We were quite a bit above 0.01 for a few days.  Yesterday (well, ~36 hours ago to ~12 hours ago) was pretty harsh.  Very few Dogecoin/Litecoin blocks even though a lot of hashes were being directed that way.  Pool simply needs more speed to reduce the variance when the most profitable coins are the ones running at a high difficulty.

A minor update since technically it's a new day:  I held off on adding new coins today, in order to make sure everything is continuing to operate smoothly with auto conversion.  I have multiple new coins sync'd up with their blockchains ready to add  tomorrow (or today, depending on what you consider 12:30 AM.  I personally call it "tomorrow" if I haven't gone to bed yet).  I am NOT adding Auroracoin.  The difficulty is obscenely high for our pool speed, and I simply despise any coin with the audacity to have a 50% premine.

Difficulty is mostly for bandwidth reduction, and it also helps reduce load on the pool.  It doesn't make you earn more (or less), it simply optimizes the amount of communication required between you and the pool.  Each step you increase difficulty reduces the chatter between you and the pool by half of the previous value.  It increases variance on your end slightly, but the recommendations I gave above are recommended because they end up having a less than 1% effect on your 24 hour variance (personal variance refers to share submission rates, not to be confused with pool variance).

3 years of experience running mining pools, 2.5 years running one of the top 3 (holding top spot for almost a full year) mining pools, and 2 of those years where my sole source of income was from it.  Custom backends, frontends, and DB schema rather than some github pool-in-a-box.

PPLNS is *always* a higher expected value.  It has good days and bad, but in the long run it averages out to > 5% better payouts than PPS.

The workers page shows you the recommended speed for each difficulty.  Basically:  Round your GH/s down to the nearest power of 2 (2, 4, 8, 16, 32, etc.).

PPLNS pays better.  Period.  In any reasonable time frame, PPLNS wins.  Jumping to PPS based on luck is gambler's fallacy.  The ONLY reason you would use PPS over PPLNS is if you have some kind of *daily* bills that you can only pay via BTC, meaning you need 100% reliable earnings on a daily basis with no variance.


As for your second question:  If you're on PPS you do not have shares in shifts.

If your concern isn't about Bitcoin mining, this should be in the Altcoin forums.  This is the BITCOIN -> Mining -> Pools section.

Sync'ing up multiple blockchains at the moment.  Also analyzing the code I had in place to live-add coins without a server restart to make sure when I restart the server this time I won't have to do it again for the next batch of coins.

Definitely needing more speed, sucks when we fall back to Litecoin due to the difficulty/block rewards of other coins not making them profitable (even if they normally bounce back within a minute).  But that still puts a lot of shares towards LTC over time, but with such a high difficulty, we can go a long time between those blocks.  Variance is a bitch when you're small.

EDIT:  Doge is kind of the same.  4 shifts since our last DOGE block due to bad luck (we've submitted enough shares to average 3 blocks).

Your first error was looking at Coinwarz.  Here's a tip:  EVERY coin estimate site is wrong.  Every, damn, one.  Because none of them are current on their chains, so they have the wrong difficulty for most newer coins (which adjust diff per block).  Additionally, they don't know what the reward on the next block is, so they use the "average" reward.




Shares are kept in the PPLNS system for 5 shifts (hours), so if we go back to a coin later and solve a block, your time spent previously impacts your rewards.




One thing people need to remember:  There is no concept of "progress" towards a block.  If the average time to find a block is 5 minutes for the pool, and we haven't found one in 10 minutes, that doesn't mean we're any "closer" to finding a block.  The average time is still 5 minutes.  As a result, there is *no* reason to remain on a coin if another coin is now more profitable.

ScryptGuild's coin switching is unique, in that ScryptGuild is on every coin at all times.  On the miner end, switching coins is no different from a new block on the network, with the exception of the old work *might* still be valid.  It means that we can switch coins without invalidating old work.  If your miner submits a share for a previous job after we switched coins, it can still be valid and still solve a block, as long as a new block hasn't been found on that coin yet (in which case it IS stale).

Here's the order things happen:

1) Shift closes (~1 minute after the hour)
2) Shift payouts are calculated/credited (~3 minutes after the hour)
3) Automatic Conversion executes (~10 minutes after the hour)
4) Automatic Payouts run (~20 minutes after the hour).


To "exclude" a coin from being converted, simply set your reserve to a very high amount.  Then set your automatic threshold to 0.01.  With this setup, every hour your balance will be sent to your wallet, and never automatically converted.

Prior to the profit switching update, new coins could be added seamlessly, without even a blip in the mining server.  However, the last time I added a coin it did cause a brief (~half second) outage.  I do know where the bug is, but it will take a server restart to actually fix (another half second outage).  My plan is to add a half dozen more coins to the system tomorrow, and hopefully fix it so future coins will not require restarts again.




LTC will take a while because I will not be using Cryptsy for it.  I almost did, then I noticed one HUGE problem:  Cryptsy's market depth on LTC is abysmal, to the point that cashing out just a few blocks would drop the price more than 10%.  So LTC conversion will be implemented later this week with BTC-e integration instead, since BTC-e has significantly more depth to absorb the sales without too much slippage.




Not much any pool can do about cryptsy API's failing.  The new auto convert feature is fault tolerant at least, so if the API fails it simply tries again in another hour.  As terrible as Cryptsy is, it's one of the only exchanges with a decent volume for most coins.  As the pool progresses, more exchange support will be added (BTC-e for LTC for example).  The real problem is each exchange has a different API, and some of them lack the ability to actually execute orders via API, making them completely useless.

Cute.  ~40% of the assets (2.8b JPY in bank balances, the easiest ones to notice missing) were unaccounted for?  This isn't just incompetence.  This is criminal negligence.

I mostly just report with "napkin math" figures.  I did bump the shift share count up slightly, but prior to my post:

Shifts were taking ~55 minutes, with a shift length of ~4 billion shares.  The translates to ~4.3 billion shares per hour.  Current network difficulty is ~3.8 billion.  The last block for the pool (at the time of my post) was 6 hours ago.  That means we had done approximately 25.8 billion shares, which is ~6.7 times the current network difficulty.


The even easier version (doesn't factor in the current shift):  Check how many open shifts currently have 0 blocks.  Multiply that by shift length (4 billion as of my previous post, 4.5 billion going forward), and compare that number to network diff.

At BTC Guild's current shift length and the network difficulty, the AVERAGE blocks for a shift (once closed) is between 10 and 11.  This means if the pool has a really long block, you can see 5-8 of the most recent open shifts with 0 blocks.

Right now we're at a little under 6 hours without a block.  Not exactly hard to believe, that's only ~6x difficulty, something which happens at least once a month (probably more).

Botnet and source of huge hashrate shouldn't be in the same sentence.  A single KNC Jupiter is faster than the majority of botnets.  Botnets haven't been a "source of huge hashrate" since mid-2013 when ASICs became widely available.  They've mostly moved on.  The botnets still around today are mostly ones that were spread in late 2012.

Avalons definitely produced higher stales than other hardware in my experience.  My Avalon averaged ~1.2-1.4% stale, while my ASICMINER USBs averaged ~0.2% on the same connection.  There's really not much that can be done about it, the problem lies in hardware/firmware.

How variance feels after dealing with it for 3 years.

Speed is only down a few hundred TH.  Nowhere near 1000.