[ANN][Pool][Profit-Switch][Optional Auto-Exchange per Coin][Vardiff] ~ Hashcows

[Gaywallet]

Is the algorithm set up to not switch until a block has been mined? Until it reaches a certain % over the usual? How does it work?

I noticed a lot of rounds that found 0 blocks, if its switching before a block is finished often, that's just wasted hash rate.

My assumption is that this was just some hiccups, but can you verify?

Can I get a response to these questions?

I see an awful lot of rounds that are highlighted in green and/or have no payout listed. That combined with a very low payout so far have me second guessing.

We check all coins profitability every few minutes, we have a base minimum time per round that we hit no matter what (because there is a small cost of lost mining time every switch), after that minimum time there are various rules to determine whether we should switch. Generally speaking, one coin has to be a decent amount more profitable than the current, to switch within X amount of time.

From time to time, we get unlucky and get no blocks but that's just the card we were dealt, so to speak

We can always increase the "base minimum time per round" so if that's something that will increase payouts we can do it.

At the absolute minimum you should never stop mining until you hit a block (exceptions at the end).

Think of it this way: Coin A takes an average of 10 minutes per block and at current trading rates is 200% more profitable than some set coin, such as litecoin.

Coin B takes an average of 20 minutes per block and is currently trading at rates of 150% that same set coin.

So you start mining A. 5 minutes in, coin B becomes more profitable.

I propose to continue mining until you get a block of coin A. If the average is 10 minutes and the minimum is 0 minutes, just about 100% of the blocks will fall between 10 and 20 minutes (yes I know the math is more complicated than that). 99.7% of the data will fall within 3 standard deviations above/below the average, so we can infer the standard deviation is at most 3.33 minutes. This also means that 95% of the blocks mined will fall between 3.33 and 16.66 minutes. Realistically, the normal distribution will be much closer to the average.

Worse case scenario, is that it takes 20 minutes to mine a block. If it takes 20 minutes, we spend 15 minutes mining a 'less profitable' coin. How profitable does the new 'more profitable' coin have to be to outweigh this difference? Well, assuming that we will mine a block of the more profitable coin at the average amount of time (20 minutes), we have the following equation:

200% * (15 minutes / 15 minutes for a block) =  B% * (15 minutes / 20 minutes for a block)

B% = 266.667%. B must be 266.667% more profitable than our coin setpoint.

However, 95% of the time, this block will mine in 16.66 minutes or less. The equation then works out to be B% = 344.82%. In this case, B needs to be almost twice as profitable as A to expect rewards from switching.

This is assuming a huge amount of variation in block size, which is unlikely. For blocks that take an average of 10 minutes, my guess is that something like 95% will fall between 6 and 14 minutes.

If you set something like litecoin or bitcoin as a setpoint, your profitability calculations to determine when to switch and when to stay should be as follows:

(current time spent mining A + average time for block of B / average time required for block A) * profitability vs. setpoint > block B profitability vs. setpoint.

The moment that > becomes < is when you should switch, even if time has been spent on coin A.

This is because on average, we should be looking at expected blocks of A vs. profitability of A as compared to B, given a switch, assuming average expected blocks of B. In fact, when we do this, we find the profitability requirements to switch are much, much higher. In fact, if you do the math of expected block time, coin B has to be 2.5 times more profitable than A to justify a switch after 5 minutes of coin A.

Of course, once a block is mined, it would be time to switch to the most profitable coin.

[coyote]

Is the algorithm set up to not switch until a block has been mined? Until it reaches a certain % over the usual? How does it work?

I noticed a lot of rounds that found 0 blocks, if its switching before a block is finished often, that's just wasted hash rate.

My assumption is that this was just some hiccups, but can you verify?

Can I get a response to these questions?

I see an awful lot of rounds that are highlighted in green and/or have no payout listed. That combined with a very low payout so far have me second guessing.

We check all coins profitability every few minutes, we have a base minimum time per round that we hit no matter what (because there is a small cost of lost mining time every switch), after that minimum time there are various rules to determine whether we should switch. Generally speaking, one coin has to be a decent amount more profitable than the current, to switch within X amount of time.

From time to time, we get unlucky and get no blocks but that's just the card we were dealt, so to speak

We can always increase the "base minimum time per round" so if that's something that will increase payouts we can do it.

At the absolute minimum you should never stop mining until you hit a block (exceptions at the end).

