Show me definable edge one can systematically test and prove. Otherwise there's no reason to believe that round numbers (30% retracement, 40% extension, etc.) work just as well...and it's just an attempt to buy low and sell high...which sometimes works and sometimes doesn't.
I tend to agree, although visually it works amazing sometimes.
The thing though is that you have to take the most common ones, the 23.6, 38.2, 50, 61.8 and 78.6, and now apply some sort of buffer, lets say 3-4% on either side since you can't always expect it to hit before a turn, so you have to front run it, and you have to give it some room if it pokes, so an early entry needs a stop to allow for that poke. So now you have 5 zones in which to look for trades, with a decent stop, and you have to be willing to take the trade at each of these fibs because how do you know which one will cause the turn? Sure you can look for confluence of other signals/indicators to cherry pick which fibs you take, but then, you're going to be missing some great entries while losing money on others since no filter will dramatically increase the win rate.
With so many fibs spread out from from 0 to 100% of the move, and all the stops you have to allow for, how do you mathematically ensure you have an edge? For something to work consistently, I would like to see a better than 50% win rate, and its not like you will get anywhere close to 50% win rate at each fib level. I think the best traders can see a move setting up, and it helps when it "happens" to also occur at a fib level, but its hardly the fib level that is the biggest factor in turning price.