Trend line prediction for Bitcoin difficulty in the next two months:
By definition, this is wrong - difficulty doesn't change per day or any particular time period, it changes per 2016 blocks.
Yes, it doesn't change per day. But if you take the change estimated on the 24th of Jan, we are only about 2% off. So you'll need to look at the estimations on the day the difficulty changes, and compare.
Part of difficulty forecasting is
when difficulty changes. If you can forecast the changes in difficulty, then you can forecast the dates when the changes occur and how much they are. Come on, don't be lazy! Do it right!
Fair enough. We put together these numbers to identify the difficulty in March, April, etc. So we could see if a mining rig would be profitable.
At least since November, the difficulty almost always changes between 11-13 days. How do you suggest we predict the *day* the difficulty changes? Should we simply estimate 12 days (historical average)?
No, it's trickier than that. The increase in difficulty is the inverse of the time taken for the previous difficulty divided by 86400*14. So if the increase in difficulty is say 120%, then the time between retargets immediately prior to the increase will be 86400*14/1.2 seconds. It'll write up something a bit more clear and post a link to it when I have some time.
But why not forecast cumulative earnings for the month? That's what I do for the most part. Difficulty prediction is tricky and tends to have unexpectedly wide confidence intervals past one or two retargets, but the cumulative earnings plot tend to be quite smooth and easy to trend-ify.
For example, if you plot date of deployment vs total income after three months, you get a very nice smooth chart that is easy to extend.
When you do write something up, please share!
Cumulative earnings require difficulty, don't they? Earnings require you to predict the number of minded coins * price. And of course the hashrate and reward, but it's fair to assume those are constant.
I'm using this formula: 24/(D*232/(H*10ˆ6)/60/60)*C
D= difficulty, H=hash, and C=reward
I have my forecasted commutative earnings, and actual (I monitor this day to day). My columns go something like this:
Date | Hashing Power | % Change | Predicted Difficulty | Actual Difficulty | Predicted BTC price | Actual BTC Price | Predicted Revenue | Actual Revenue
- that's not shipping).
i just seen a tweet...