Think of it this way: Coin A takes an average of 10 minutes per block and at current trading rates is 200% more profitable than some set coin, such as litecoin.

Coin B takes an average of 20 minutes per block and is currently trading at rates of 150% that same set coin.

So you start mining A. 5 minutes in, coin B becomes more profitable.

I propose to continue mining until you get a block of coin A. If the average is 10 minutes and the minimum is 0 minutes, just about 100% of the blocks will fall between 10 and 20 minutes (yes I know the math is more complicated than that). 99.7% of the data will fall within 3 standard deviations above/below the average, so we can infer the standard deviation is at most 3.33 minutes. This also means that 95% of the blocks mined will fall between 3.33 and 16.66 minutes. Realistically, the normal distribution will be much closer to the average.

Worse case scenario, is that it takes 20 minutes to mine a block. If it takes 20 minutes, we spend 15 minutes mining a 'less profitable' coin. How profitable does the new 'more profitable' coin have to be to outweigh this difference? Well, assuming that we will mine a block of the more profitable coin at the average amount of time (20 minutes), we have the following equation:

200% * (15 minutes / 15 minutes for a block) =  B% * (15 minutes / 20 minutes for a block)

B% = 266.667%. B must be 266.667% more profitable than our coin setpoint.

However, 95% of the time, this block will mine in 16.66 minutes or less. The equation then works out to be B% = 344.82%. In this case, B needs to be almost twice as profitable as A to expect rewards from switching.

This is assuming a huge amount of variation in block size, which is unlikely. For blocks that take an average of 10 minutes, my guess is that something like 95% will fall between 6 and 14 minutes.

If you set something like litecoin or bitcoin as a setpoint, your profitability calculations to determine when to switch and when to stay should be as follows:

(current time spent mining A + average time for block of B / average time required for block A) * profitability vs. setpoint > average time for block B * profitability vs. setpoint.

The moment that > becomes < is when you should switch, even if time has been spent on coin A.

This is because on average, we should be looking at expected blocks of A vs. profitability of A as compared to B, given a switch, assuming average expected blocks of B. In fact, when we do this, we find the profitability requirements to switch are much, much higher.

Of course, once a block is mined, it would be time to switch to another coin.

I don't think that fruitless rounds are a significant problem at the moment but I think you raise an interesting general suggestion that is worth considering.

Make the check for profitability, and change if appropriate, upon the discovery of a block or after a longer period of time.

I believe this pool has a great framework with unlocked potential profit that can hopefully be found by tinkering with the rules and carrying out some structured testing.

Keep up the good work

[matsonj]

Okay so I did some calculations based on just the past 30 rounds for today. We all know the estimations are hosed but I wasn't aware how much they were hosed. My estimated BTC earnings for today is currently sitting at 0.01455106. I have mined 0.144726 in the past 30 rounds which includes the past 14:35:31. I'm estimating a payout around 0.23803709 BTC. Now this is of course taking into the fact that we have been having some VERY good rounds but, my calculations have averaged what I make per second and then I just extrapolate. This estimation will get more accurate when we get closer to the 24 hr mark but for now, is still wildly inaccurate (as is to be expected, statistically speaking).

To help everyone understand what kind of payout we can expect, I'm currently running 5 cards for a total of 3.1 MH/s which means I expect 0.076786158 BTC per MH/s. Very good day indeed! I hope this helps everyone with estimated payout questions for today.

Hash cows to the moon!

EDIT: Since doing the calculations and rechecking my estimated earnings on the statistics page, my payout has been reduced dramatically even though the round was paid in full before and after my calculations. Is there something amiss in the table as well?

Changed for me also, even though the round had been paid in full. It told me for one round .316601 BTC now for that same round it's telling me .003268

 Of course the numbers change. First it is estimating based on what the scripts HOPE to sell the coins you just mined for. The revised numbers probably reflect what they ACTUALLY were able to sell for.  Watch out later today when all the WDC estimates drop 20% to reflect wdc only sells for .0008 not .001 like cryptsy has been claiming all day.
 No script can KNOW how much you are going to get, until the coins are sold, thats why estimates are estimates

The explanation makes sense but what doesn't make sense is how the rounds are labeled. The label is "Rounds Paid in Full" not "What We Hope the Round Will Pay." The label is misleading to say the least. Paid in full sounds like it's been taken to the market, sold and paid out to us. I suggest wording that more closely resembles what is actually happening..."Rounds to Be Sold" and "Rounds Sold and Distributed." This will make it much clearer what is happening.

