Topic: 2024 Diff thread happy New Years. - page 19. (Read 7261 times)

To me it seems that a reduction of -2% more normal, than 5 or 6, after such a big rise in the previous cycle.

But, the sample is still small (for the moment we are analyzing). We need to see how another cycle goes in order to draw better conclusions.

If in a third cycle we have a stagnation of variation, it is a sign that we have had a real increase in power of about 5%. But if the variation is higher, then the increase was greater. If it's lower, then it's weird that +8% spike.

2% variance in a whole epoch has a chance of 18.42%, mind that given the K (the shape parameter) is constant the variance in either direction would yield the same result, although in my previous code the code would get the wrong result if you enter -2 instead of 2 in the percent parameter given this line


removing the 1 - and changing the line to


Would work fine for negative % but obviously not for positive, I could add a simple function to check the value and format the result accordingly, but I don't think it's needed.




Using your logic 2% is 1 in 5.426, so two consecutive events would be 1 in 30.72, that is 3.25% in percentage terms.

But then, this is based on the assumption that in the second epoch (the current one for our case) is indeed affected by 2% or more variance, but if we assume that the previous epoch had a 2% luck, then his epoch would end up with -2% change while having 0% variance because the expect variance is 0%.

if we do indeed end up with a -4% which is 2% variance then that's 3.25%, however, we have no way of knowing what the heck is going on, 2016 blocks make a very small sample, if Satoshi made the difficult adjustment twice a year, we would have enough blocks to guess with.

With the size of 6 months' worth of blocks, the probability would be

1>% = 5.2%
2>% = 0.06%
3>% =0.00007% (almost never)

VS 2 weeks' worth of blocks

1>% = 32.4%
2>% = 18.4%
3>% = 9%

With the 2 weeks epochs, you can't be too sure that this epoch isn't having a 3% variance since 9% is relatively a huge figure, if we have diff adjustments only twice a year, then you could almost always assume that variance is never above 2%, in fact anything above 1.5% drops to below 0.7% so ya, assuming no epoch has a 1.5% variance would be a safe thing to do and you would really tell how much hashpower was added or left, but with two weeks worth of blocks even 5% variance has a good 1.3% chance of happening, so there is nothing you can do to safely eliminate a 5% variance, let alone 2% that happens nearly 20% of the time.

It's always a mix of variance. down clocking , overclocking , upgrades. Power shortages. internet outages. bankrupt companies, natural disasters.  Lastly government interference.   To do a +8 and now a -5 is a very big swing.



I forgot to mention.

cents per th. we were at 13 cents a th in December we are now at 8 cents a th

some may shut down due to that.


back to luck if you want to argue let jump was 3% extra luck

and this jump is 3% less luck.

you get .09 x .09 = 0.0081 or 1 in 120 chance.

it is possible as the time period is 4 weeks x 120 = every 480 weeks which is 1 Time in 9+ years.


more likely variance was +2% and then -2%

Mike plug in 2%  and that is more like a 1 in 4 shot. then back to back 1 in 16 shot.  which is around 1 time every 64 weeks


you did do a 4% which is 1 in 25 cycles so back to back it is 1 in 625 cycles

So I lean to back to back 2% 1 of 16 shot. (mike can plug this in for exact number)

not back to back 3% variance 1 of 120 shot
not back to back 4% variance 1 of 625 shot


it is simply way more likely that variance was 2% up then 2% down

and cent per th dropped from 13 cents in dec to 8 cents now

and retooling s19 to s21

and some under clock


they are coming back strong !

Latest Block:   832042  (5 minutes ago)

Current Pace:   97.6502%  (1451 / 1485.92 expected, 34.92 behind)

Previous Difficulty:   75502165623893.72                           
Current Difficulty:   81725299822043.22                           
Next Difficulty:   between 79992421406053 and 80125865742021
Next Difficulty Change:   between -2.1204% and -1.9571%
Previous Retarget:   February 15, 2024 at 9:07 AM  (+8.2423%)
Next Retarget (earliest):   Thursday at 4:34 PM  (in 3d 23h 47m 51s)
Next Retarget (latest):   Thursday at 5:12 PM  (in 4d 0h 25m 57s)
Projected Epoch Length:   between 14d 7h 27m 1s and 14d 8h 5m 7s

3% or more luck variance happens in 9% of epochs, that is a high probability, 4% variance happens in almost 4% of cycles which is still high.

