Currency idea: Block reward based on mining difficulty

[JohnDoe]

EDIT: New reward algorithm is nSubsidy = log2(difficulty)

This method of money creation to achieve a stable monetary policy has been suggested before but has been shot down quickly on the assumption that it would create hyperinflation. I agree that, because of technological progress, increasing difficulty by X amount will be cheaper in the future than it is now, but I propose that the relative effort of increasing difficulty by X% should have little variance over time. In other words, going from 1,000,000 difficulty to 2,000,000 at some point in the future will be as cheap as going from 1,000 to 2,000 right now. Thus the block reward could be a function of the difficulty growth rate to maintain stability.

[Bobnova]

That goes directly against the concept of putting out X coins per Y time.  In bitcoins case, 50 coins per 10 minutes.

[Etlase2]

"but I propose that the relative effort of increasing difficulty by X% should have little variance over time. In other words, going from 1,000,000 difficulty to 2,000,000 at some point in the future will be as cheap as going from 1,000 to 2,000 right now."

You are making a major assumption. It depends on what you want to achieve. Do you want more coins to be awarded as more people join the network? Because that is the primary force behind bitcoin's difficulty at this point. Down the road it will be improvements in hardware and so on, but it is not possible to account for both the number of computers and the power of those computers in one equation. At least not in the way bitcoin does it.

[Explodicle]

It's interesting, but will be very tricky to do properly. Assuming miners spend a constant amount on hardware over time, block rewards would go up 25% every 18 months, right? That's probably too much inflation even for a Keynesian currency. I'm assuming you want below 3% annual inflation, which means your payoff function should factor in % increase in the money supply too. You would have to get your econ math right the first time.

Good luck!

[Explodicle]

Moore's law. Roughly speaking, the same money will buy you twice as many hash/s in 18 months.

[Bobnova]

Ahh but the new $1000 card produces twice the hashes at roughly the same power consumption.

[ArtForz]

http://en.wikipedia.org/wiki/Koomey's_law

[ArtForz]

Doesn't sound too far-fetched, wouldn't be terribly surprised if the biggest single-GPU card ends up around 3.2-3.5Mh/J or so and a downclocked dual-GPU variant at least close to 4Mh/J.

[johnj]

From what I've read about the 7x series, it'll give (very roughly) 1.25x more mh/s, while only requiring half the power.

IIRC, 800 ALU @ 50w, 1600 ALU @ 90w (where as the 5770 is 800 @ ~100w and the 5870 is 1600 @ ~200w)

(i may have some of my numbers mixed up, but i believe that's the jist)

[Explodicle]

Most impressive. So I guess we should continue with nSubsidy = log(difficulty) for the purpose of discussion? Raise hands if you prefer logarithm.

(Raises hand)

[d.james]

what about: Coins mined during a higher difficulty is worth more than coins mined during lower difficulty.

[Bobnova]

6990 isn't 18 months newer than 5970.
Wait for the 7990.

[t3a]

This would change confirm times, but if confirm times were quick like with Geist Geld then it could work well.

Correct me if I'm wrong, but it would have to be a linear relationship between difficulty and coin/block rather than exponential like you have in that chart. Otherwise you would the number of blocks/(arbitrary time unit) go down exponentially.

[Explodicle]

what about: Coins mined during a higher difficulty is worth more than coins mined during lower difficulty.

That would make it difficult to price goods. More coins of equal value accomplishes the same thing, but you could be more certain how much they're worth.

[ElectricMucus]

I like the idea of a logarithmic inflation curve and thought about that myself.
You pretty much want to approach the natural logarithm with it and model the rewards to meet it.

If done right this could give more long term stability since economical development follows the natural logarithm.

[ElectricMucus]

I like the idea of a logarithmic inflation curve and thought about that myself.
You pretty much want to approach the natural logarithm with it and model the rewards to meet it.

What do you mean by "approaching" the natural logarithm? There's no difference in using base e, base 10 or base 2, the reward growth rate slows down at the same pace with any base.

