Topic: NEW: Bitcoin Mining Calculator with Difficulty Change Adjusted Calculations (Read 17307 times)

LOL - I put in some numbers and every single miner currently available for sale and pre-order will never make ROI even with free electricity except for Cointerra IV miner - but that's assuming it is delivered on time.

I used 30% increment on difficulty.  Now it may not go up 30% every increment, but the last 2 have been over 30% and it's irrelevant if the last few cycles are near 30% if your devices is already down to mining satoshis.

I'm going to sit on my sit on the sidelines with my 300GH and wait for the dust to settle.  My 2011 GPU farm has already made more than coins than any ASIC could provide (short of shelling out 7 digits).

check out the calculator at http://www.vnbitcoin.org/bitcoincalculator.php

it is more sophisticated

http://www.vnbitcoin.org/bitcoincalculator.php

@KaptainBlaZzed: Your calculator agrees 100% with mine on it's calculations.

Nah.

What someone really needs to do is have a graphing function, and after about a months extrapolation it says something like this:

A good place to start is to look at the correlation between a moving average of price with Difficulty. I think you will find that a 10-12 week moving average will give you a fairly linear correlation. You can apply Pearson to give you an objective measure of the degree of correlation. http://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient

Once you have that you will find that Difficulty can be approximated by multiplying the moving average of price (the one you found to have a high degree of correaltion) with some factor. Let's say you use a factor that is in between all the ratios of Difficulty to the moving average of price going back as far as your data goes. You will find that these ratios fall on a more or less normal probability distribution. http://en.wikipedia.org/wiki/Normal_distribution So, you can characterize the likelihood that the actual Difficulty will fall so far from your projection.

There are several methods you can use from there to dial in results that give you less variance and a more reliable outcome. One method I use is to max out the degree to which the moving average of price is correlated with Difficulty by applying a custom weighting to the average. You could try to go for accuracy and optimize it by minimum average deviation. It will occasionally give you some spectacularly close hits, but this is statistically fragile. That is, it can produce more outliers. So, what I prefer to do is to minimize the expected maximum deviation. This method specifically targets outliers to help ensure your projection falls within a certain range of probability. It won't always give you the most accurate result, but it will give you more reliable results with in an advertised range.

All this can give you pretty reliable results for a projection that is one re-target out, like I have been getting. But the real trick is to try and figure out what Difficulty is likely to be going further out. This is because future Difficulty is highly dependent on what the exchange rate is going to be. What I have been experimenting with to try and go out farther with the projection, without trying to determine what price is going to be, is to simply drop terms from the moving average going forward. But the shorter the average gets the less correlated it gets and the more variance it produces. So I have to use a bunch of tricks to try and keep the variance down. One of those is to apply a weighting to the truncated average, another is to optimize it to reduce outliers as described above. Yet another is to eliminate outliers by discarding earlier data samples that do not correlate well. But this is a slippery slope, because it can make your data more statistically fragile and actually produce more outliers.

The next step, which I am still working on, is to come up with meaningful way to do actual price projections. And more importantly, to attribute a meaningful probability distribution to price projections. This is a work in progress.

I wish there were some kind of simple formula to use but as soon as you put something like that together it breaks. I continually re-evaluate and update the models I use as more data comes in.

Take a look at this.
https://forum.bitcoin.org/index.php?topic=13339.msg183351#msg183351

The thing is, it is not like figuring out an exact Difficulty in the future, but a probability distribution. This is what you need to determine risk vs. reward for an investment. Figure out what your risk tolerance is and then decide how much you are willing to risk.

It is not OK to do this: "Price and Difficulty will be X two months from now so I can calculate my return from that."

But it is OK to do this: "There is a 50% probability that Price and Difficulty will be above/below level X two months from now. That is/is not an acceptable risk to me."

I would gladly write that in if somebody would help me with the mathematical calculations to do this.

Anybody have an idea on how to do this?

Anybody have an idea on how to do this?

Looks good.  Now we need someone to write one that does do a conservative estimate of additional difficulty increases.  Difficulty increases aren't completely predictable, but writing an estimator that uses the exponential curve that best fits previous increases would be a good start.

Looks good.  Now we need someone to write one that does do a conservative estimate of additional difficulty increases.  Difficulty increases aren't completely predictable, but writing an estimator that uses the exponential curve that best fits previous increases would be a good start.