[PrintMule]

How fast are cah out requests processed? I've been waiting on mine for some time now.

[Gaywallet]

I don't think that fruitless rounds are a significant problem at the moment but I think you raise an interesting general suggestion that is worth considering.

Make the check for profitability, and change if appropriate, upon the discovery of a block or after a longer period of time.

I believe this pool has a great framework with unlocked potential profit that can hopefully be found by tinkering with the rules and carrying out some structured testing.

Keep up the good work

There are a ton of round that say 0 blocks found right now, according to the tracker.

The dev response sounds like this happens fairly often. It should almost never happen.

Just providing the math as to why its almost always pointless to change mid-block.

[nwfella]

Things starting to look bleak once again :/

[aTriz]

How fast are cah out requests processed? I've been waiting on mine for some time now.

A BTC cashout? It should be within a few minutes, Send me a message with you're username and wallet address.

Things starting to look bleak once again :/

Trades are once again backed up due to exchanges.

[nwfella]

Exchanges backed up :/  Sounds highly plausible given my experiences with them lately...hanging in there in the hopes that things improve soon

[jerrybusey]

As far as round round switching goes I would hope the major consideration isn't how long has been spent on the current coin, but instead how long the difficulty will stay at its current level on the next coin. When I'm manually pool jumping I always try to check what the current block is in relation to the next difficulty adjustment and average block time so I can figure out if I'm going to see enough rounds for the switch to be worthwhile. If the profit differential is guaranteed to be high enough for long enough to make a switch worthwhile then the switch should be made, regardless of how long has been spent on the current coin. That's the optimal move anyway. I've found it can get a little annoying with some of the coins that have unusual difficulty changing systems (e.g. gradual change each round limited by a factor of x which I can never seem to find published, grr) but that's easy to model around.

[TheQuin]

I don't think that fruitless rounds are a significant problem at the moment but I think you raise an interesting general suggestion that is worth considering.

Make the check for profitability, and change if appropriate, upon the discovery of a block or after a longer period of time.

I believe this pool has a great framework with unlocked potential profit that can hopefully be found by tinkering with the rules and carrying out some structured testing.

Keep up the good work

There are a ton of round that say 0 blocks found right now, according to the tracker.

The dev response sounds like this happens fairly often. It should almost never happen.

Just providing the math as to why its almost always pointless to change mid-block.

Wrong, just plain wrong.

It is most profitable in the long run to always mine the most profitable coin. Variance will always even out in the long run, you just need to understand probability. Start by reading this thread from the beginning and also the multipool.us thread where it has been addresses many times before.

[Tempered81]

There should be a trollbox right there on the pools homepage so that peole can shout, "hey switch to such and such coin" and the admins can just do a quick swap...

[2RoyalSox]

I'm still really confused about the estimated daily earnings...I check it earlier and it said .023, but then I got a payout for .067?

Estimated payouts seem incredibly inaccurate compared to the final payout. I'd much rather see a this is how many of each coin you personally have mined type thing, so you can self estimate it against say cryptsy prices to figure out how much you will likely have once the pool sells the coins for you.

[aTriz]

The estimate was completely screwed up today because of exchanges. Nothing we can do, we send the coins to the exchange and we have to wait for them to pop up so we can trade them.

Of course all un-traded coins will be credited to tomorrow's payout.

[TheQuin]

The estimate was completely screwed up today because of exchanges. Nothing we can do, we send the coins to the exchange and we have to wait for them to pop up so we can trade them.

Of course all un-traded coins will be credited to tomorrow's payout.

My payout was 3x the estimate, which was nice.

It really would be a good idea to implement a running total of exchanged so far since the last payout. I know the Cows are busy, and you spend most of your time on keeping the profitability good, but please make it high up on the todo list.

[Epinnoia]

Guys,

Do u have any plan to add some new coins ?

Yes we are, we talked about a few prospects last night actually. If there is a coin you guys want me to lool at please send me a PM.

Why not make sure you're covering the top dozen or so coins at coinchoose?  I think the name of the coin is irrelevant.  The profitability is the all-important stat.  Sort it by the past week's performance, and get those into the system as fast as possible.

If you wait until half the world already knows about the coin before you add it, you're doing us no favors.  Profitability% over BTC...  That's all that matters.

