Topic: 2024 Diff thread happy New Years. - page 13. (Read 3704 times)

One thing I found interesting, in the example model I presented, Binance, despite only having a 6% hash weight, is the one that profits the most from mining fees. They tend to find more blocks when rates are higher than when they are lower.

Of course, it is not yet possible to see a pattern here, because we are only analyzing a period of one month. But I found this fact curious.

It's not about coping Phil, it's about seeing both sides, you look for a pattern I look to see if that pattern is not broken for weeks in a row.

In my last post I mentioned that we had 103 in 24 hours:
https://blockchair.com/bitcoin/blocks?q=time(2024-01-16%2011:50:23..2024-01-17%2011:50:23))
At that time feebudy was saying this:

103 blocks after a previous day of 135 blocks:
https://blockchair.com/bitcoin/blocks?q=time(2024-01-15%2011:50:23..2024-01-16%2011:50:23))
and only 58sat/b!!!

But then you have a thing like this:

with 6 blocks in 2 hours and 10 in 3 hours, that's 50% of the capacity and you finally got fees up to 80sat/b.

So how much of the variance in fees and big fee blocks is just really bad timing between blocks and how much is average fee increase?

Also, I would like to point a thing you're forgetting in this hashrate turn-off:


You're going to mine more for the same reward, although your 5% is now out of 95% because the network has lost that hashrate it's not going to be 144 blocks per day but also 138 blocks per day for 15 days, it's not a real gain, it's just like a competitor has gone out of business but he has taken the clients with him also.

Also when you look at block rewards, it's not like every spike is triggered just by traffic, there are cases like this:
https://mempool.space/tx/254115f95440409f70a1b93d3870712d4c0a3865fef4d1d86542ac96be18e081
This guy paid 616 sat/vB , for sure it's not about the next block fees sitting at 80sat/b so it would be far more accurate to think these guys would have anyhow paid a ton more, even with fees at 20sat/vb.

Back two weeks ago and next blocks fee was 30 and this guy overpaid 48x to 1,501 sat/vB.
https://mempool.space/tx/b9b4bc87c93bb06bacf1980c2a9fa07ff149ada6981ec496e52dc887697c9788
So when looking at the fees should we not consider what the real floor increase is and not how influenced is by somebody like Binance consolidating and paying 240sat/vb for half a block worth of space in one day?

So how much for these fees is actually because of the congestion and some inscribing monkeys at 4 times the price?

One thing I noticed, is that the "unknown" classification, maybe not so unknown. In collecting information for this test, I noticed that there are identifiable pools within the "unknown" field. I didn't explore it, because I didn't think it was relevant to the test.

This test is not a test of theory, it is just a test of data collection.

In this sense, what data do you think I can collect, and how can I work with it, so that we can study this issue together?

The idea is to collect data every month or every cycle. (For management, each month will be easier.)

Anyway, what do you suggest I add or change to the "model" data I presented?

fee shares should be fairly random.

Really watch the unknown group

you have 10 groups 9 pools and the 1 unknown.

By sheer random luck the odds of being in the top 3 fee average would be 30%

The least likely group to try fee manipulation would be the unknown group

so 2 months out of the top 3 are .7x.7 = .49
3 months out are .7x.7x.7= .343
4 out = .2401
5 out = .16807
6 put = .117649
7 months= 0.0823543
8 months = 0.05764801
9 months = 0.040353607
10 months= 0.0282475249. if you get to this I would consider that the under over clock is in effect.

but as fuzzy has shown this drop is very like;y just a cold weather snap


fees are  as follows.





there are multiple factors and teasing it all out is not going to happen

1 fomo rises more chain action = fact
2 power  up and down due to  contracts in Texas = fact
3 ordinals intentional spam action = fact
4 hash rate manipulation = theory but I am in a 90% sure belief of this.
5 unknown shit  which does exist but is unknown
6 random luck or variance = fact and should be very hard to predict

I am using random luck to attempt to rule in or out  number 4

as fees should more or less be random.

tracking the top ten pools well 9 plus unknown could show some interesting fee patterns.

