Quote from: kTimesG on Today at 05:17:38 PM

Quote from: newsecurity1986 on Today at 04:52:56 PM

Quote from: kTimesG on Today at 03:18:46 PM

Congrats on discovering and demonstrating that there's a 63% chance of success to hit X at least once, when p=1/N, in N tries, btw. No one really bothered to do the math on that one. You're definitely the first.

So then, if you're saying the prefix method is mathematically sound, right? Why are you giving it shit?

N balls in an urn. one is blue, the rest are red.
Extract one ball, if it's blue = success. If it's red: failure.
Put the ball back in the urn.

This gives you an exact 1:1 analogy with either searching for an H160 full hash, or whatever prefix size of it you desire. You can use any other classification criteria as you see fit, the analogy doesn't change.

p = 1/n
trials = n

Probability of failure per extraction: (n-1)/n = 1 - 1/n

Total probability of failure over N extractions: (1 - 1/n) ** n

Probability of success to get the blue ball at least once over N extractions:

p_success = 1 - probability_failure = 1 - (1 - 1/n)**n

This function converges towards 1 - 1/e = 63.212...% as N goes to infinity. It's a very exact number to calculate.

Now: what in the freaking world changes after you put the extracted ball back in the urn?

Answer: NOTHING. That's why the prefix theory is total bullshit, and is one of the reasons this thread is junk science, which I'm retreating from. Feel free to go and pat mcdouglasx, he needs a hug.

Listen mate, your fancy urn model with ball replacement shares that same 63% probability on paper, yeah, we all know 1 - 1/e ≈ 0.6321. But quit pretending it's the real deal for crypto hunting. Your abstract math misses three massive practical points:

Keys ain't balls:
Crypto keys are unique and never reused, no "putting back" like your urn nonsense.

Real-world:
Prefix search actually saves work by bailing early when collisions happen. Your urn keeps uselessly drawing balls forever.

Physical constraints matter:
The prefix search exploit actual hardware behavior (sequential scanning + early termination) to boost efficiency. Your model just jerks off in theory land.

The results prove it works in practice 63% success rates don't lie. You're dodging the actual method to wank over abstract models that physically don't function like real crypto searches. Just 'cause the numbers match doesn't make your urn relevant. Stop confusing textbook toys with engine-level optimization.