have to go from building up a lot of “this is working!” steam into “guess this isn’t working after all.”

Finally, ours never asks about the very uncommon letters “B” and “G.” The former is particularly egregious to name given that it appears in only one star sign (although if you’re going to name it, doing so right before you get the right answer is the only time it makes sense, so I’ll let the competitor off the hook this time).

But, as I said, we recommend also seeing what can be achieved using the Frontloaded PA criteria. If you fire that one up and let it cook for a while, eventually a little miracle pops out:

Yes, at first glance, this one is actually worse than our anagram above. You miss 3 times in a row for a Leo, and overall you usually have to ask more questions. So what do we gain in exchange for these losses?

Well, the main thing you gain is that the sequence of questions is far easier to remember: You’re only ever asking the letters CIRASUT in that order. And given that all the branching happens on C, I, or R, and after you know those answers you have a perfect PA ahead, it follows that the order those three letters are asked has no effect on the anagram as a whole. If you don’t like asking “C” first given that two signs start with “C”, our tool provides you with the ability to edit and construct your own anagram manually. You could manually edit it so “I” comes before “C”:

The fact that it can be rearranged thusly is what makes it so flexible for constructing nice presentations around. That’s why I consider it a miracle.

Grouping

One common issue with anagrams is the fact that small variations on words in English can often have quite different letters in them. And you can’t always be sure what spelling a participant has in mind. The only ways to account for this variation are either to carefully word your prompts to avoid problematic variants or to add every possible variant to your anagram. For example, if you ask someone to think of a type of flower without careful wording, they could be thinking of a single flower or a whole bunch. Sure, in some cases, that doesn’t make a difference: “Roses” uses the same letter set as “rose” and no one is going to be thinking of “lavenders.” But it can make a huge difference whether they’re thinking of “daisy” or “daisies.” If you want to cover all those bases using the “add every variant” strategy on a competing generator, you could get this tree:

Notice how it branches on “S” so that “violet,” “daisy,” and “poppy” end up nowhere near “violets,” “daisies,” and “poppies.” Meanwhile, our generator allows for the use of syntax that lets you group together words as if they were a single word and prevents them from being split up:

Of course, you can use the same feature to group together wildly different words that you don’t need to distinguish between as well. For instance, maybe you are doing a zoo animal anagram in the zoo and hope to group together animals based on what area of the zoo they can be found in. You can drop [gorilla/monkey/orangutan] in as a word as well. (However, if this added complexity creates a situation where no anagram even exists, unfortunately our generator won’t be able to detect that fact and warn you.)

Conclusion

We’re not guaranteeing results like the above when you use our generator. There will be tweaks to the algorithm as we improve its ability to target our preferences. As mentioned above, the space of possible PAs for a given set of terms is vast, and even tiny changes to our criteria can result in vastly different results. However, we do guarantee that, as of this writing, you won’t get better results from any other solvers you can find. And if, for some reason, you’re finding our results aren’t suiting your particular needs, feel free to shoot us an example of the kind of results you’d rather be seeing and maybe we’ll be able to coax outputs more to your liking out of the algorithm. (Likewise, let us know about any bugs you encounter!)