Network hash just spiked about 50%

[cpt_howdy]

Wow. Just after we come to the end of another 2016 block period, the hash rate ramps up massively to ~15Thash/s.

This says to me that (a) Miners were deliberately keeping their rigs in 2nd-gear until the difficulty rolled over (stayed constant at ~1380000) (b) They did this to maximise their profits (in BTC at least) during the next 2 weeks.

Why would they do this? Well, probably because they're predicting a big price crash, and are going to power down those rigs at the end of this 2 week period. And why not? If the hash rate stays like this for 2 weeks, come the end of this 2016 blocks, the price/difficulty ratio is going to be so low it will be unprofitable to mine for all but the most efficient. Wheat from the chaff...

[RandyFolds]

No, the global hashrate is extrapolated from the rate that blocks are found. A lucky streak will make it look like the hashrate is higher than it actually is.

Beyond that, the sample period is too small immediately following a difficulty change so the numbers are always way off for a few days.

[the joint]

No, the global hashrate is extrapolated from the rate that blocks are found. A lucky streak will make it look like the hashrate is higher than it actually is.

Beyond that, the sample period is too small immediately following a difficulty change so the numbers are always way off for a few days.

Looks to me that the "luck" we're seeing is more than simple variance.

[Kluge]

If miners really quit mining after price drops (which I still don't think is nearly as strong a correlation as many've said), then I'd assume GPUMax's extremely generous PPS rate has attracted the additional miners.

[DeepBit]

Deepbit's luck for last 24h is 20% better than expected, so I can believe that this spike is caused by luck.

[cpt_howdy]

No, the global hashrate is extrapolated from the rate that blocks are found. A lucky streak will make it look like the hashrate is higher than it actually is.

Beyond that, the sample period is too small immediately following a difficulty change so the numbers are always way off for a few days.

I agree with you first point - it is true that it's block discovery rate, not hashrate, that is directly measured, and that it's possible we're just seeing a lucky period. However, your second point doesn't make any sense: what do you mean by the sample period? The block discovery rate is calculated by adding up N blocks over the past T minutes/hours, and then calculating N/T. The calculation of block discovery rate should be completely independent of both the difficulty, and where we are within a given 2016 block period. So, why should we inevitably see more variation in hash rate at the beginning of a 2016 block period? What you're suggesting is that T (the 'period' you mentioned) automatically resets to zero at the end of a 2016 block period. Well, I was taking the rate value from bitcoincharts.com, so if that's the way they do it, then yes I would expect a large over or under-estimate *just after* a change of difficulty. For a few days though? That makes no sense at all - think about it.

[RandyFolds]

No, the global hashrate is extrapolated from the rate that blocks are found. A lucky streak will make it look like the hashrate is higher than it actually is.

Beyond that, the sample period is too small immediately following a difficulty change so the numbers are always way off for a few days.

I agree with you first point - it is true that it's block discovery rate, not hashrate, that is directly measured, and that it's possible we're just seeing a lucky period. However, your second point doesn't make any sense: what do you mean by the sample period? The block discovery rate is calculated by adding up N blocks over the past T minutes/hours, and then calculating N/T. The calculation of block discovery rate should be completely independent of both the difficulty, and where we are within a given 2016 block period. So, why should we inevitably see more variation in hash rate at the beginning of a 2016 block period? What you're suggesting is that T (the 'period' you mentioned) automatically resets to zero at the end of a 2016 block period. Well, I was taking the rate value from bitcoincharts.com, so if that's the way they do it, then yes I would expect a large over or under-estimate *just after* a change of difficulty. For a few days though? That makes no sense at all - think about it.

That is the way they do it, and a few days is a small sample period compared to two weeks. Probability, homie...shit's all over the place.

Try out the search function. This has been beaten to death since this forum's inception.

[cpt_howdy]

Right, well I think we're talking at cross-purposes here. I was seeing a big change in the block discovery rate, and attributing it to a real change in network hash; the truth is that the block discovery rate is dictated by the combination of total *real* network hash, and luck, which randomly varies from block to block.
Forget that issue, In the last post I was just trying to point out that this estimation process does not change when the difficulty changes - check out the graph below.



See the changes in difficulty? No indication that they provoke wild changes in the the estimated hash rate.

[The-Real-Link]

That's quite a spike.  I'm going to simply guess a lot of people powered on rigs since the difficulty did not spike so they know they can have two more weeks of "consistent" block generation / earnings.

[RandyFolds]

That's quite a spike.  I'm going to simply guess a lot of people powered on rigs since the difficulty did not spike so they know they can have two more weeks of "consistent" block generation / earnings.

I am failing to see a statistically significant jump in that graph...certainly not 50%.

Why would they have powered off and then powered on with very little change in the situation, save for a drop in the price of BTC, and thus a drop in their profitability?

[The-Real-Link]

That's quite a spike.  I'm going to simply guess a lot of people powered on rigs since the difficulty did not spike so they know they can have two more weeks of "consistent" block generation / earnings.

I am failing to see a statistically significant jump in that graph...certainly not 50%.

Why would they have powered off and then powered on with very little change in the situation, save for a drop in the price of BTC, and thus a drop in their profitability?

Perhaps they thought the difficulty would spike upwards and decrease their profits even further?  Who knows?  While there can be good estimation on all of this, it's seemingly as much speculation as it is fact sometimes.

[Littleshop]

These threads appeared every difficulty change for a while but we went a bit without them. 

The rate did not rise dramatically. It is just the formula used to display the hash rate. 

[BinaryMage]

And voila, back to normal. Worry less, think more. 

[joulesbeef]

And voila, back to normal. Worry less, think more. 

That's what THEY would like you to think.

[DeathAndTaxes]

These threads appeared every difficulty change for a while but we went a bit without them. 

The rate did not rise dramatically. It is just the formula used to display the hash rate. 

Yeah they make a good difficulty change alarm.  I knew what the thread, proposed scenario, and real outcome was before clicking.  Good news is I learned we just had a difficulty adjustment and it was flat so wasn't a wasted thread.

[DeathAndTaxes]

See the changes in difficulty? No indication that they provoke wild changes in the the estimated hash rate.

See a 15TH data point there?

My guess is you got your "15TH/s" from somewhere else.  Somewhere that to avoid complex calculations simply compared # of blocks and timespan since last difficulty reset.  Why since last difficulty reset?  It avoids having to do more complex calculations involving multiple difficulties.  Those sites tend to be insanely in accurate in the first 30 or so blocks after difficulty change.

Care to point out where you saw "15TH/s"?

[RandyFolds]

See the changes in difficulty? No indication that they provoke wild changes in the the estimated hash rate.

See a 15TH data point there?

My guess is you got your "15TH/s" from somewhere else.  Somewhere that to avoid complex calculations simply looks at blocks and time since last difficulty reset (it avoids needing to adjust for multiple difficulty values).  Those sites tend to be insanely in accurate in the first 30 or so blocks after difficulty change.

Care to point out where you say 15TH/s?

Doesn't it feel good to be right? I rarely have the luxury, but when I do...oh boy...

[eleuthria]

Deepbit's luck for last 24h is 20% better than expected, so I can believe that this spike is caused by luck.

BTC Guild also had extremely good luck on the 18th (and some of the 19th).  Of course then we had a 6 hour block and 10 hour block yesterday/today.  Luck can have a dramatic influence on the rate of blocks in a short window (and 24 hours is DEFINITELY short).