Basically, nearly 1 in 10 epochs is subject to a 3% or more luck, nothing out of the ordinary here, run the code i posted above and you will understand how small of a sample is 2016 blocks with mean of 10 minutes.

But, having a combination of luck during an entire cycle? Well, it really takes a lot of luck.
On the other hand, I also think it's unlikely that anyone will turn their business around selling machines based on difficulty cycles.

So it was probably a combination of these and other factors, which we will most likely never find out.

Not that I don't want to take sides here, but I am guessing both theories are valid, and I would side with stompy a bit, not fully discarding the abuse-before-you-sell approach which is probably true for many farms (I have traded thousands of used gears and I know a thing or two about the subject), but I think it's more of a variance than the other option, maybe 70%-30% or 60%-40%, but noway it was all just overclocking before you they could sell.

My reasoning behind that is if there was room for 50EH to come online from a sustained overclocking it would have been the case long before, if you can safely abuse these gears for 2 weeks and have enough cooling not to kill them right away, why wait till now? also, usually, and from my own experience, most large sales are done in test and pick-up rather than ship away, someone who wants to buy 10M$ worth of gear would come to inspect the miners, see them run, then by them, those ready to ship are usually small in size.

So, is it possible that 5EH or so was the result of overclocking either for abuse or stress test for clients? yup it is, is a 50EH overclocking sustained for a whole epoch? probably not, so ya I'd guess it's likely a combination of both things with luck having the bigger impact.

I think we will find out soon tho, don't you think? that missing 50EH would come online soon, accompanied with the new gears replacing them, we should then see at least 15-20% increase in difficulty in a month or so time, which I don't think is going to happen, obviously, I could be wrong.


Food makes me feel sleepy, especially if it has red meat in it, I sure won't be able to write shit lol, but ya I just had my dinner at 2 AM, going to be my last reply for the night.


Luckily, I know what's causing this one, terrible weather and a lot of real-life work, too much thinking, and lack of sleep.


The halving is going to be frustrating to all miners that's for sure, I would hate to see it happen with a price below 60-70k at least, just imagine the price drops to 30-40k right before the halving, that's going to take a lot of popcorn dipped in a chocolate fudge to digest.

Sorry to hear that, but you could take Phil's approach and have a dinner tin german restaurant , some greasy bratwurst mit sauerkraut would either make your write that in 2 seconds or quit doing anything for the weekend.
But seriously, if you're not feeling well take care, with all the shit in this world you don't know what a headache might be caused by.


I think that the previous 8% was a fluke by a few percentages, so it was more like 4% than 8% that time and while this one might end -1-2% in reality is actually positive.

Phils' version is :


I'm saving all the popcorn for the halvening, these few adjustments are like the commercials before the movie, once that kicks in I'm going to watch the diff block by block, hutting all the press releases and hoping it will finally bring some sense in the market.

I think we are just bullshitting around waiting  for the next epoch.

I side more against variance for this drop and more for retooling.

With last jump being overclock the fuck out of the older s19 pros before you take them off line to sell.


He sides more with luck altering the numbers.

I think ?

I tried to figure out what the exact argument is about but I am not feeling too well, took me like 30 mins to write a 10 mins code, so is it true that you think this is all just luck while phill thinks the 50EH or so was actually present in the previous epoch and vanished today? would appreciate a TL;DR summary of what are you debating so that I don't take sides blindly, although, as usual, if it's a guessing game I rather be on the other side that you are debating given your track record of terrible speculation.

just went to a nice german restaurant and eat up 🆙 a storm.

and here I am back to the diff thread.