Not exactly e is quite unique in this matter, it's not really about the growth rate but rather about the "divisibility" of the currency in regards to spending. Base 2 is the second best choice for a physical currency, best is 3 however in our case we can exploit the base e which for a physical currency is just a idealistic assumption but in our case a practical possibility.
see this article: (I have to admit I might not have fully understood it and this might be a wrong assumption, but you might look into it )
www.americanscientist.org/issues/pub/third-base

[t3a]

How would it change confirm times? nSubsidy and nTargetSpacing have nothing to do with each other.

If the value/block was increasing and the target spacing was staying the same then you would have exponentially more coins as the difficulty increased.

I assumed you would have the target spacing decrease so there was a constant number coins being produced.

[t3a]

Not exactly e is quite unique in this matter, it's not really about the growth rate but rather about the "divisibility" of the currency in regards to spending. Base 2 is the second best choice for a physical currency, best is 3 however in our case we can exploit the base e which for a physical currency is just a idealistic assumption but in our case a practical possibility.

That sounds very vague to me. How does base e make a cryptocurrency more divisible than base 2?

If the value/block was increasing and the target spacing was staying the same then you would have exponentially more coins as the difficulty increased.

I assumed you would have the target spacing decrease so there was a constant number coins being produced.

No, I'm not trying to make production either exponential or constant. If I wanted exponential I would use a linear relationship between reward and difficulty, like 1 coin per point of difficulty. If I wanted constant I would just make the reward 50 coins forever and be done with it.

Yes, that's what I'm saying.
If you have one block every ten minutes, then you have one block every ten minutes.

If the difficulty goes up, then you have more coins/block, but still one block every ten minutes.

This means that you either expect the difficulty to stay constant, or to go down. Otherwise you will have an exponentially increasing number of coins.

[Explodicle]

Not exactly e is quite unique in this matter, it's not really about the growth rate but rather about the "divisibility" of the currency in regards to spending. Base 2 is the second best choice for a physical currency, best is 3 however in our case we can exploit the base e which for a physical currency is just a idealistic assumption but in our case a practical possibility.

That sounds very vague to me. How does base e make a cryptocurrency more divisible than base 2?

If the value/block was increasing and the target spacing was staying the same then you would have exponentially more coins as the difficulty increased.

I assumed you would have the target spacing decrease so there was a constant number coins being produced.

No, I'm not trying to make production either exponential or constant. If I wanted exponential I would use a linear relationship between reward and difficulty, like 1 coin per point of difficulty. If I wanted constant I would just make the reward 50 coins forever and be done with it.

Yes, that's what I'm saying.
If you have one block every ten minutes, then you have one block every ten minutes.

If the difficulty goes up, then you have more coins/block, but still one block every ten minutes.

This means that you either expect the difficulty to stay constant, or to go down. Otherwise you will have an exponentially increasing number of coins.

The number of coins per block is f(x). Just because x goes up exponentially does not mean f(x) does. For example, if he was using f(x)=1/x, then the number of coins per block would approach zero.

[t3a]

Not exactly e is quite unique in this matter, it's not really about the growth rate but rather about the "divisibility" of the currency in regards to spending. Base 2 is the second best choice for a physical currency, best is 3 however in our case we can exploit the base e which for a physical currency is just a idealistic assumption but in our case a practical possibility.

That sounds very vague to me. How does base e make a cryptocurrency more divisible than base 2?

If the value/block was increasing and the target spacing was staying the same then you would have exponentially more coins as the difficulty increased.

I assumed you would have the target spacing decrease so there was a constant number coins being produced.

No, I'm not trying to make production either exponential or constant. If I wanted exponential I would use a linear relationship between reward and difficulty, like 1 coin per point of difficulty. If I wanted constant I would just make the reward 50 coins forever and be done with it.

Yes, that's what I'm saying.
If you have one block every ten minutes, then you have one block every ten minutes.

If the difficulty goes up, then you have more coins/block, but still one block every ten minutes.

This means that you either expect the difficulty to stay constant, or to go down. Otherwise you will have an exponentially increasing number of coins.

The number of coins per block is f(x). Just because x goes up exponentially does not mean f(x) does. For example, if he was using f(x)=1/x, then the number of coins per block would approach zero.

If that is true, his difficulty/reward chart is completely misleading.