2024 is a very good year to do this as it is a 1/2 ing year.

Based on what was said, I made adjustments to the analysis and added another analysis table.



^{Days of each cycle (in order in the table): 10 | 13 | 8}



This is just a test with the month of December, the idea afterwards will be to use the cycles as a reference. But first I want to adjust the way of analysis.

What other suggestions do you have?

There is no 'if' about it. ERCOT's pricing & usage deals with the major mines in Texas has been oft discussed here. Same thing happens in the very hot summertime for the same reason.

Those usage deals are what has enabled ERCOT to invest heavily in solar & wind power. A historic problem with solar & wind energy has been how to deal with how much and when power is produced. Because the grid in Texas is mostly isolated from the rest of the US grid ERCOT cannot bring in power when needed and cannot ship out excess power production. The deals with miners has let ERCOT vastly overbuild solar/wind production because the miners will act as a massive base load to use it and like now, when conditions warrant the miners will throttle down as needed with ERCOT paying them usage credits for doing it. I for one cannot think of any other industry that can easily & quickly per-site switch on/off 10's to 100's of MW worth of power usage with very short notice.

A rare win-win for all involved.

The first logical explanation, for a drastic hash change. If they really did that, I think it was a very sensible decision.

Although it does not explain similar cases in the past.

Good call.

A good part of the drop is no doubt the extremely cold weather in Texas and the rest of the USA. ERCOT no doubt has Foundry and others throttling back so power can be used in the rest of the state https://www.telegraph.co.uk/us/comment/2024/01/16/green-wind-solar-energy-freezing-texas-power-grid-blackout/

the theory is not 7 under 7 over.  But you can stick with that if it helps you cope.

the theory is under till mempool jams and over til mempool clears up

so it could be 3 under 11 over

or 3 under 4 over 3 under 4 over. 

the best way to say it is wrong is to see unknown block winners get in the top 3 fee average at least 2 times in a year.

One thing is for sure, they are having a run of bad luck. Because they are mining 20% less of the blocks they normally mine.

Foundry only found 25 blocks in the last 24 hours. Against the 32 blocks they had found in the previous 24 hours. And in the 24 hours before, they had made 39 blocks.

EDIT: And as I had the filter on, I took advantage and saw that on the 13th there were 32 blocks, on the 11th and 12th there were 42 blocks. Anyway, they really are very unlucky...

@joker_josue
 
So, how was that theory about 7 days of slow blocks then 7 days of fast blocks? 


In the last 24 hours we had 103 blocks so if we would say that the hashrate went truly offline it would mean only 71% is mining right now to get those blocks, which would mean Foundry itself on whole to have shut down in the last 24h    Now we need to blame antpool for this since they are the ones losing 5% marketshare.

Anyhow, nice breather we got here, it was really getting weird with tens of exas over exas piling up!

first add all blocks made

total blocks 4599

4599/31 = 148.354 blocks a day for December

then do antpool

1228/4599 = 26.70%

then do foundry

1338/4599 = 29.09%

antpool should do 26.70% x  148.354 = 39.6 blocks a day as a norm

foundry should do  29.09% x 148.354 = 43.15 blocks a day as a norm

see how much they deviate.



next do this

unknown   991.17/ 284  =  fee of 3.48 BTC a block
AntPoll    4656.08/1228 = fee of 3.79 BTC a block
ViaBTC    1899.55/  521 = fee of 3.64 BTC a block
Foundry   4905.87/1338 = fee of 3.66 BTC a block
F2Pool     2047.94/  540 = fee of 3.79 BTC a block
Mara         837.72/  221 = fee of 3.79 BTC a block
SBIC         309.73/   95 = fee of 3.26 BTC a  block
Binance   1099.82/ 270 = fee of 4.07 BTC a block
Poolin.      170.99/  48 = fee of 3.56 BTC a block
SlushPool  198.26/54 = fee  of   3.67   BTC a block



I bold typed the three worse pools.

