Difficulty speculation for the round started on 29/6/13

[Mabsark]

I'll track the estimated difficulty increase over 126 block periods. The increase shown is the increase in difficulty from the last round, not an increase between blocks. I missed the first 2 periods but I'll try not to miss any more. Just in case though, feel free to post the right value if I haven't updated the post and nobody else has posted the value.

From the initial values, 3.16% increase 1/8 into the round, it's looking to be a 24% increase for this round. I suspect that the rate of increase will itself be increasing though, so my initial estimate for this round is about a 30% increase.

1,720 blocks remaining (just over 1/8 into the round) = 3.16 % increase
1,638 (3/16) = 4.72%
1,512 (4/16) = 5.12%
1,386 (5/16) = 6.76%
1,260 (6/16) = 7.4%
1,134 (7/16) = 7.95%
1,008 (8/16) = 8.94%
882 881 (9/16) = 9.99%
756 (10/16) = 12.95%
630 629 (11/16) = 15.48%
504 503 (12/16) = 15.52%
378 (13/16) = 18.33%
252 (14/16) = 19.98%
126 (15/16) = 20.21%
0 (new round) = 22.63%

So, we're half way through now and here's a quick summary. The difficulty round was seeing impressive increases between periods of about 1.5%. That fell to 0.4% and has steadily increase backup to 0.99%. Based on the initial results, the difficulty increase for the next round should be at least 18% based on doubling what it's at now. Taking into account the accelerating rate of increase, it should be a bit higher, so say at least 20%.

3/4 into the round and we're at 15.5%. We've seen some big 3% gains in a couple of periods, yet over the 11th, the increase was only 0.04%, the smallest so far. Was is just an extremely unlucky period or did it correspond to the lack of shipping on previous weekends?

So, at the end of the round, my initial estimate of 24% based on the first 252 blocks turned out to be pretty close. The difficulty increases varied wildly, from 0.04% to 2.96%.

[ujka]

For a 30% increase in diff, network needs a minimum of 50T of new hashing power. Not very likely in next 10 days.
Anyway, nice effort on your part. How are you calculating the data? Is there some webpage with calculations, or are you pulling data from blockchain?

[Mabsark]

For a 30% increase in diff, network need a minimum of 50T of new hashing power. Not very likely in next 10 days.
Anyway, nice effort on your part. How are you calculating the data? Is there some webpage with calculations, or are you pulling data from blockchain?

10 Minirigs per day = 50 Th/s in 10 days. There's Avalon, ASICMiner, and BFL units coming online. Between the 3 of them, 50 Th/s over the next 10 days is certainly a possibility.

As for the data I'm using, it's just taken from http://bitcoincharts.com/

Also, my estimate will be based off the data I post here. That initial 30% guess is based on observations of the data over the past few days. We'll be able to see how it's progressing over the next few data points though.

[ujka]

As for the data I'm using, it's just taken from http://bitcoincharts.com/

Do you know what metod is bitcoincharts.com using to estimate next difficulty? Is it sliding window of 2016 blocks?

I'm watching allchains.info.  Author says he is using the exact calcluation since last diff change. Right now there is 1652 blocks left in this round (18.06% blocks measured), with 9.43% increase. Bitcoincharts: 4.38%.

[Mabsark]

As for the data I'm using, it's just taken from http://bitcoincharts.com/

Do you know what metod is bitcoincharts.com using to estimate next difficulty? Is it sliding window of 2016 blocks?

I'm watching allchains.info.  Author says he is using the exact calcluation since last diff change. Right now there is 1652 blocks left in this round (18.06% blocks measured), with 9.43% increase. Bitcoincharts: 4.38%.

I'm not sure what method it's using.

[PaperClip]

As for the data I'm using, it's just taken from http://bitcoincharts.com/

Do you know what metod is bitcoincharts.com using to estimate next difficulty? Is it sliding window of 2016 blocks?

I'm watching allchains.info.  Author says he is using the exact calcluation since last diff change. Right now there is 1652 blocks left in this round (18.06% blocks measured), with 9.43% increase. Bitcoincharts: 4.38%.

allchains.info use linear extrapolation method, you can easily do all these calculations by yourself:

1) go to http://blockexplorer.com/q/nethash/2016, take latest block number (currently: 243936). This is first block after difficulty change.
A = 243936

2) go to http://blockexplorer.com/b/243936 (or insert new block number from step 1 in case of difficulty change), click on "Raw block" link, copy value from field "time" (in this case: 1372515725). This
B = 1372515725

3) now take block number from step 1, add to it any value between 1 and 2015. In this case it will be 126, or 252, or 378, etc..
C = 243936 + 126 = 244062

4) go to http://blockexplorer.com/b/244062, click on link "Raw block", copy value from field "time" (1372584672)
D = 1372584672

