I refrained from mentioning GHash and Discus Fish because they don't have publicly available historical stats, at least that I was able to find.  Eligius has a CSV export for it's block history, and p2pool has a 3-month luck chart which is reported in the same way BTC Guild reports luck (% compared to expected earnings, rather than CDF).  Whether or not they have had good luck, neutral luck, or similar bad luck is a mystery to me, so I only report on the comparable pools that I'm aware of.

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.

Upgrade to the newest cgminer.  It won't allow client.reconnect to send you to a server on a different domain name than the one you're mining against.

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.

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.

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).

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.

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.

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

No problem.  It's something very common with coins who have abstract reward days or frequent halvings.  Since people aren't locked into mining a specific coin with their hardware, they'll simply mine something else after a reward change.  Eventually the difficulty will end up being (roughly) half of what it would have been previously.  A good example of this is Earthcoin, which has 2x and 5x reward periods.  During this time, the difficulty basically increases to 2x or 5x what it normally is, so for profitability it is virtually the same as any other day.

It's not an exact halving of difficulty since Doge has a decent number of loyal miners, but with multipools being such a huge part of the average DOGE hashing rate, the difficulty still dropped significantly so that it's still worth mining as a multipool, just not for quite as long as it normally would be.

Current expectation is BTC Guild's scrypt mining will be starting next Friday.  So the timeline I posted previously is still roughly accurate.  An update will be posted here and on the website once the BTC Guild scrypt pool is up and running.

"Still" piling up is not a good term for this, considering they stopped piling up and were sold the last time it was reported.   CSC requires manual cryptsy refilling because of how many coins we have that can't be sent until they have more confirmations.

The difficulty dropped sharply after the reward was reduced, because multipools wouldn't touch it til it dropped lower.  The halving had very little effect on profitability, though that is why we've been hitting LTC more often (less time spent on DOGE because the diff goes up fast once it drops low enough for multipools to hit it).

You've already admitted you're not an expert.  That is also obvious by what you've said.  I have kept what you've posted in this thread even though it is making assumptions that are completely false.

Ghash, Discus Fish, and KNC do not post their luck figures.  They also don't publish any historical data to calculate it externally.  For all we know their luck is as bad or worse.  Even if it isn't bad though, that doesn't mean anything because variance doesn't only swing one way.  Just like how BTC Guild had over 6 consecutive months of positive luck.  But nobody ever bitches when things are working better than expected, but they sure do forget it quickly.

However, they did not "figure something out" and the fact that you even mention that shows you have no clue what you're talking about.  If they "figured something out" it would mean SHA256 is fundamentally broken and Bitcoin is insecure.  Mining is a RANDOM process.  That is a *FACT* unless SHA256 is broken.  There is no way to "more efficiently distribute work".  Finding a difficulty 1 share is a 1 in ~4.2 billion chance per hash.  Finding a share at current network difficulty is a 1 in ~33,600,000,000,000,000,000 chance per hash.

I really just need to remove that chart.  Eligius' thread doesn't get derailed this frequently over this crap and their 90 day luck is almost identical to ours. (Guild: 92.3%, Eligius 92.5% for the last 2000 blocks [calculated yesterday], which is ~88 days, p2pool 90.8%).

For BTC, I will adjust the payout threshold to 0.0001 for all users when mining is closed on ScryptGuild.  It will deduct 0.0001 from your payout in order to cover the txfee if you do not have > 0.001.

It's been less than 1% since the paid orphans were removed.  We haven't had one in the last 300 blocks now.  The historical orphan rate (not counting the 0.8 hardfork event) has been about as close to 1% you can get given the sample size in comparison to the expected frequency.

Yep.  The pool no longer pays for orphans, but it doesn't require a 100-120 confirmation wait.  The likelihood of a block getting orphaned after it is confirmed is extremely low, to the point it just doesn't seem worth delaying the application of rewards to wait for even more confirmations.

When another block on the network confirms it.  That only shows up when BTC Guild just found a block and no new block on the network has built off it (IE:  We don't know if it's going to be orphaned or not yet).

BTC payouts have been working fine.  They have not failed to go out.  0.0x+ goes out 30 minutes after the hour.  0.00x goes out ~30 minutes after midnight server time.