Getting a block with a lot of fees should be random

So do this for 4 months in a row and see if these three finish in the bottom 3 over and over and over.

the other 7 could be doing cartel or could simply be doing the ideas I talked about here.

unknown are the little guys they should lose out in average fees if bigger pools do what I am saying.

SBIC and POOLIN may move up and down but the other block hitters  should be in the bottom 3 on a constnat monthly basis if the hash down then up then down idea I talk about is really being down.

It should get worse in April.

I'm thinking about doing this type of analysis:




Do you think it makes sense? How could this data be processed?

Well if I am correct it will continue and become more noticeable after the ½ ing.

My theory is to down clock to crowd the mempool first 2-6 days of a 14 day jump.

Then overclock the last 8 days to reap high paying blocks.

This also keeps hashrate in check allowing for a 1 or 2 % jump vs a 5% jump.

When my partner was in China he watched Canaan assembly factories.

I can tell you China can make 2x to 3x the hash we see at low costs.

Ie https://shop.canaan.io/products/avalon-miner-a1246-85t-3420w?VariantsId=10019

only 680 about 8 bucks a th.

I suspect that there is more than 1000eh of so called modern hashing gear with another 300eh of older gear.


If you are big enough and have paid the gear off having 1 megawatt of power and 2.1megawatts of different gear all paid off is not that dumb as it sounds.

btc hot mine all your btc gear. >>>>>>>>>> say 700kwatts
ltc/doge hot mine all your LTC/doge gear.> another 700kwatts
gpus hot mine all your gpus. >>>>>>>> another 700 kwatts


you only have 1000kwatts it is simple to  do 700 of you best and only 150 and 150 of the other 2.


s19s have a 5-6 year life
L7s have a 5-6 year Life
gpus have a 5-6 year life

lots of easy movement in a big place can be done.

plenty of software allows shutting gear on  and off.

at stompix

kano did not answer

my questions for this jump

we were at 527 vs 573. whats the math on that

and we have done  705 vs 658 whats the math on that

and we did that back to back whats the math on that.

I know that it has to be fairly long odds.



by the way they can pick any part of the 14 days to over and under clock




fees are lower

Kano's answer on probability!
Then go here:
https://bitinfocharts.com/comparison/bitcoin-confirmationtime.html#3y.
If still wondering, do an experiment, throw dices for 10 minutes see how many you got then repeat see the differences 
I find it always amusing how when there are fewer blocks there is obvious something wrong but when there are more it's no longer an issue!

The epoch in October started with 156,155, the one in middle November had first 161 then 121, the one in mid December went 158,144,149 so it doesn't always happen like it's some kind of pattern.
Go to the other topic for the 2013 and see how many times we had completely unrealistic positive values in the first days, this is just from my own history:

https://bitcointalksearch.org/topic/--5431167

I'm not a miner - unfortunately, so I ask: What logical explanation is there for this?

How is it that during 7 days the hash drops significantly, after these 7 days the hash returns to the values it was before? What happened during those 7 days?

in the plus zone +8 blocks

We are in the middle of the difficult cycle, and after several days, with a negative rate close to 8~10%, it has now returned to 0~1%, and in turn a new increase in rates.



It's really interesting to analyze these numbers. It is difficult to think that this is something normal.

so far 27 cycles from jan 2023 to now.

I would think at least nine cycles in the last 27 match the pattern.

look at the concept I describe . if you make it work and it pays more and it will as rewards decrease you simply do not need to do it every jump.


 the patterns happened a lot in2023 and the first month of  2024.

as rewards drop with 2024 ½ ing it is easier to do.

and getting closer to 0 blocks behind. from 50 behind now to 20 behind




but this round of down then up did not give them as many fees as other rounds of this game. most fees were under 0.75 btc a block



so if the next four years happen with 2 of 3 jumps doing what i say pools will make solid coin.



back to 0%