How Does BitcoinCharts.com Know The Network Total Hashes/Second?

[gigabytecoin]

As seen in the top all black nav bar of http://bitcoincharts.com/markets/

How do they determine the number of terahashes/second, currently at 0.735 Thash/s

Isn't a hash only submit to the network once it has been deemed to be below the target value?

Is the network hash rate just an estimate?

[1715483717]

1715483717

[1715483717]

1715483717

[1715483717]

1715483717

[Grinder]

I think it's guessing using a combination of the known current difficulty and how quickly blocks are currently being solved.

[Pieter Wuille]

Is the network hash rate just an estimate?

Yes, and a very tough one to make.

[gigabytecoin]

Is the network hash rate just an estimate?

Yes, and a very tough one to make.

Ok thank god. I thought I was going nuts there for a second.

I'm a man of logic. And the speed with which that number sways every few hours doesn't make sense given the nature of mining.

[Jim Hyslop]

Is the network hash rate just an estimate?

Yes, and a very tough one to make.

Actually, it's not that hard.

The wiki states somewhere (can't find the page now, of course) that the probability of a single hash calculation solving a block is

      target
     -------
       2²⁵⁶

and the average number of hashes required to solve a block is its inverse:

       2²⁵⁶
     -------
     target

Let's call that last formula A. On average, then, it takes A hashes to solve a block. We know how long it takes to create a particular block (just take its creation timestamp and subtract the timestamp of the previous block from it). So the network hash rate is simply A / timespan.

Of course, individual blocks will be highly variable due to the random nature of finding a solution (there were about 3 or 4 blocks recently that were created within 5 minutes; the block before that series took 20 minutes, as did the block after). To get a more accurate estimate, you'd want to take the time over 50 or 100 blocks. The formula is:

      2²⁵⁶ * n
-----------------------
target * (ts_N - ts_N-n)

where
n is the number of blocks you're sampling
ts_x is the timestamp of block number x
N is the latest block number (or whatever block number you want to end your sample range with)
and target is currently 0xCBBD000000000000000000000000000000000000000000000000