Potential change to live profitability calculation

Hello,

Because of the recent developments with the X11 algorithm and some coins producing an extremely high profitability for a very short period I believe there may be a need to change how the live profitability is presented. The current system reflects the average value of the last 250 shares, which by definition treats all share values as equally likely, when in actuality they are not. It would be akin to averaging out the winnings of a lottery without taking into account the odds.

As an example, if over the last 250 shares 0.01% of the pool submitted shares worth 1000 cents and 99.99% submitted shares worth 15 cents, then I would do the following calculation:

0.0001*1000 + .9999 * 15 = ~15.1 cents

Similarly if only high value shares are being submitted the calculation would look as follows:

0.0001*1000 = 0.1 cents

This seems accurate to me as most of the pool did not submit any shares over the last 250 shares and as such the value they earned was nothing.

The above is the value that a connected miner would be expected to be worth over the last 250 shares. The way the profitability is presented now only provides what was submitted, which has nothing to do with what a connected miner would be expected to produce because the system does not connect miners to coins evenly (nor should it).

This metric doesn't necessarily need to take the place of the live profitability, but I believe it would be useful, especially for those of us using bots to determine whether we want to mine here or not. I would do this calculation myself, but I do not have access to the necessary data to do so.

Thoughts?

ryguy wrote:......It would be akin to averaging out the winnings of a lottery without taking into account the odds.
...
...
This metric doesn't necessarily need to take the place of the live profitability, but I believe it would be useful, especially for those of us using bots to determine whether we want to mine here or not. I would do this calculation myself, but I do not have access to the necessary data to do so.

Thoughts?

Does knowing what last week's winning numbers were help you win the following week?

You want to use the values from the previous 250 shares to determine if you should switch miners to mine on this pool in the future.
It's a live running tally. It's not a crystal ball. You can follow a rising trend and make a guess where it might continue to go, but it's just a guess.
If you are going to base a pool switching bot on the live profitability data, I think you will just be spending time swapping from pool to pool.

I think you would be better off looking at the CHART data that displayed what expected payout should have been for the previous 24 hours and use that to base your decision for the day. Looking back 250 shares or 24 hours, neither one will tell you what you will make the following day or the following 5 mins.

Even if they changed the algo as you suggested, suppose everyone on X-11 in the past 250 shares DID make 1000 cents each.
That would trigger your bot to come here, and 5 seconds later the data would update to go back to .01 cents each share.
How long before the bot recalculates to swap again? Whats the restart time? Is there a spool up of the hash each time it switches pools?
Are you loosing more in the hunt than if you stayed static? Lots of variables and things to consider but I'm not sure how increasing accuracy of a past event will affect future predictions, sorry, not seeing it still I guess.

Who cares?

Just tell people to connect their miners and be happy.

At the end of the day, you know how well the pool did.

Mlangford75 wrote:Who cares?

Just tell people to connect their miners and be happy.

At the end of the day, you know how well the pool did.

obviously, someone other than yourself cares.

CSZiggy wrote:

ryguy wrote:......It would be akin to averaging out the winnings of a lottery without taking into account the odds.
...
...
This metric doesn't necessarily need to take the place of the live profitability, but I believe it would be useful, especially for those of us using bots to determine whether we want to mine here or not. I would do this calculation myself, but I do not have access to the necessary data to do so.

Thoughts?

Does knowing what last week's winning numbers were help you win the following week?

Of course not, they're independent events (I had said mutually exclusive before, which isn't correct). I want to know the odds of winning, not the actual winning numbers.

CSZiggy wrote:
You want to use the values from the previous 250 shares to determine if you should switch miners to mine on this pool in the future.
It's a live running tally. It's not a crystal ball.

I'm not expecting it to be a crystal ball. I'm looking to determine what I would have earned had I been connected so that I can use that data to determine if the trends are holding in a way I believe will benefit me. Knowing that the last 250 shares produced earning of 1000 cents doesn't mean much if I don't know how likely I am to earn that if I were to be connected. That's the purpose of a weighting.

