[CLOSED] BTC Guild - Pays TxFees+NMC, Stratum, VarDiff, Private Servers

[-ck]

I have been researching pool centralization and manipulation, one of the things I found is that discus fish and ghash set their accepted block size very low.  Discus fish obscenely low.
This allows them to plow through more of the smaller blocks faster than pools that play fair.

There's no such thing. "Smaller blocks" (presumably by that you mean ones with less transactions) take just as long on average to solve as larger ones.

The time the block should be the same, in statistical sense. But the time needed to propagate a smaller block to the network should be slightly short, giving the miner a slightly advantage in orphan block race.

How many orphans do you see here from btcguild?

[1713889341]

1713889341

[1713889341]

1713889341

[1713889341]

1713889341

[1713889341]

1713889341

[eleuthria]

How many orphans do you see here from btcguild?

Less than 1% (the general 'rule of thumb' that has been fairly accurate for the past 3 years is ~1%) in 2014.

[-ck]

How many orphans do you see here from btcguild?

Less than 1% (the general 'rule of thumb' that has been fairly accurate for the past 3 years is ~1%) in 2014.

Right.

But the time needed to propagate a smaller block to the network should be slightly shorter, giving the miner a slightly advantage in orphan block race.

And what's discus fish and ghash's orphan rate?

[eleuthria]

How many orphans do you see here from btcguild?

Less than 1% (the general 'rule of thumb' that has been fairly accurate for the past 3 years is ~1%) in 2014.

Right.

But the time needed to propagate a smaller block to the network should be slightly shorter, giving the miner a slightly advantage in orphan block race.

And what's discus fish and ghash's orphan rate?

GHash and Discus Fish use smaller blocks because they are by far the two worst connected pools on the network.  GHash's history of winning orphan races is pretty poor when they aren't the one that mines the next block.  Based on looking at organofcorti's charts (http://organofcorti.blogspot.com), it looks like GHash's average over their last 6000 blocks is around 1.5% (not seeing the actual stat anywhere over the full timeline, so going based on visual estimate).

Discus Fish doesn't have many (any?) stats publicly available, so their orphan rates aren't known.  I would bet they probably have a worse orphan rate considering they're in China, meaning their connectivity to the majority of the mining power on the network is very poor.

EDIT: And speak of the devil, Guild's most recent block was orphaned by Discus Fish even though BTC Guild's block was seen by most of the network well before theirs, since they found the following block.  Guild's block was seen *14 seconds* before Discus Fish's block by blockchain.info, and 96% of the nodes blockchain.info is connected to reported Guild's block over Discus Fish's.

[Sir Alan]

EDIT: And speak of the devil, Guild's most recent block was orphaned by Discus Fish even though BTC Guild's block was seen by most of the network well before theirs, since they found the following block.  Guild's block was seen *14 seconds* before Discus Fish's block by blockchain.info, and 96% of the nodes blockchain.info is connected to reported Guild's block over Discus Fish's.

Selfish mining is alive and well, then.  Any indication that DF might have been perfectly well aware that Guild had already reported 298926 before building on their own preceding block (298925)?

[eleuthria]

EDIT: And speak of the devil, Guild's most recent block was orphaned by Discus Fish even though BTC Guild's block was seen by most of the network well before theirs, since they found the following block.  Guild's block was seen *14 seconds* before Discus Fish's block by blockchain.info, and 96% of the nodes blockchain.info is connected to reported Guild's block over Discus Fish's.

Selfish mining is alive and well, then.  Any indication that DF might have been perfectly well aware that Guild had already reported 298926 before building on their own preceding block (298925)?

Selfish mining is withholding a block until you have built another on top of it already, or waiting for someone else to announce a block before releasing yours.  It's unlikely that this was a malicious act, but it could have been.  Discus Fish has always had terrible connections to the network due to being in China, which blocks the IPs to a *lot* of pools/bitcoin services which run some of the fastest relaying nodes on the network.

Discus Fish is a PPS only Chinese pool.  Selfish mining with ~15% of the network on a PPS pool would mean bankruptcy (assuming they're legit and not running a fractional reserve) due to the significantly higher orphan rates that selfish mining will always cause.


EDIT:  Of course that doesn't rule out Discus Fish doing selfish mining, just that it's unlikely to make economic sense.  Of all the pools out there, they'd be the easiest ones to pass off selfish mining as standard inefficiency since their stats are limited at best, the majority of the community can't even read their website, and they are in a country with 3rd world connectivity when it comes to worldwide networking.

[hurricandave]

Do you remove the orphans from the Dashboard manually? I still see them listed in the pool stats but not in the Recent Block rewards. Just curious, since I rarely click on Pool Stats.

[eleuthria]

Do you remove the orphans from the Dashboard manually? I still see them listed in the pool stats but not in the Recent Block rewards. Just curious, since I rarely click on Pool Stats.

