Wordle Analytics

Wordle is a fun game, but I just couldn't resist knowing what a great first / second word choice would be.

There's multiple strategies available here in this repo, the best strategy we have so far is listed at the top. Feel free to add ideas as an issue! Orginally I went with something really naive as I didn't want to think too hard, but then I opened pandoras box, and here we are!

One core tennat here is that we don't cheat while we're playing the game, we just want to know what an optimal strategy is going into the game.

Wordle design choices

Wordle has every possible 5 letter word available as a guess, and there's a seperate list of possible answers due to the fact that a lot of the 5 letter words are very rarely used in english. Due to these two lists our best strategy takes the possible answers into account while determining what the best first word to use is! Note that not every strategy on this list has this feature due to not knowing about all of this before hand! https://notfunatparties.substack.com/p/wordle-solver was a good resource that explained the answers vs question list.

Best first word (if best is defined as most number of letters matched in the first guess on average)

Test every combination of Question (~12k words) to Answer (~2k words from Wordle's list). Pick the word with the most number of letter matches.

Some improvements available like right now we double count letters if the answer word has the letter duplicated.

SQL

https://github.com/AutoIDM/wordle/blob/main/actual_best_first_word.sql

Results

guess	sum	avg	stddev	max	min
roate	4594	1.9844492440604752	1.0355479830658308	5	0
oater	4594	1.9844492440604752	1.0355479830658308	5	0
orate	4594	1.9844492440604752	1.0355479830658308	5	0
realo	4584	1.9801295896328294	1.0105506429868332	5	0
alert	4559	1.9693304535637149	1.0694998512424874	5	0
alter	4559	1.9693304535637149	1.0694998512424874	5	0
taler	4559	1.9693304535637149	1.0694998512424874	5	0
later	4559	1.9693304535637149	1.0694998512424874	5	0
artel	4559	1.9693304535637149	1.0694998512424874	5	0
ratel	4559	1.9693304535637149	1.0694998512424874	5	0
arose	4534	1.9585313174946004	1.0107501271667612	5	0
soare	4534	1.9585313174946004	1.0107501271667612	5	0
aeros	4534	1.9585313174946004	1.0107501271667612	5	0
retia	4511	1.9485961123110151	1.0308304749950321	5	0

First word by letter matching

Just looking at the number of letters that match the answer word, what's the best word for generating one letter match? What's the best for generating 2 letter matches?

SQL

https://github.com/AutoIDM/wordle/blob/main/best_first_word_by_word_matches.sql

Results

matcher_strategy	word	words matched	total_word_count
oneletter_matcher	toeas	12417	12972
oneletter_matcher	stoae	12417	12972
oneletter_matcher	aloes	12412	12972
oneletter_matcher	aeons	12407	12972
oneletter_matcher	soare	12395	12972
twoletter_matcher	rales	5995	12972
twoletter_matcher	laers	5995	12972
twoletter_matcher	lears	5995	12972
twoletter_matcher	earls	5995	12972
twoletter_matcher	laser	5995	12972
threeletter_matcher	stear	1724	12972
threeletter_matcher	resat	1724	12972
threeletter_matcher	aster	1724	12972
threeletter_matcher	rates	1724	12972
threeletter_matcher	reast	1724	12972
fourletter_matcher	stear	541	12972
fourletter_matcher	resat	541	12972
fourletter_matcher	rates	541	12972
fourletter_matcher	aster	541	12972
fourletter_matcher	reast	541	12972
fiveletter_matcher	stear	111	12972
fiveletter_matcher	resat	111	12972
fiveletter_matcher	aster	111	12972
fiveletter_matcher	rates	111	12972
fiveletter_matcher	reast	111	12972

Naive approach looking at frequency of letters used

Looking at all of the 5 letter words available what's the most commonly used letter?

Most common letter

SQL: https://github.com/AutoIDM/wordle/blob/main/most_used_letter.sql

Results

Most used letters

Letter	Count
s	6665
e	6662
a	5990
o	4438
r	4158
i	3759
l	3371
t	3295
n	2952
u	2511

First Word

Take the top 5 letters from the most common letter and find a word

SQL

select * from public.wordle
where 
word like ('%s%')
and
word like ('%e%')
and
word like ('%a%')
and
word like ('%o%')
and
word like ('%r%')

Results

1. arose
2. aeros
3. soare

Best second word

If no matches on the first word, best second word based on the next 5 most commonly used letters

SQL

--
select * from public.wordle
where 
word like ('%i%')
and
word like ('%l%')
and
word like ('%t%')
and
word like ('%n%')
and
word like ('%u%')

Results

1. unlit
2. until

Run analytics yourself

1. Setup postgres ( podman run -e POSTGRES_PASSWORD=postgres -p 5432:5432 -h postgres -d postgres )
2. Run Initial Setup
3. Run Most Used Letter
4. Have fun!