CSZiggy wrote:
Even if they changed the algo as you suggested, suppose everyone on X-11 in the past 250 shares DID make 1000 cents each.
That would trigger your bot to come here, and 5 seconds later the data would update to go back to .01 cents each share.

If last 250 shares were each worth 1000 cents and I have a 0.01% chance of being assigned to mine the coin with those shares, then I would end up calculating:

250 * 0.0001 * 1000 = 25 cents

but what I'm presenting is the totaled calculation of

0.0001 * 1000 = 0.1 cents (I'm assuming the last 250 shares total a value of 1000 cents. Sorry if that wasn't clear).

Now would I use a single calculation to determine whether to switch? Depends on my strategy.

CSZiggy wrote:
How long before the bot recalculates to swap again? Whats the restart time? Is there a spool up of the hash each time it switches pools?
Are you loosing more in the hunt than if you stayed static? Lots of variables and things to consider but I'm not sure how increasing accuracy of a past event will affect future predictions, sorry, not seeing it still I guess.

It's not increasing the accuracy of a past event, it's telling me what I should have expected in the past so I can determine if I should also expect that in the future. That's what trend data is. Will it give me 100% accuracy to what's going to happen? No, but it can be a useful guide.

EDIT FOR ADDITIONAL INFORMATION/EXAMPLE:

If someone tells you that the average winnings from the past lottery was $10K should you play? Hard to tell right? It could be that everyone that plays the lottery wins 10K, it could be one very large prize and a bunch of $0 or it could be something in between. However, if someone tells you that the expected return from having a ticket is $2 you have much better information on which to base your decision. I want the latter, not the former.

I think, that profitability must be calculated with percentage weight of assigned miners:
1% -> weight 0.01 -> 1% of assigned miners to coin_1
10% -> weight 0.1 -> 10% of assigned miners to coin_1
70% -> weight 0.7 -> 70% of assigned miners to coin_1 and so on

(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_1
(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_2
.
.
.
(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_250

(value_for_cycle_1 + value_for_cycle_2 + ... + value_for_cycle_250)/250 = final_profitability_value

balzsam wrote:I think, that profitability must be calculated with percentage weight of assigned miners:
1% -> weight 0.01 -> 1% of assigned miners to coin_1
10% -> weight 0.1 -> 10% of assigned miners to coin_1
70% -> weight 0.7 -> 70% of assigned miners to coin_1 and so on

(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_1
(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_2
.
.
.
(weight_1 * coin_1_value + weight_2 * coin_2_value + weight_3 * coin_3_value + .....) = value_for_cycle_250

(value_for_cycle_1 + value_for_cycle_2 + ... + value_for_cycle_250)/250 = final_profitability_value

Are you saying that's how you believe it currently is calculated or how it should be calculated?

It is my explanation how it should be calculated.
weights for pool can be calculated as:
weight1 = assigned_miners_to_coin1 / all_miners
weight2 = assigned_miners_to_coin2 / all_miners
... and so on
for all coins with assigned miners

The current profitability calculation is wrong and useless. It can be used only when no extreme values occures, but my calculation works always correctly

Or more exactly, weights for pool can be calculated as:
weight1 = assigned_hash_power_to_coin1 / overall_hash_power
weight2 = assigned_hash_power_to_coin2 / overall_hash_power
... and so on
for all coins with assigned hash power

balzsam wrote:The current profitability calculation is wrong and useless. It can be used only when no extreme values occures, but my calculation works always correctly

I'm pretty sure the calculations come right from the pool's data so I would wager they are NOT wrong.
They may be useless for your intended purposes, but as an average overall idea of what the data just was, I think it does a pretty good job.

I'm still a bit confused, didn't I read that nicehash contracts were for 24 hours and couldnt be changed or something?
So it's important to wait to start the clock going only when the profit is at its highest point? because there is no chance of it falling lower later on in the following 24 hours? If you are going to get a contract to mine for 24 hours, set it and forget it or don't.