I've been tweaking the dashboard a little bit the last few days.  Orphans weren't supposed to show under Recent Block Rewards at the moment, it's going to be a configurable option (Default off) because of how many emails they generate from people who don't know what it means.  They were defaulting On due to some other changes I'm making related to Scrypt mining (not currently visible to the public).  They've actually been Default 'On' until today.  We just hadn't had any orphans out of the last ~400 blocks (until today).

They've always been shown under Pool Stats (though they used to show the block # instead of identifying the block as an orphan up until about 1 month ago).

[DevonMiner]

We just hadn't had any orphans out of the last ~400 blocks (until today).

Yeah, I was watching the confirmation status ... took ages ... and I suspected all was not well .. and then it orphaned. It happens.

[eleuthria]

We just hadn't had any orphans out of the last ~400 blocks (until today).

Yeah, I was watching the confirmation status ... took ages ... and I suspected all was not well .. and then it orphaned. It happens.

It was actually orphaned much earlier than that, the problem was the old way the dashboard filtered orphan vs waiting for confirmation required the pool to find another block first.  I updated it so it will recognize an orphan earlier.

It's *very slightly* possible that a block might show up 'Orphaned' on the dashboard (once you can re-enable that display) and then end up getting confirmed & paid.  This would happen if the hot wallet server [used to identify new blocks and confirmations] sees the block after a competing block, since the hot wallet server is not actually one of the mining servers that relays our block solves.  Another pool (or our own mining servers) might then build off the erroneously-marked orphan.  The mining servers and the hot wallet are in the same datacenter though, so this probably won't ever happen.

[eleuthria]

The hot wallet lost connectivity (tweaking some upstream filtering to prep for Scrypt mining) earlier.  Payouts weren't being broadcast over the network as a result.  It's been fixed and they should be rebroadcast shortly.  Sorry for the delay in getting confirmations for those payouts.

[os2sam]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.

Great idea.

[DPoS]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.

Great idea.

Well, seeing how things get removed from this thread, why bother?

let the data over time speak loudly on this one


Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%

just a bad streak.. seems legit

[stan258]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.

Great idea.

Well, seeing how things get removed from this thread, why bother?

let the data over time speak loudly on this one


Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%

just a bad streak.. seems legit

Pool is below 7.5 P now :{

[kendog77]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.

Great idea.

Well, seeing how things get removed from this thread, why bother?

let the data over time speak loudly on this one


Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%

just a bad streak.. seems legit

Bad luck can't last forever unless it's not just luck and there is something more nefarious going on. Other pools may have also figured out how to gain some sort of mining advantage and are good about keeping it a secret. If someone discovered some sort of advantage that allowed them to earn more BTC, sharing it with anyone while it could still be exploited would be silly because it would eliminate the advantage.

I like the fact that eleuthria is so transparent regarding BTCGuild and urge eleuthria not to remove useful pool information like luck statistics. Most other pools don't even bother posting information about luck, so why have it on your site and open yourself up to scrutiny if you're trying to pull some shenanigans? That doesn't make any sense to me.

Since BtcGuild and Eligius are both underperforming right now, some other pools must be over performing since the difficulty keeps going up. Does anyone know which pools are experiencing better than average luck right now?

[Minor Miner]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.
Great idea.

Well, seeing how things get removed from this thread, why bother?
let the data over time speak loudly on this one
Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%
just a bad streak.. seems legit

Pool is below 7.5 P now :{

Perhaps that the actually size of the pool for a while.   Is it not possible that 15% of the HR rate (the same 15% of miners) were consistently unlucky?   I believe in variance but at some point you have to ask if there are not some free riders on this bus.

[kendog77]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.
Great idea.

Well, seeing how things get removed from this thread, why bother?
let the data over time speak loudly on this one
Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%
just a bad streak.. seems legit

Pool is below 7.5 P now :{

Perhaps that the actually size of the pool for a while.   Is it not possible that 15% of the HR rate (the same 15% of miners) were consistently unlucky?   I believe in variance but at some point you have to ask if there are not some free riders on this bus.

That's an interesting concept. Perhaps there is some sort of pool bug that is providing some miners with duplicate or useless work and counting shares that should not be counted? I do trust eleuthria and think he is an honest pool operator, but people are human and humans make mistakes.

Did the bad luck start around the time of all the backend pool changes and rollout of scrypt guild?

There also have been cases of pool users being re-directed to other pools via man-in-the-middle attacks recently, so perhaps some attackers figured out how to steal a small number of solved blocks for themselves.

We're all just speculating here, and the only one that can really look into this and know for sure is eleuthria.

[stan258]

It reminds me of how Deepbit disappeared from the pools list.

It's no wonder [Tycho], and others, have disappeared from these forums.  Let's accuse everyone else, with NO evidence, of malfeasance as well so that we have no responsible people left in the community.
Great idea.

Well, seeing how things get removed from this thread, why bother?
let the data over time speak loudly on this one
Approximate Pool Luck* (24H / 3D / 1W / 2W / 1M / 3M / All Time): 57.525% / 66.081% / 76.537% / 80.554% / 84.873% / 90.909% / 98.545%
just a bad streak.. seems legit

Pool is below 7.5 P now :{

Perhaps that the actually size of the pool for a while.   Is it not possible that 15% of the HR rate (the same 15% of miners) were consistently unlucky?   I believe in variance but at some point you have to ask if there are not some free riders on this bus.

