2021, time for a new general & diff speculation thread...

[alh]

[...]

So you are suggesting that there is some kind of "hash shaping" being done on a large scale that presumably benefits somebody by capturing more blacks during the first 25-30% of the epoch? That's an interesting possibility.

Just as a reminder to folks, while 1% increases are way better than 5% increases, it is still exponential growth, it's just that the exponent is smaller. It works out to a pretty impressive growth rate of 30% per year. If you could get a 30% interest rate on you savings account you would be very happy.

[1715120222]

1715120222

[1715120222]

1715120222

[NotFuzzyWarm]

My take on it is that it is at least partly a result of how the diff setting equation works. Remember, the 1st block after a diff change does not directly count. How much that matters in the Real World - probably just a little - but again, when an Epoch starts the small sample size means inaccurate data and as we near the end of each Epoch the data set becomes larger and more refined.

[wndsnb]

Well ... if I'm doing/applying my poisson calcs correctly, getting 770 blocks or more when the expected # for the actual hashrate is 700 would only happen 1 in 209 times. Doing that 3 times consecutively is not likely unless there's something going on with the hashrate.

10% ahead after the first day is more like a 1 in 8 chance.

[NotFuzzyWarm]

And to throw more fuel into the Bitcoin frenzy - a group from Dubi has announced they are investing 100kBTC or roughly 5 BILLION $ into BTC ventures... https://finance.yahoo.com/news/dubai-ibc-group-pledges-100-175500092.html

[mikeywith]

Yeah, another great diff jump for mining.  2021 seems on making up for 2020 is on its mind.

If we do 1% jumps over and over we will whale.

I wish I had enough time to dig back into my old posts to praise myself,   I have been "bearish" on the difficulty for a while, IIRC i said we will do 3-5% for most epochs to come, I might was wrong with one adjustment, I didn't keep track but overall I am still on track, which makes everyone happy I suppose, including the difficulty bulls.

at blocks 1-700 we are way up like 10% and then we drift back to 1-3%

A huge part of it is what Kano explained, I am not sure if everyone knows that already but even if the hashrate was 100% flat throughout the 2016 blocks, we will always adjust positively, that's because given the meantime for blocks is 600 seconds, 63.2% blocks are subject to be solved in <600 seconds aka; less than 10 mins and only 36.7% of blocks are found in more than 600 seconds.

So when we first start a fresh epoch, even with static hashrate blocks will be mined in less than 10 mins on average, the cryptodiff website we use isn't inaccurate because the dev lacks intelligence as someone tries to make them seem like, but that variance over a small sample results in some inaccurate results due to how statistics are, the website uses simple math with no statistics involved, they could, of course, factor the CDF of the exponential distribution in their website, but that will take a lot of work.

[philipma1957]

Well ... if I'm doing/applying my poisson calcs correctly, getting 770 blocks or more when the expected # for the actual hashrate is 700 would only happen 1 in 209 times. Doing that 3 times consecutively is not likely unless there's something going on with the hashrate.

10% ahead after the first day is more like a 1 in 8 chance.

Well it has been 3 jumps of first 33% of blocks lucky last 66% of blocks not lucky

so I consider it to be a  bit unlikely.

it is over a 42 day period of time.

say

5days lucky
9 days unlucky
Jump
5days lucky
9 days unlucky
jump
5 days lucky
9 days unlucky
jump

if intentional it would be bitmain and viabtc cloud mining doing it.
it started when they announced join venture about Jan 1 2021
think s19pros  with 80th setting firmware and of course 110th normal max
80 is more efficient and it is enough hash to cover all buyers of cloud hashing so when set to 80 bitmain and viabtc would earn a little extra on power savings and cover the contracts.

so set to 110th for 5 days and get 30 extra th per miner
drop to 80th for 9 days and the power saved covers the money spent  extra power on the 5 days ay 110th
if so maybe it is 16eh low 22 eh high so 6 eh for free for 5 of every 14 days.

and you are fully covering your cloud hashing contract. kind of legal (I am not a lawyer)
it comes to close to a constant 2eh for free.