[Epinnoia]

Should be evaluated on how much money can be exchanged for a large number of altcoins. Should not be taken for calculating the best bid on the exchange, as it may be just the bait.

I agree.  The depth table should be used, not just the current best price.  In fact, I think it would be better if it didn't look at anomalous trading prices at all.  It should look at the price at X number of exchanges, add them together, and divide by X.  Otherwise, you're entirely too vulnerable to people gaming the system.  Remember -- people can see what coins are being mined.  And they know there's a delay between mining and payout.  And they can just about predict when you're going to attempt to sell, and where.  Some of that can't be avoided, but it can be minimized.

It should also look at the depth table to ensure that trading is actually occurring at that price, as a sanity check.  If coins are NOT moving on the exchange(s), then almost certainly the price is nonsense.  

Better still, it should not use the reported current prices at all.  It should calculate an average price paid for the last XX coins of that type on that exchange in the last XX minutes.  The time to make sure the price is valid is not AFTER you have a whole bag full of them and are trying to sell them off.  Tiny sales (ie. 0.0000001) could be performed periodically (every few minutes) to ensure that the price is legitimate.

[B.T.Coin]

We'll look at doing a reduced fee promotion or something similar to compensate for the reduced mining time once I gauge the scope.   Apologies for the issues.

Any news on this?
Are things back to normal now?

Things are back to normal yes.

Things do look pretty normal again but before moving my hashes back to your site I'm interested to know if precautions have been taken so this can not happen again in the future. A pool going down is one thing, we have failover settings for that, but a pool accepting my work and not forwarding it to the network to claim the blocks is just a waste of electricity, time and valuable profits.

Did you gauge the actual scope of the losses already and will you do some kind of reduced fees to compensate?

[aTriz]

Guys,

Do u have any plan to add some new coins ?

Yes we are, we talked about a few prospects last night actually. If there is a coin you guys want me to lool at please send me a PM.

Why not make sure you're covering the top dozen or so coins at coinchoose?  I think the name of the coin is irrelevant.  The profitability is the all-important stat.  Sort it by the past week's performance, and get those into the system as fast as possible.

If you wait until half the world already knows about the coin before you add it, you're doing us no favors.  Profitability% over BTC...  That's all that matters.

I assure you we have looked at 90% of the coins out there, you can't just look at coinchoose and say this one would be the best to add, there are other factors to consider with us having so much hash. Low difficulty coins with FAST re-target time (every block etc) doesn't do anywhere near as good as the profitability states, we have tested it and it doesn't do well for the cows.

We added Stablecoin to the line-up last night looks to be a decent coin for payouts, we will continue to monitor it in the near future.

[CoinBuzz]

I think Auto-Trade credit is not credit to my account.

I'm mining for some days but my BTC balance is not changed and also in log i can't see any AT_Credit .

[nearmiss]

Should be evaluated on how much money can be exchanged for a large number of altcoins. Should not be taken for calculating the best bid on the exchange, as it may be just the bait.

I agree.  The depth table should be used, not just the current best price.  In fact, I think it would be better if it didn't look at anomalous trading prices at all.  It should look at the price at X number of exchanges, add them together, and divide by X.  Otherwise, you're entirely too vulnerable to people gaming the system.  Remember -- people can see what coins are being mined.  And they know there's a delay between mining and payout.  And they can just about predict when you're going to attempt to sell, and where.  Some of that can't be avoided, but it can be minimized.

It should also look at the depth table to ensure that trading is actually occurring at that price, as a sanity check.  If coins are NOT moving on the exchange(s), then almost certainly the price is nonsense.  

Better still, it should not use the reported current prices at all.  It should calculate an average price paid for the last XX coins of that type on that exchange in the last XX minutes.  The time to make sure the price is valid is not AFTER you have a whole bag full of them and are trying to sell them off.  Tiny sales (ie. 0.0000001) could be performed periodically (every few minutes) to ensure that the price is legitimate.

Depth of coins is taken into account, both when determining profitability, and when trading coins.  Your post does include an idea or two I may look at adding however. 

The backlog of current coins/rounds also has nothing to do with us mining a coin we shouldn't have been and now being stuck with it due to prices.  There was only 1 coin affected at the exchange end due to pricing (the rest were api issues), and our decided average price when we hit said coin were all within reason/market levels at that time.