Independent Leyton Orient Forum

Mick McQuaid wrote: ↑Fri Jul 22, 2022 11:24 am https://freshman.dev/wordle/#/leaderboard

I spent a diverting half an hour or so looking into how people are coding wordle bots. Lots have got down to an average of 3.41, I haven't been able to find/understand whether someone has actually proved that this is the optimal strategy but I would think its close.

I spent another 10 minutes or so going through tuffers last 100 scores posted on here. Amazingly, despite occasionally using what us mere mortals would consider a slightly less than optimal strategy of using starting words with repeated letters he has managed an average of 3.23.

Mick McQuaid wrote: ↑Tue Jul 26, 2022 12:10 pm Your 4 is way more suspicious than BoniO's 3 there actually tuffers. BoniO excluded 9 letters before getting a result, you excluded 3.

Mick McQuaid wrote: ↑Tue Jul 26, 2022 12:10 pm Your 4 is way more suspicious than BoniO's 3 there actually tuffers. BoniO excluded 9 letters before getting a result, you excluded 3.

Mick McQuaid wrote: ↑Wed Jul 27, 2022 9:22 am
Up until there were a few suggestions that tuffers was posting suspiciously good results he was running at 88% of games solved in 3 tries or less, after which his results have normalised a little for some reason.

Mick McQuaid wrote: ↑Wed Jul 27, 2022 9:22 am I'll make this my last bit of maths on the subject.

The best bots solve in 3 goes or less 56% of the time. Up until there were a few suggestions that tuffers was posting suspiciously good results he was running at 88% of games solved in 3 tries or less, after which his results have normalised a little for some reason.

Putting the results of the first 63 posts into a binomial probability calculator (I had to try 3 before I found one which calculates to enough decimal places) gives a probability of this happening, when using a mathematically perfect strategy, of 0.00000001649151% or about 10 times less likely than winning the lottery.

By the way, I cheated today after 4 goes by using a crossword solver after staring at it for about 5 minutes and not being able to think of any word.

Mick McQuaid wrote: ↑Wed Jul 27, 2022 9:22 am I'll make this my last bit of maths on the subject.

The best bots solve in 3 goes or less 56% of the time. Up until there were a few suggestions that tuffers was posting suspiciously good results he was running at 88% of games solved in 3 tries or less, after which his results have normalised a little for some reason.

Putting the results of the first 63 posts into a binomial probability calculator (I had to try 3 before I found one which calculates to enough decimal places) gives a probability of this happening, when using a mathematically perfect strategy, of 0.00000001649151% or about 10 times less likely than winning the lottery.

By the way, I cheated today after 4 goes by using a crossword solver after staring at it for about 5 minutes and not being able to think of any word.