Title: The reason we do see long delays between blocks Post by: ripper234 on December 13, 2012, 12:34:11 AM (Nothing in this post except ultra simple math - feel free to skip it if you already know how many blocks take more than 53 minutes to be mined)
I just happened to spot this block (http://blockchain.info/block-index/324520), 53 minutes after it was created. There was no other block in site ... 53 minutes without any blocked being mined. Wondered a bit about what the odds for this event are. So here are my (very simple) calculations, in case you were curious. 1. I dug up this old question (http://bitcoin.stackexchange.com/questions/4690/what-is-the-standard-deviation-of-block-generation-times), which reminded me Bitcoin block times followed the exponential distribution (http://en.wikipedia.org/wiki/Exponential_distribution). It's basically like radioactive decay - each block has a "halflife time" of 10 minutes. 2. Realized that gamma = 10 minutes, and inputted 53 minutes into the cumulative distribution function. 3. Got 1-(e^-53/10) = 0.005 So, about 5 out of every 1,000 blocks take > 53 minutes to confirm. Nothing unusual about spotting a block like this every now and then. A few interesting numbers: 1. The chance a block is not found within 10 minutes = 36.7% 2. Not found within 20 minutes = 13.5% 3. 30 minutes = 4.9% 4. 40 minutes = 1.8% 5. 50 minutes = 0.67% 6. An hour = 0.24% (Divide by e = 2.718 to get the next number in line) Title: Re: The reason we do see long delays between blocks Post by: MPOE-PR on December 13, 2012, 12:55:55 AM each block has a "halflife time" of 10 minutes. This part is factually incorrect. What half life means is that given a sample of radioactive material, half of it will have disintegrated in that time. This is not the case for BTC blocks at all, probably l'Hospital would be more adequate to calculate likeliness of blocks not being solved in 10 minutes (they have 1/diff chances to be solved each hash, so after diff hashes their chances to be solved are X). Title: Re: The reason we do see long delays between blocks Post by: kyotoku on December 13, 2012, 01:46:27 AM each block has a "halflife time" of 10 minutes. This part is factually incorrect. What half life means is that given a sample of radioactive material, half of it will have disintegrated in that time. This is not the case for BTC blocks at all, probably l'Hospital would be more adequate to calculate likeliness of blocks not being solved in 10 minutes (they have 1/diff chances to be solved each hash, so after diff hashes their chances to be solved are X). Hence the double quotes. Title: Re: The reason we do see long delays between blocks Post by: ripper234 on December 13, 2012, 01:55:18 AM each block has a "halflife time" of 10 minutes. This part is factually incorrect. What half life means is that given a sample of radioactive material, half of it will have disintegrated in that time. This is not the case for BTC blocks at all, probably l'Hospital would be more adequate to calculate likeliness of blocks not being solved in 10 minutes (they have 1/diff chances to be solved each hash, so after diff hashes their chances to be solved are X). I wrote half life in quotes. It is not exactly like half life, but it is analogous. Imagine a million parallel universes all identical to ours (except for the seeds of random number generators in miners' computers). After ten minutes from now, in exactly 1/e of these universes a block is not found, and in (1-1/e) of them, a block is found. Technically 10 minutes isn't a half life, but rather x = -ln(1/2) * 10 minutes = 6.931 minutes. Each 6.931 minutes, in half of the universes (at least one) block is found, and in half of them no block is found. (Thanks for pushing me to review this) Title: Re: The reason we do see long delays between blocks Post by: jjiimm_64 on December 13, 2012, 10:44:31 PM Not a math wiz... shouldn't the network hashrate of something to do with the percentages? So I should assume that network hashrate is a constant during these calculations? perfect to keep a 10 minuter per block avg? Cause my thoughts are, if half the miners dropout, a block will take longer to find with the same diff....yes? Title: Re: The reason we do see long delays between blocks Post by: hamdi on December 13, 2012, 11:11:32 PM sometimes i happened to make payment and exactly then such a block comes up :)
Title: Re: The reason we do see long delays between blocks Post by: Yuhfhrh on December 14, 2012, 12:07:33 AM each block has a "halflife time" of 10 minutes. This part is factually incorrect. What half life means is that given a sample of radioactive material, half of it will have disintegrated in that time. This is not the case for BTC blocks at all, probably l'Hospital would be more adequate to calculate likeliness of blocks not being solved in 10 minutes (they have 1/diff chances to be solved each hash, so after diff hashes their chances to be solved are X). I wrote half life in quotes. It is not exactly like half life, but it is analogous. Imagine a million parallel universes all identical to ours (except for the seeds of random number generators in miners' computers). After ten minutes from now, in exactly 1/e of these universes a block is not found, and in (1-1/e) of them, a block is found. Technically 10 minutes isn't a half life, but rather x = -ln(1/2) * 10 minutes = 6.931 minutes. Each 6.931 minutes, in half of the universes (at least one) block is found, and in half of them no block is found. (Thanks for pushing me to review this) In one of those universes, a block will take 1 year to solve :D Title: Re: The reason we do see long delays between blocks Post by: bcpokey on December 14, 2012, 11:27:30 AM (Nothing in this post except ultra simple math - feel free to skip it if you already know how many blocks take more than 53 minutes to be mined) I just happened to spot this block (http://blockchain.info/block-index/324520), 53 minutes after it was created. There was no other block in site ... 53 minutes without any blocked being mined. Wondered a bit about what the odds for this event are. So here are my (very simple) calculations, in case you were curious. 1. I dug up this old question (http://bitcoin.stackexchange.com/questions/4690/what-is-the-standard-deviation-of-block-generation-times), which reminded me Bitcoin block times followed the exponential distribution (http://en.wikipedia.org/wiki/Exponential_distribution). It's basically like radioactive decay - each block has a "halflife time" of 10 minutes. 2. Realized that gamma = 10 minutes, and inputted 53 minutes into the cumulative distribution function. 3. Got 1-(e^-53/10) = 0.005 So, about 5 out of every 1,000 blocks take > 53 minutes to confirm. Nothing unusual about spotting a block like this every now and then. A few interesting numbers: 1. The chance a block is not found within 10 minutes = 36.7% 2. Not found within 20 minutes = 13.5% 3. 30 minutes = 4.9% 4. 40 minutes = 1.8% 5. 50 minutes = 0.67% 6. An hour = 0.24% (Divide by e = 2.718 to get the next number in line) So it should happen roughly 10 times every difficulty period, not too unlikely. I wonder how often this occurs though Quote Height Age Transactions 212168 52 minutes 677 25,884.30 BTC Deepbit 260.72 212167 1 hour 34 minutes 535 10,720.04 BTC BTC Guild 243.20 212166 1 hour 30 minutes 501 6,880.35 BTC EclipseMC 326.54 EDIT: Fairly often I guess Quote 212149 (Main Chain) 2012-12-14 05:39:51 00000000000002d26537d9eb939ed212171b45f2d70eccbf5b0d0939ee1e80e2 BTC Guild 189 104.49 212148 (Main Chain) 2012-12-14 05:31:39 000000000000039ca22d7de84a951c305535ec8f08449e5b059575308c266456 BTC Guild 365 226.70 212147 (Main Chain) 2012-12-14 05:33:58 00000000000002b54baa99eb63d0fd8b7e07d8d277be713257a7c23098a99b66 69.121.83.181 300 103.45 Silly timestamps. |