That's an interesting concept. Perhaps there is some sort of pool bug that is providing some miners with duplicate or useless work and counting shares that should not be counted? I do trust eleuthria and think he is an honest pool operator, but people are human and humans make mistakes.

Did the bad luck start around the time of all the backend pool changes and rollout of scrypt guild?

There also have been cases of pool users being re-directed to other pools via man-in-the-middle attacks recently, so perhaps some attackers figured out how to steal a small amount of solved blocks for themselves.

We're all just speculating here, and the only one that can really look into this and know for sure is eleuthria.

Didnt pps pool 50btc have this issue back in October?

[Minor Miner]

That's an interesting concept. Perhaps there is some sort of pool bug that is providing some miners with duplicate or useless work and counting shares that should not be counted? I do trust eleuthria and think he is an honest pool operator, but people are human and humans make mistakes.

Did the bad luck start around the time of all the backend pool changes and rollout of scrypt guild?

We're all just speculating here, and the only one that can really look into this and know for sure is eleuthria.

I prefer the more tinfoil hat theory of someone spending $1MM to make a chip and boards and like everyone who made ASICs, they have a few flaws.   They work great on the test set, and maybe even worked great for a while on real work, but because of some screwup they just never return solutions for work higher than difficulty xxxx (insert your own paranoid number here).   At some point the people that made these miners figured it out and thought (oh shit, we are solo mining and our best share is NEVER above 3,000,000,000, but we have invested all this money).   What do they do?    They would have to point the miners at pools right?    And they could not point them at small pools because they might get noticed.   So, they make accounts at the largest pools and produce a ton of work (but never above a certain difficulty because of the flaw in their miner) and get a proportionate share of the blocks mined.

Is that just too crazy?   Maybe.  I know most people would assume those people would just stop mining and write off their investment right? 

[eleuthria]

1) That's an interesting concept. Perhaps there is some sort of pool bug that is providing some miners with duplicate or useless work and counting shares that should not be counted? I do trust eleuthria and think he is an honest pool operator, but people are human and humans make mistakes.

2) Did the bad luck start around the time of all the backend pool changes and rollout of scrypt guild?

3) There also have been cases of pool users being re-directed to other pools via man-in-the-middle attacks recently, so perhaps some attackers figured out how to steal a small number of solved blocks for themselves.

We're all just speculating here, and the only one that can really look into this and know for sure is eleuthria.

1) Dupe work is impossible.  Every share you've submitted for the current block is held in memory and compared against to determine if you've submitted a dupe.  It can't be done across multiple accounts because each account has an independent ExtraNonce1, making work for one account not valid for another.

1b)  As stan258 mentioned, 50BTC had a problem with over-crediting work.  That's because they completely screwed up user definable difficulty and didn't realize the most obvious exploit available for doing that.

2)  There have been no backend changes for ScryptGuild.  There have been no changes to the backends period in the last 9 months.  6 of those months were positive on luck.  The last 2 were not.  The 9th month isn't available since luck only started being tracked across multiple difficulties 8 months ago (prior to that it only showed the most recent shifts of the current difficulty).

3) The MITM attacks are not MITM.  They're a flaw at the end user, clearly identified by the fact that the same users get hit repeatedly on multiple pools, even though the vast majority of users are never hit.  However, assuming it was a MITM, it would not be possible to steal a block for yourself.  A hash/share is only valid if it pays to the pool wallet.  If you try to change the destination the work is immediately worthless because it will not produce the same result on a different payment address.



This is not an isolated incident.  Eligius has had low luck in the last 3 months as well, which is roughly when Guild's luck took a turn for the worse.  So is p2pool, though they're so small it's very hard to call a line between normal variance and abnormal when comparing it to Eligius/BTC Guild which are significantly larger.  Most of the long term pools (BTC Guild, Eligius, BitMinter, Ozcoin, triplemining) have operators which communicate regularly, and we have been discussing this recent swing in downward luck.  This isn't something I'm brushing off or ignoring, but it's also completely out of my control, thus the anger/short temper when it comes to people bitching about it.  It may be a problem in firmware/hardware of ASICs which have been released in the last few months.  It may be a coincidence.  It may be an actual withholding attack in the wild, though this is unlikely because doing a withholding attack on a fee-charging PPLNS pool makes no economic sense.  You'd essentially be paying a fee twice, because you lose income from the pool fee *and* even more income in the form of withholding blocks from the pool.

One thing it isn't, without any doubt, is a backend problem.  There has been only one change to the backend in the last 8 months that even brushes against the share processing.  That change was updating how it stores difficulty (moving from a 32-bit integer to a double).  This doesn't actually affect how the server treats the share though, it's purely a display value used in internal reporting.  Aside from that, the share processing side of BTC Guild has been untouched since the day stratum was launched.