Topic: [ANN][Pool] HashBros Private Alpha [Multi][Profit Switch][Transparent] - page 21. (Read 18993 times)

Hash Bros ‏@hashbrospool 20m

First 25 alpha users are being enabled now.

We also just added a uservoice active in the bottom right hand corner of every page. If you want to check it out and give us feedback directly check it out here : https://hashbros.uservoice.com/

Just hit 330 sign ups. Pretty amazing guys. We will be rolling out the alpha to users tonight.

Alpha starting tomorrow night hopefully.

How long does it take to approve someone?  I signed up earlier today and no email confirmation.

I would but there are some final things that need to be worked out. Tomorrow doe.

You should go ahead and approve mine

brb talking to my finance buddy. but seriously very insightful. I imagined developing some sort of algorithm thats better than the most profitable at the moment but this is something else.

You're going to want to create some system of hysteresis for profit switching that takes a number of factors into account. For instance, HashCows offen jumps to coins which are *slightly* more profitable than the current coin, only to jump back a minute later with no gains to show for the switch, when it would be more profitable to simply stay on the current coin. Properly filtering this will need to bias the coin choices based on past history. Some other things to consider include:

- Price at multiple exchanges
- Time of exchange confirms (cryptsy can take hours, and end up trading the coins at a much lower rate)
- Average reject rate for users (for instance, FastCoin = high reject rate on my rig)
- Average profitability of the coin over time
- Coin parameters (finishing blocks vs switching now, etc)

Likely scenario would be to score all of these factors as a low-pass filtered average over several time windows, then assign weights to them for each time window and finally weight them against each other. Rather than determining the weight manually, you should consider a system that evaluates different weightings and finds the ideal one.

For instance, imagine we construct min/average/max profitability for each coin for various time windows (say, 20 minutes, hourly, 6 hour, days, etc, based purely on exchange value). Assuming we're using cryptsy, we can look at the average time between transfer and trade to determine a possible variance on price over that time (based on our min value during a time period matching the average trade time for cryptsy). This gives us a reasonable projection of price and variance by the time we get to the exchange. Factor in the historical reject rate for the coin (averaged by users), and perhaps the slope of the coin over a long time window (for instance, doge has been going down for days), and you get a pretty decent score for your coin. More importantly, it's unlikely that a small spike in the data (single bogus trade, etc) is going to cause the system to thrash over to the wrong coin.

Anyway, you get the idea..

Great discussion going on reddit : http://www.reddit.com/r/litecoinmining/comments/1ua4d6/hashbros_multicoin_pool_with_profit_switch_alpha/

I haven't given this thought but this is actually a fantastic idea.

I think what we can do when making such a decision is look at the average time it took find a block and see if its worth staying to finish till the next block.

ofeefee++

Signed up

How do you handle the work that's already been spent in a round and you jump to the next coin?  Will you finish that block out before jumping?

Thanks. Also if anyone has any feature requests that may be a bit of a reach we'd love to hear them. Something that you've never seen other pools do.

Signed up!  

Another HashBro engineer here!

I'm super excited to launch the private alpha which will set the bar for a truly advanced and transparent pool.

Thanks for all the signups so far, I'm positive you will like what we've been cooking up!

Thanks for registering up guys. We've already got 20 sign ups. Let's keep it coming.

HashBro here checking in.

You can reach to me either @hashbrospool, travisby in #digitalcoin (on Freenode), or here.  I would love to hear suggestions on our profit switching abilities.

One suggestion to decrease downtime is to set scantime via your miner.  I have mine set to 5 seconds, and the downtime between coin switching is only five seconds now.  Our backend implementation is an instantaneous switch; the bottleneck is in the client.

Hash On Bros

I'm deploying code regularly which makes the site unresponsive at times. It should be up.

Right now we are using coinchoose to determine what coin to mine. For the most part its reliable but I have been thinking about developing my own version of coinchoose because it lacks coins. For profit switching its pretty basic to calculate whats profitable to mine. As long as you have an exchange like cryptsy to bounce prices off of. What's hard for most to do is to adapt for new coins. Thankfully we can support new coins in under 15 mins.

If you simply mean the coin that is being mine, it will be displayed 24/7

signed up

Thanks