It would be great to get someone to double-check this work? Repo link above. The solid red line is June 15. The dashed red lines are each time the bomb’s period was at period 31 for comparison. (Period 31 is where it was reset in December.)
Latest installment. I guess we’re starting to see a very slight effect. Notice from previous bombs how quickly it comes once it starts to come. Red vertical line is June 15.
Latest installment. I see a slight dip. I expect we will start seeing an effect in the coming weeks. A higher hash rate delays the first ‘appearance’ of the bomb, but there’s a concern that once the bomb appears it will be more virulently.
The horizontal grey lines are the number of seconds between blocks given the number of blocks produced in a week on the y-axis.
I’m going to post another chart shortly to show another view.
This chart shows the real block number (in red), the fake block number (in green), and the bomb’s period (the staggered blue line which changes discontinuously every 100,000 blocks).
The dashed grey horizontal line is what I call the “We should start paying attention” line. If you look closely at previous bombs, you can see the slight diminishment in the block number over time. We’re not seeing that yet, but stay tuned.
is a “bomb” a stark decrease in the block-production rate over a certain period of time? what is it about the time preceding hard-forks that cause such down-spikes? is it because node-runners deactivate their nodes to update the software?
what does this mean for the transaction rate for ethereum? does it remain at a stable rate? do these bombs have measure-able impact on gas prices?
I’ve written about a lot of this here: Adventures in Difficulty Bombing. An exercise in predicting the future… | by Thomas Jay Rush | Coinmonks | Medium. That link also has links to a bunch of other articles as well.
About the impact on gas price, I’m not sure. It has had, mistakenly I think, a serious effect (at least on the arguments surrounding) the block reward, but I don’t know about gas prices.
At block 14,600,000, which happened this week, we entered into period 39, which based on previous bombs is when we usually start seeing a diminishment in block times. I’m not seeing that yet. This is not yet concerning, but is becoming “interesting.” (See below).
Here is an estimate, assuming no growth in block times, of when the next three doublings in difficulty will happen. (A doubling happens when the period changes – period = fakeBlockNum/100000 – fakeBlockNum = realBlockNum-10,700,000 – the additional difficulty at each block is 2^period).
blockNumber timestamp date name 14000000 1642114795 2022-01-13 22:59:55 UTC 14100000 1643450273 2022-01-29 09:57:53 UTC 14200000 1644784417 2022-02-13 20:33:37 UTC 14300000 1646123109 2022-03-01 08:25:09 UTC 14400000 1647466065 2022-03-16 21:27:45 UTC 14500000 1648811420 2022-04-01 11:10:20 UTC 14600000 1650160464 2022-04-17 01:54:24 UTC 14700000 1651492241 2022-05-02 11:50:41 UTC (est flat) 14800000 1652822241 2022-05-17 21:17:21 UTC (est flat) 14900000 1654152241 2022-06-02 06:44:01 UTC (est flat)
In this estimate, which extends to period 42 (the same period just prior to the Byzantium fork–see above charts), I add NO additional seconds to each block. This may seem conservative, but it tries to take into account the possibility that the much higher hash rate we’re seeing now (compared to the past) may be masking the increasing difficulty of each change of period (bomblet).
This next group of data is a slightly less conservative estimate showing a one-second slowing down of blocks with each doubling. If this happened, we may be at 16-second blocks by the end of June):
blockNumber timestamp date name 14000000 1642114795 2022-01-13 22:59:55 UTC 14100000 1643450273 2022-01-29 09:57:53 UTC 14200000 1644784417 2022-02-13 20:33:37 UTC 14300000 1646123109 2022-03-01 08:25:09 UTC 14400000 1647466065 2022-03-16 21:27:45 UTC 14500000 1648811420 2022-04-01 11:10:20 UTC 14600000 1650160464 2022-04-17 01:54:24 UTC 14700000 1651837086 2022-05-06 11:38:06 UTC (est onesec) 14800000 1653730218 2022-05-28 09:30:18 UTC (est onesec) 14900000 1655823350 2022-06-21 14:55:50 UTC (est onesec)
My growing concern is that due to a much higher hash rate, unlike previous bombs, this bomb’s effect may not appear at all until it appears massively. (In other words, this bomb may truly explode!)
The higher hash rate may be hiding the initial, smaller shocks to the system (the bomblets) which are being recovered from more quickly, so much so that they don’t reveal themselves in the weekly production. (I looked at a daily chart that also does not show much of a slow-down.)
The concern is that once the system loses its ability to recover from the shock of a doubling, it will lose that ability in a serious way.
I’ll keep watching it.
We should start seeing a recognizable diminishment in block times soon. If not, we’ll need to dig deeper.
You can also track this on Coin Metrics here
Chart as of today. The bomb is clearly starting to show. It’s reasonable to estimate its future progress based on previous bombs.
I guess a new hardfork for delaying the difficulty bomb is around the corner
Some new charts. Back in December, I was asked to predict 15-second blocks by mid-June. Impossible to be certain, but I wouldn’t be surprised if this were the case.
The red line is June 15.
The dashed red lines are the last time we were at the same period in each of the previous forks. Much higher hash rates now has tamped down the effect, but that won’t last forever.
The blue dashed vertical lines are the previous forks.
The table tries to compare past forks with the current one and predicts the next bomblet to be around May 18.
Latest chart as of last night…I’m quite confident that the is the correct data, but I would be happy to have others produce similar results independently if possible.
It looks to me like it’s getting close the time to make a decision. I think we can expect a fairly quick decline from here (based on previous bombs).
Here’s my prediction for blocktimes assuming hashrate stays consistent. Disclaimer: This is based on reported block timestamps which may differ materially from the actual time between blocks.
Why do you say that reported block times may differ from actual block times? Where are your “reported” block times coming from?
I used the recorded block timestamps (as set by the miners) when looking at historical blocktimes for a given hashrate. These timestamps are often quite different from the actual time the block was produced. So, for instance, it may have taken 20 seconds to mine the block, but the block timestamp could be set at just 5 seconds ahead of the previous block. My data would read that as a 5 second blocktime, whereas the 20 seconds is really what we are interested in for purposes of network congestion etc…
I haven’t thought through exactly how or if this would impact the results when averaged over a large dataset, but I thought it was worth mentioning.
ETA: This (potentially) matters for my data since I plugged one simulated block at a time into the difficulty formula to get this result. If you use blocks/day or something like that as a metric it wouldn’t matter.
Thanks. While I don’t really make predictions (purposefully), it seems that the actual data as shown by my chart and your prediction (at least for late May and early June) align fairly well. Would you say that’s true?
Just using my eyeballs, my charts is hinting at 15-second blocks through mid-June and 16-second blocks into late June / early July which align pretty much with your prediction.
Apparently, there’s talk of setting back the bomb. I think that’s totally reasonable.
There’s a link above to an article I wrote in November trying to figure out how far back and when to set the bomb for Arrow Glacier. I’d love any comments/thoughts about that.
Also, here’s a spreadsheet I put together to try to figure out how far back and when to set back the bomb this time.
Again, I’d love for others to look it over.
If the sharing permissions are not correct, please let me know and I’ll share it directly.