How does a pool correlate your work to vardiff?

[bitpop]

If I'm submitting vardiff 2048, hows the pool know about all my work in between? Is it guessing? To my understanding work between is tossed, not added up to a 2048.
Is there any penalty to using very high vardiff? I do know that in p2pool you may miss out on shares because of new work being so quick.

[1715051989]

1715051989

[1715051989]

1715051989

[1715051989]

1715051989

[-ck]

It ignores all your work below that diff. By definition it is 2048 times less likely you will find diff 2048 shares so the pool knows how much work you've been doing only indirectly. The only disadvantage to higher diffs is more variance but the significance of this is grossly exaggerated and for the most part it is of no detriment at all and only benefit.

[bitpop]

So shares are guessed though at some level?

[-ck]

There's no guessing. If you submit a share it counts as 2048 shares. If you submit none, it counts none. There is no inbetween.

[bitpop]

Ok thanks.

[eleuthria]

Just to clarify something when it comes to difficulty.

The work you're doing is the same whether the pool has your difficulty at 1 or 1,000,000.


Difficulty 1 mining didn't return all of your results either, only about 1 in 2^32 hashes produced a difficulty 1 share (4.2 billion).  However, your share wasn't "exactly" difficulty 1 (normally).  About 50% of those diff1 shares would also qualify as diff2 shares.  About 50% of those (25% of all diff1) would also qualify as diff4 shares.  Repeat infinitely.


When the pools adjust your difficulty (or you do via a web interface), what is being agreed upon is how high of a difficulty your minimum is.  You're not credited at all for anything worse than that difficulty.  In return, you're credited for more work with each submission.  So if you're only submitting diff2048+ shares, each submission is counted as 2048 shares, because the odds of a diff1 share also qualifying as diff2048 is 1 in 2048.

In the short term, you may have done much more or less work than what it would take to submit 2048 difficulty 1 shares.  Just like pools have luck, each hash is luck based.  But over time, the distribution of your hash results will match what you've been credited.  The proof of this is the fact that the pool interface should show an estimate of your hash rate that is very close to your actual hash rate (unless you have your difficulty set too high, in which case it my deviate significantly from variance).