Exploring Wordle — Has the game become tougher after NYT took over?
There have been quite a few articles over the last few days about how Wordle has become harder after the game has been bought by New York Times. While NYT has said that they’ve not made any changes, we will have enough people who will not believe this.
I was wondering if there is a sensible way to look at this problem with the help of data. I was aware that there are many scales for readability that can be used to determine whether a piece of writing is complex or simple. One of the more famous measures of readability is the Flesch-Kincaid scale. An example of this scale in action was when someone looked at the 1st 30,000 words spoken by presidents using the Flesch-Kincaid scale. What came out was that Trump spoke at the level of a fourth-grader while Obama spoke at the level of a ninth-grader
There is a limitation, though, in the way this particular scale (and other readability scales) can be used for any analysis on the difficulty of words used in Wordle — these scales require full-fledged sentences whereas what we have as input data is just words and more specifically, five-letter words.
I was looking for an equivalent that can be used for words and found one in the form of Age of Acquisition(AoA). AoA is a very simple metric that tells us the age at which a word is typically learned. One analysis done by Kuperman et al, the word MOMMA had an AoA of 1.58 while the word BIRTHDAY had an AoA of 2.85 and MEGALOMANIA an AoA of 18.92. Makes sense!
I wanted to look at 3 data sets.
1. The Wordle solutions from the start of the game till the NYT announcement of the acquisition (i.e. Jan 31st)
2. The time from NYT acquisition announcement till date (Feb 01 — Feb 18, 2022)
3. The post-acquisition Wordle solutions (we know the solutions till Oct 2027) — Feb 01, 2022, to Oct 2027
My hypothesis was very simple:
The average AoA of Wordle answers pre-NYT and post-NYT would be comparable. The notion of things getting more difficult is psychological rather than borne out by the data.
I used the 31,124 words in the Kuperman AoA data set to check for scores — the data set contains AoA for 2,792 five-letter words.
Of the 2,315 words in the Wordle data set, there were 263 “no matches” when compared with the Kuperman AoA data set — i.e. Words like BIOME, ISLET, MOULT or CAULK did not have any AoA attached in the Kuperman data set. I excluded these words from the analysis.
When I compared the average AoA scores, the results are tabulated below:
In the long-term, the AoA goes up marginally i.e. from the current 8.96 to 9.09.
For the last 17 games since NYT took over (note: CAULK has no AoA), the games have been a lot simpler in reality — the AoA has gone down from 8.96 to 8.62.
So, for all the people crying out that the game has become a little more difficult, (That includes you @TrevorNoah), the answer is a BIG, RESOUNDING NOPE!
Image Credits: mali maeder (Pexels) and Andrea Piacquadio (Pexels)