Result formula:
(60*60*24*14)/((D-B)/(C-A)*2016)-1
Result is 0.096494409, that is ~9.65%


I don't know about http://bitcoincharts.com/ algorithm, unfortunately it is impossible for me to find missing values there. Anyway, with allchains approach difficulty increase estimation every 126 block is following:
126 (1/16): 9.65%
252 (2/16): 9.98%
378 (3/16): 10.04%
504 (4/16): 9.65%

[ujka]

Thanks. I think I understand:
A is first block in round (first block on diff change)
B is time of that block
C is block 'offset' into round
D is time of that block

Network is targeting to generate a block every 10 minutes (600 seconds), and is calculating new difficulty for that target every 2016 blocks. That gives 14 days (or 60*60*24*14 in seconds - 'target time').

(C-A) is number of blocks found in current round
(D-B) is time elapsed in current round (from first block A up to block C)
(D-B)/(C-A) is 'average time per block' in current round

Estimated new diff change we can get with:
(target time) / (average time per block * 2016) - 1

[PaperClip]

Yes, correct.
Actually, equation for "Estimated new diff change" is larger, and involves difficulty. But it can be simplified to that simple form.

(D-B)/(C-A)*2016 is time (in seconds) that will be needed for full round, according to current average speed.
60*60*24*14 is ideal time for full round in case if network hash speed does not change. Basically, average hash speed from last round would take this much seconds on current difficulty for this round.

DIF - current difficulty
DIF_NEXT - next difficulty

next difficulty can be calculated from this:
DIF * (60*60*24*14) = DIF_NEXT * ((D-B)/(C-A)*2016)
or
DIF_NEXT = DIF * (60*60*24*14) / ((D-B)/(C-A)*2016)

Increase = (DIF_NEXT - DIF) / DIF
Increase = (DIF * (60*60*24*14) / ((D-B)/(C-A)*2016) - DIF) / DIF

if you simplify last formula then you will get
(target time) / (average time per block * 2016) - 1

[ujka]

Some crazy hashing today - 185 blocks generated! That's lots of luck, or 200,000TH/s!

[ujka]

1,260 (6/16) = 7.4%

[Mabsark]

1,260 (6/16) = 7.4%

Cheers. I woke up 20 blocks to late for that one.

[Mabsark]

Half way through the round now. There were major increases in difficulty at the beginning of the round. Do you think that was mostly due to hashing power or luck?

[ujka]

Half way through the round now. There were major increases in difficulty at the beginning of the round. Do you think that was mostly due to hashing power or luck?

Last difficulty increase didn't really 'catch up' to the hashrate (http://bitcoin.sipa.be/speed-lin-10k.png). That's why calculations started high, and there is also a major increase in hashing power last few days. Mostly 'unknown' portion - http://blockorigin.pfoe.be/top.php has NO DATA for 10.5% of blocks found. Was 7% last week.

[Mabsark]

Over the 9th period, the difficulty increased by about 3%. If the following periods were all at least 3% then the difficulty at the end of the round would be at least 30%. If difficulty starts to increase by more than 3% per period then the difficulty increase could be as high as 40% at the end of the round. The next period of two should shed some insight on whether to expect 20% or 40%.

[HackneyB]

May I ask why you're doing this? 

There's a math puzzle factor to this and maybe a flip a coin aspect too.  But over the last 3-4 months difficulty changes have been sort of unpredictable - there was even a decrease if I remember correctly. 

All due to a looong list of factors that affect the total network hashrate, including the less obvious ones like homemade rig uptime lapses, to FPGAs being sent back to BFL, people going on holidays and just not caring, to individual miners switching to other pools, to solo, to other cryptocurrencies ....

Don't get me wrong I'm tracking things too but why such a real time approach?

[Mabsark]

May I ask why you're doing this? 

There's a math puzzle factor to this and maybe a flip a coin aspect too.  But over the last 3-4 months difficulty changes have been sort of unpredictable - there was even a decrease if I remember correctly. 

All due to a looong list of factors that affect the total network hashrate, including the less obvious ones like homemade rig uptime lapses, to FPGAs being sent back to BFL, people going on holidays and just not caring, to individual miners switching to other pools, to solo, to other cryptocurrencies ....

Don't get me wrong I'm tracking things too but why such a real time approach?

No reason really. Just curious to see how its progressing.

[ujka]

503: 24647153 / 21335329 = 1.1552        15.52%

[Mabsark]

503: 24647153 / 21335329 = 1.1552        15.52%

Thanks again.

[ujka]

Linear extrapolation of elapsed time (seconds) per each step:


21.6% extrapolated increase.
24.35% 

[Nemo1024]

Next difficulty is predicted to be 25668640 in 2.2 days. Buckle up.