2th is 66 cents a day
2gh is 660 usd a day
2 eh is 660000 usd a day for "free"

I can't say they are doing it but why wouldn't  they do it.

I think its legal since they are selling an s19pro contract and giving that to the cloud hash renter.
I think my math checks out.

BTW if they are not doing  they should be doing it.
as I think it is good business.
I think it is legal and they should give me a cut since if they are not doing it the idea is worth 660,000 x 365 = 240,900,000 a year for them


I will take 1% please send it to

146UJM5kgzLVUV23CXCf33KQKHckoX1gx3

that would be 6600 for each day .

[mikeywith]

Phill, I like the theory but I don't think it's remotely possible, at the current price range it makes 0 sense to turn off 6EH worth of hash power, if Bitmain and Via were sitting on a 6EH they would never turn them off, at the current prices, it doesn't matter if you pay 5 cents or 10 cents, profits are still decent, your theory would make sense if the drop in hashrate was right after a price drop, say Bitmain and Via directly and instantly converting BTC to USD with a good enough margin to the downside, and the bot triggered at x price range to switch from 100TH back to 80TH mode, that would make some sense.

However, the price after the first 5 days is a lot higher, which means if anything, bitmain would be overclocking the heck out of those gears if they had them, bitmain is not under any obligation or law that forces them to run miners at x speed, you pay for 100th you get 100th, how and what they do to get it for you is nobody's concern, so if they have say 20EH worth of power and only 10EH worth of contracts is sold, with BTC at 50k I would bet my house that they would be mining full blast with it.

In other words, the power they save switching from 110th to 80th is nothing compared to the reward, and to make this a lot easier think just 1 S19 and 6 cents per khw.

110TH / 3500w = profit $42.69 and 5.04$ for the power bill = $37.65 daily net profit
80TH /   1600w = profit $30     and  2.3 $ for the pwoer bill = $27$ daily profit = a net loss of $10.6 per miner

Keep in mind that 80th/1600w is based on 20w/th efficiency which isn't even possible.

[wndsnb]

A huge part of it is what Kano explained, I am not sure if everyone knows that already but even if the hashrate was 100% flat throughout the 2016 blocks, we will always adjust positively, that's because given the meantime for blocks is 600 seconds, 63.2% blocks are subject to be solved in <600 seconds aka; less than 10 mins and only 36.7% of blocks are found in more than 600 seconds.

Um, no. If the hashrate was 100% flat, over long enough time scale the average diff would be flat. By definition, the average time to find a block for the given hashrate is 600 seconds.

The Poisson distribution gives you the probability of X events happening in a given time period when you know on average Y events happen. So if on average we expect one block in 600 seconds, you can calculate the chance of getting 0, 1, 2, ..etc for the time period instead.

Although the chance of finding 1 or more blocks in 600 seconds is 63.2%, and the other 36.8% you find 0 blocks, that does not mean that the average expected time is lower. The chance of finding 1 or more is higher because the "or more" contains the chance of 2 through infinity number of blocks.

On average, it still settles to 1 block in 600 seconds.

This is the Poisson distribution that shows the chance of finding a number of blocks in 600 seconds:

blocks    chance
0      0.3678794412
1      0.3678794412
2      0.1839397206
3      0.0613132402
4      0.01532831005
5      0.00306566201
6      5.109436683E-4
7      7.299195261E-5
8      9.123994075E-6
9      1.01377712E-6
10      1.01377712E-7

The upper cumulative distribution gives the chance of finding X or more blocks in 600 seconds

0      1
1      0.6321205588
2      0.2642411177
3      0.08030139707
4      0.01898815688
5      0.003659846827
6      5.941848176E-4
7      8.324114929E-5
8      1.024919667E-5
9      1.125202598E-6
10      1.114254783E-7

And the lower cumulative distribution gives the chance of finding X or fewer

0   0.3678794412
1   0.7357588823
2   0.9196986029
3   0.9810118431
4   0.9963401532
5   0.9994058152
6   0.9999167589
7   0.9999897508
8   0.9999988748
9   0.9999998886
10   0.99999999

The other way to use Poisson is based on rate of success,  if your average expected success rate is 50% for finding one block in a time period (like finding one or more blocks with 50% diff work, or in 1/2 the expected time), the probability is 39.3%.