We're discussing and speculating each giving some arguments and not getting convinced by the other's.
Not that the outcome would matter, but what are you going to do on a rainy Friday?
As you said in the other reply, chit chatting..


and none of us as far as I understand claim that it's a fully 8% luck and more like half of it or one third or something around which would again increase the possibilities but, again not that it matters to the zeros.

I would be more interested in the actual technical challenges of having at least 50exhash in the datacenter and 50 out the next day, even assuming some larger farms managed to do this the same time.  I still believe that unless you're willing to spend extra for extra hands and shifts, and a lot here this won't be possible unless all the farms came to some agreement and all of Foundry and Bitmain and Antpool did it the same time. Why would they? That is another speculation!

Correct, it assumes no change in the actual factor that directly impacts the frequency of the events, which is why you just can't tell, looking at the previous epoch which had an 8.24% increase, assuming there was no change in the hashrate in either direction, the chance of that happening would be 0.016% which is pretty damn low, but if we were to assume half of that increase was actually caused by gears combining online and only the 4% was the result of luck, the chance of that happening increasing dramatically to 3.9%, if we assume it's 5% actually gears then that 3% is 9.3% chance.

The same would apply to the 288 blocks, if half of that assumed 25% is the result of gears combining online, and we talking only 12.5%, then that's a 2.27% chance, if we apply the same logic to 5% vs 3% and say roughly 17% was actual gears and 8% was luck than that's roughly 10% chance.

So ya, there isn't much you can takeaway from all of these guessing, except having a nice conversation with fine gentlemen like yourself and stompy.


edit: in the above, I used Poisson distribution function which is mainly "discrete probability distribution that models the number of events occurring in a fixed interval of time or space", I changed my code to use Erlang distribution which is " a continuous probability distribution that models the time between events in a Poisson process", while both have nearly similar results there is a slight difference, and I believe for the case in hand, using Erlang is more accurate.

Here is the code I wrote, which is pretty straight forward and would get 100% accurate result.


You can copy and paste this code to any online python compiler like this one https://www.onlinegdb.com/online_python_compiler if you want to run your own numbers, all you need to change would be the mean and the percent, the code I pasted above uses 5% or more blocks in a single day (144) blocks, if you want a whole epoch and 2% you would change the 144 to 2016 and the 5 to 2, let me know if you have any troubles running the code.

that formula is technically accurate.

but it can be flatter if you did the 2016 blocks with 1 or 2 million S19s or s21s in this case you need the difficulty of 84t

vs 1 s19 or s21. in this test you need a low difficulty to show the network is only 200th


I am saying the formula will work with either case .  but that the larger sample of gear flattens the luck more than the smaller sample.


the formula assumes perfectly equal hash . ie 1th anywhere is completely equal.

So if you have 1 machine for the whole network it is less likely the all th are true.

Not sure what are you guys trying to reach to, but there is a 0.0048% chance you get 25% or more variance in 288 blocks vs a 0.0245% chance of getting 8% or more variance in 2016 blocks, so the latter is more likely to happen based on my simple code which uses pyton erlang.cdf function.

I still think you're underestimating luck and probabilities:
11th of February till 12th 24 hours 177 blocks
https://blockchair.com/bitcoin/blocks?q=time(2024-02-11%2019:29:29..2024-02-12%2019:29:29)
next 24 hours: 141 blocks
https://blockchair.com/bitcoin/blocks?q=time(2024-02-12%2019:29:29..2024-02-13%2019:29:29)
If you have a 25% variable in 48 hours why is 8% instead of 4% so impossible in 13 days?

that is a fallacy.  the more you do an event the more likely it comes closer to 'normal' luck.



So 1,000,000 machines are more likely to be close to standard luck than 1 machine.



think coin tosses.

 1 flip = 1 machine and the result will be terrible luck or perfect luck

1,000,000 flips = 1,000,000 machines the result will not be 100% good luck or 100% bad luck.  it likely will be 490,000 heads 510,000 tails or some place near that.

When you do a luck calculation for the network now we are not one machine or 1 coin flip.

we are millions of machines or millions of coins flips.

 this far more likely means a 12 day adjustment due to magical luck or a 16 day adjustment due to badest luck around does not happen.

because that many 'flips' should show the bias if there is bias.


