Hacker News Reader

Comment by ghshephard | Hacker News Reader

Would anyone like to translate this? I'm reasonably familiar with bitcoin mechanics, but I couldn't understand what had occurred based on this post.

View on HN

Normally, Bitcoin mining has fairly high variance - you only get paid when you find a block, and because blocks are relatively rare and found at random intervals your income varies a lot too. MtRed were what's known as a PPS pool, which meant they basically ate the variance and paid miners based on the amount of mining work they did.

The trouble is that running a PPS pool requires a huge buffer of your own funds to cover payouts during periods of bad luck, and if the buffer isn't big enough the pool will inevitably go bankrupt. It appears this has now happened to MtRed and it sounds like they only have enough funds to pay out half the money they already owe miners. (Technically, over a long enough period of time PPS pools will go bankrupt with 100% certainty no matter how big the pool buffer.)

As I've rapidly discovered, trying to calculate aggregate stats over the basis of Poisson arrivals... is a pain. To work out anything interesting, you end up studying piecewise-deterministic markov processes in more detail than you'd wish.

Would adding a fee allow them to continue?

What if they only paid out when a block was found?

Adding a fee might have helped if they'd done it from the start, but it's too late now. Sticking with PPS but only paying out when a block was found also wouldn't help; Eligius tried that and as soon as they hit a long period of negative luck everyone stopped mining there and the pool had to be restarted with a different payout method, screwing everyone who'd mined there previously in the process. (That issue with Eligius's payout method was actually predicted well before it actually happened.)

> What if they only paid out when a block was found?

Then the miners would have no impetus to help them and would instead mine for themselves

I'm not sure this is true. For example, consider the following scenario at $200 USD per bitcoin: 5 people pooled mining capabilities, and mined a single bitcoin. They could then split that, and effectively pocket $40/ea (or 0.20 bitcoins each). If you mine solo, is it possible to mine a partial bitcoin, rather than increments of a whole bitcoin?

Right: mining produces whole bitcoins, and does it infrequently. That's why pools exist; people weren't willing to wait months to randomly win the bitcoin lottery. (Maybe that's a design flaw!) Most pools pay proportions of their actual winnings, not their predicted ones.

A mining pool collects work from miners who are too small to mine on their own. In Bitcoin today, this is basically every miner (other than the bigtime dudes who already got ASICs). Time to complete a block is in the years for GPU miners. Put together hundreds of miners at varying rates and collect their work into a pool, now you can afford to pay small miners something even if their work would not approach a single block in the amount of time that you see them.

A negative buffer means they've been paying per share the going rate per share (however they've calculated that), and not some other way of calculating that avoids negative buffers (say, dynamically calculating the value of the shares based on how many buffers they touched and how much is in the buffer.)

They have missed their projections for too long and they are closing their doors to avoid paying out of pocket, rather than change to a scheme that does not depend so much on the pool's luck.

Reply on news.ycombinator.com