Finding one or more blocks in 1/4 of the time would be 22.1%, and finding one or more blocks in 2X the time would be 85.4%.

FYI, here are a couple of online Poisson distribution calculators,

https://keisan.casio.com/exec/system/1180573180
https://stattrek.com/online-calculator/poisson.aspx

Excel also has a poisson function, POISSON.DIST()

Back to my earlier post about the chances of being 10% ahead after 700 blocks when the hashrate was actually flat, that is the chance of getting 700 blocks or more when on average we would expect 636 (not 770 vs 700 like I said before), that would be 0.649% (1 in 154 times). So you might expect that to happen once every 6 or so years... doing that 3 times consecutively is 1 in 3.6 million.... so once in 140,000 years... lol.

[philipma1957]

well my math skills are not what they used to be.

but unusual pattern detection I am very good at.

the new jump starts in 28 blocks.

lets see the  first ⅓ vs last ⅔ this time.

And the I said 10% in 700 blocks it was ball park  but we did repeat.

jump
5 days good luck
9 days bad luck as compared to first 5 days
jump
5 days good luck
9 days bad luck as compared to first 5 days
jump
5 days good luck
9 days bad luck as compared to first 5 days

frankly if it is just exception streak lets see it up.
if it is hash rate manipulation by bitmain and viabtc cloud farms it works for me as a miner
lets keep it up.

53k price and 42 days of small jumps = score

so in 2 weeks

61k price and 56 days of small jumps = better score.

I will be watching and hoping for a repeat.

[mikeywith]

Um, no. If the hashrate was 100% flat, over long enough time scale the average diff would be flat.

By long enough I suppose you mean infinity?

I believe we are on the same boat but with a little bit of confusion, let's "for fun" work with the time between blocks which follows the exponential distribution as it's a lot easier to understand than the Poisson distribution which is the number of blocks found in x time follows.

Using the cumulative distribution function we get this nice and simple result.



Notice how almost 40% of blocks come after less than 5 mins from the previous one, we will just assume the epoch just started and we are now at block 10, based on the CDF

Block 0 > =<5 mins  > block 1
Block 1 > =<5 mins > block 2
Block 2 > =<5 mins > block 3
BLock3  >=<5 mins > block 4
Block 5  >=<10 mins > block 6
Block 6  >=<10 mins > block 7
Block 7  >=<15 mins >block 8
Block 8  >=<15mins  > block 9
Block 9  >=<20 mins >block 10

The order does not matter, and this isn't really how it works by the way , but without a doubt finding the first 10 blocks will in theory (based on static hashrate) take less than 100 mins, let's just assume it takes 95 mins, this will automatically set the cryptdiff to a base of +5% which would indicate that we have 5% more hashrate when we are not even close to that.

The deviation from the mean will eventually drop closer to zero as we move further, but when the sample is too small the slightest deviation makes a large impact,

Back to my earlier post about the chances of being 10% ahead after 700 blocks when the hashrate was actually flat, that is the chance of getting 700 blocks or more when on average we would expect 636 (not 770 vs 700 like I said before), that would be 0.649% (1 in 154 times). So you might expect that to happen once every 6 or so years... doing that 3 times consecutively is 1 in 3.6 million.... so once in 140,000 years... lol.

It was indeed not flat or else it can't go to 10%, chances are slim to nothing as you have explained since we are now only 4 blocks away and now the number shows 1.3% we can safely assume that this is going to be accurate, and we can for the sake of argument say this has been it since the beginning of this epoch.