I doubt very much we are ever 5% above or below for a 2 week adjustment due to luck, we just never count in the  leveling effect that millions of units of gear has on luck.

And they have the ability to also increase the luck to ridiculous levels, too, because that's how luck works.

Of course the 350k could be done you just have to employ more people, you just have to have a lot of MW available for god knows what reason, you have to have extra racks at hand, again why would, and spend a ton of $ to do that as fast as possible for, what really? A extra week of mining?

The reason why I don't believe in it being possible is because I see all those press release from Mara, Riot, Hut8 and if you look at the rate they put in new gear despite having it in order you will see they barely manage to bring each of them 1-2 exahash online constantly, to think that they would get their gear at the same time and decided all of them to play like this and keep the s19 running for exactly the same time, this to me sounds like a possibility just as high of getting 1 extra block for the 6 you mine usually without doing anything extra.

As mikey said above, a combination of luck, extra gear delivered, some farm testing a setup, a farm maybe going down for maintenance as we speak yes, but just one single factor, I find it hard to believe!

Pace has gone up a bit:


Yup, weeks is the thing, it just doesn't happen in two weeks, on/off.
Besides, the whole renovation thing, let's assume you have 10 000 gears , you plan on giving away 3000 and replacing them with 3000 new ones, why would that company had the space, racks, cables, transformers, cooling already in place  for 3000 extra gear? When you renovate something you don't have already things set up in place and working for extra 20 or 25% of capacity.

I hear your points , both of you, both make sense and unfortunately probability is not math, you can't be too sure of what caused an event to happen when you have no control over it, it's likely a combination of both theories with different ratio, there has to be luck involved for that 8% figure to happen, I simply don't believe there was an addition of nearly 50EH in a single epoch, but I also know giving the halving is just around the corner -- many gear switching is probably taking place, gear prices are falling down rapidly especially the used once.

this truck can move 500 pieces

https://www.pensketruckrental.com/quote/#/reservation/truck-selection/truck-details/4

yellow truck
video icon
Watch to learn more about our 26 Foot Truck.
26 Foot Truck

3-5 BR Home (5-7 rooms)

Cargo Size: 25'11" L x 8'1" W x 8'1" H
Exterior Size: 35'5"L x 8'6"W x 12'6"H
Capacity: 1700 cu. ft., 10,000 lbs.
Fuel: Diesel, Up to 13 mpg
Tank: Up to 70 gal.
Seating: 3-person bench
Loading Ramp: 1,000 lb.
Tow Hitch: Yes
Transmission: Automatic
Air Conditioning: Yes


if the gear is on pallets it loads in 1 hour.

I am 67 and I just loaded and unloaded that truck with 9700 pounds of eth cables. we loaded it in an hour.

offloading was slower. 2 hours. Two guys. the other guy was 30 years old and did more work than me.

So yeah its work but 4 guys and two trucks could do four short trips in a day. that is 2000 pieces a day. or 10,000 pieces in a week.

We can argue good luck and bad luck which can be involved. But the larger the network the more level luck is.

 If 1  s19 was the entire network it can do 180 blocks in a day with luck
 If 1,000,000 s19 is the entire network they can do 180 blocks in a day with luck

In theory both are true , but you are neglecting a secondary factor the 1,000,000 separate pieces running independently have the ability to flatten the luck.

and the 1 piece has the ability to skew the luck easier.


Although I can bet you bottom dollar someone will say no to that idea.  The difference in difficulty cancels that factor.  It does in theory only if every machine is a perfectly equal machine running at perfectly equal conditions .

Then they would be like perfect casino dice. with zero balance.

The reality is a 1 machine network could have a great machine with great parts and it is a luckier machine.

It also could have a great connection to the blockchain.  So it could be biased to higher luck.


1,000,000 machines will flatten that a lot.  It will not show in the formulas but it is true.

Why not? If the miner is planning an equipment renovation, they may well do that.
Are you going to do this overnight? No, it will be a gradual process that will last a few days or weeks.