In 20 mins or so we will hit a new diff epoch, let's try to take as many screenshots/copy-paste the data so we can discuss it further when we actually get to 700 blocks.

[wndsnb]

OK, I think I see what you are saying.

If you look at a the chances of being ahead or behind at each block of an epoch, you are more likely to be equal to or ahead than behind through the whole epoch, but getting closer and closer to 50/50 as you get to the end. But if you look at just the chance of just being ahead it changes the picture drastically.



Well, the data is available, so maybe if I get ambitious I'll look at the blocks from the last couple epochs to see what we might be able infer. All you need are the timestamps of the blocks to re-create the % ahead/behind.

[alh]

For what it's worth: 74/74.3 blocks in/expected. BTC price: $55440.

[mikeywith]

If you look at a the chances of being ahead or behind at each block of an epoch, you are more likely to be equal to or ahead than behind through the whole epoch, but getting closer and closer to 50/50 as you get to the end.

Aha, exactly what I mean, of course by 700 blocks we get close to 1/3 of the epoch lifespan which would get as closer to the mean, unlike when we view the results at say block no.100 or so.

Well, the data is available, so maybe if I get ambitious I'll look at the blocks from the last couple epochs to see what we might be able infer. All you need are the timestamps of the blocks to re-create the % ahead/behind.

The timestamp is inaccurate, you will see blocks mined in the past or way ahead into the future, the most accurate would be the actual time a block was submitted, not sure if any of the explorers out there keep track of that, but if you want just the timestamps LoyceV has it here in a single file.

[alh]

If there is some kind of deliberate effort to adjust hash rate, those folks don't know the epoch has switched. Right now we are at 156/156.28 blocks. BTC price is $56081.

Of course it's possible they are just messing with Phil. 

[mikeywith]

Imagine someone would turn off 6EH just to mess with someone.  

I wanted to take screenshots of 250, 500, 750, and 1000 blocks but I missed 250 and I got the 285.

Latest Block:   671612  (8 minutes ago)
Current Pace:   101.6779%  (285 / 280.30 expected, 4.7 ahead)
Previous Difficulty:   21434395961348.92
Current Difficulty:   21724134900047.27
Next Difficulty:   between 21827669607849 and 22123513877023
Next Difficulty Change:   between +0.4766% and +1.8384%
Previous Retarget:   last Friday at 11:56 PM  (+1.3517%)
Next Retarget (earliest):   March 5, 2021 at 6:23 PM  (in 11d 19h 44m 20s)
Next Retarget (latest):   March 5, 2021 at 10:28 PM  (in 11d 23h 49m 37s)
Projected Epoch Length:   between 13d 18h 27m 18s and 13d 22h 32m 35s

I also got the 5 blocks status earlier on.

Latest Block:   671332  (3 minutes ago)
Current Pace:   83.0581%  (5 / 6.02 expected, 1.02 behind)
Previous Difficulty:   21434395961348.92
Current Difficulty:   21724134900047.27
Next Difficulty:   between 19472305559050 and 21711695987187
Next Difficulty Change:   between -10.3656% and -0.0573%
Previous Retarget:   Yesterday at 11:56 PM  (+1.3517%)
Next Retarget (earliest):   March 6, 2021 at 12:16 AM  (in 13d 23h 20m 10s)
Next Retarget (latest):   March 8, 2021 at 8:28 PM  (in 16d 19h 31m 57s)
Projected Epoch Length:   between 14d 0h 20m 22s and 16d 20h 32m 9s

[alh]

It could very well be that someone(s) with a lot of hashrate have decided that at the present time it might be better to just plug away and let BTC price increases do the work for profit. That 6EH would come at a cost, so maybe you would "turn some off" and reduce their cost for next couple of weeks. After all, a difficulty increase hurts everyone, including the Multi-EH folks as well as the TH folks like me.

[wndsnb]

Here are some plots for the last 3 epochs I made using the data from LoyceV's file. The green and orange are the probability of more than and less than the number of blocks at that point in the epoch assuming the actual hashrate for the whole epoch was flat. So when they diverge a lot, it suggests a better chance that the hashrate actually changed and is not just the error you'd assume.




And breaking out my old profitability table ... looks a lot different than the last time I posted this when S9s were "dead"... still dead at my 21c/kWh I pay at home though. But at least I can test out the S17s I'm repairing without losing $.

If someone was shutting down 6EH of S17s at 5c power, they'd only be giving up $69,700 per hour.... lol.

[alh]

Clearly nothing "odd" during the 1st 700 blocks so far: 338/344 expected, so a small decrease so far. BTC price: $55830.

Just another small update on the "early 700 block hash rate" discussion. At this time, we are behind just a bit.Either those folks are asleep at the switch, or maybe it's not going to materialize this epoch.

422/431 blocks expected (behind by 9).   BTC price has also slid some at $54,554.

Another update on the current epoch (aka the 700 block hash discussion): We are currently at 460/483 expected. This is currently projected to end with a decrease of -3.6% to -1.5% (plenty of time for that to change).

Unfortunately BTC price continues to slide, currently at $49,599. BTC has been above $50K for the last week.

[philipma1957]

did not happen.  Since I spoke that cloud mining likely supplies the hash it is interesting to note that it did not happen the exact same jump that these emails advertising a new bitmain based cloud came.

https://www.bitfufu.com/subject/newbie#utm_source=ic&utm_medium=bmemail&utm_campaign=bmcorp

So first three jumps matched the bitmain+ viabtc hookup.

and as Mikeywith said why not just use the hash we know have a way to use the hash bitfufu.

So 4 jumps correlate to cloud mining being used to manipulate hash rate.

All spec zero evidence

[mikeywith]

Here are some plots for the last 3 epochs I made using the data from LoyceV's file...

Nice analysis, I wish the source of the data was more accurate, I believe Kano and/or CK would have such data since they both have mining nodes running, I suppose they keep a record of when they received every block regardless of the timestamp on the block itself.

If someone was shutting down 6EH of S17s at 5c power, they'd only be giving up $69,700 per hour.... lol.

What's 70k between fellow miners?  


Ok price dropped a nice 23%, many things have changed from the last 2 epochs, hard to assume why didn't phill theory apply this time, anyway, I am happy to see us cool down and with some potential of a drop in difficulty which is massive.

Latest Block:   671863  (11 minutes ago)
Current Pace:   94.0643%  (536 / 569.82 expected, 33.82 behind)
Previous Difficulty:   21434395961348.92
Current Difficulty:   21724134900047.27
Next Difficulty:   between 20450514442221 and 21117763283211
Next Difficulty Change:   between -5.8627% and -2.7912%
Previous Retarget:   last Friday at 11:56 PM  (+1.3517%)
Next Retarget (earliest):   March 6, 2021 at 9:42 AM  (in 10d 10h 48m 18s)
Next Retarget (latest):   March 6, 2021 at 9:08 PM  (in 10d 22h 13m 54s)
Projected Epoch Length:   between 14d 9h 46m 31s and 14d 21h 12m 8s

Going with the assumption that this is accurate, 5% of the total hashrate from the last epoch is 8EH, since this can't be the 19 series since those are still profitable even with high rates of power, let's assume those are old S9s and T17s, so maybe in the range of 70w/t per miner on average, so anywhere between 25-30th per miner in respect to w/th, this means at least 266,666  (quarter of a million) are not profitable below 50k.

The above is far from accurate obviously, but it's an indication that the majority of the hashrate does not come from the new generation miners, or else, 20% in price shouldn't shake anybody, but to shake out those many gears, well. maybe this explains why I failed to source any S9i for one of my clients the past couple of days in China.