Bitcoin Forum

Hello, I have 10+ years of experience in financial markets, especially statistical analysis of markets (https://en.wikipedia.org/wiki/Time_series), so I can offer my forecasting skills as a service to people who want to approximate the price of any market into the future.

I have also developed some free demo tools, written mostly in Python 3.5+, which is FREE that I give out for free so that you can check my claims by yourself. Whatever I will sell to you will be much better that that!

Free Tools:

• [DEMO] LINEAR CRYPTO FORECASTING TOOL 2.6 (https://pastebin.com/M2fGxNhi), read explanation (https://steemit.com/bitcoin/@profitgenerator/diy-full-linear-forecasting-tool-cryptocurrency-prediction)
• Data Processing & Correcting Tool for Blockchain.info's Data (https://pastebin.com/uRnX5S8Z), read explanation (https://steemit.com/programming/@profitgenerator/forecasting-adventures-4-processing-blockchain-info-s-data)

What you Need To Know:

First you need to know some things about my methods, I use LOG based absolute error measurements, not RMS (https://en.wikipedia.org/wiki/Root_mean_square) and bullshit like that, since that is heavily biased and I don't know why any statistician would use such things, as explained here (https://steemit.com/mathematics/@profitgenerator/variability-in-a-heteroskedastic-market).

So when you see things like 0.1 LN, that means an error rate of that relative to the natural logarithm (https://en.wikipedia.org/wiki/Natural_logarithm). Since logs are additive, we can also calculate an average for them. So it's much better to work with that. I call this LN Error.

I also have another benchmark which I call Predictive Edge, which is based on Theil’s U value (https://docs.oracle.com/cd/E17236_01/epm.1112/cb_statistical/frameset.htm?ch07s02s03s04.html). Basically it counts how much better our model is compared to random guessing. Anything over 0 is a theoretical edge over the market, so it should theoretically be profitable, giving a positive expectancy for a trader, and if it's below 0, then that model is not profitable.




Public Model:

• Explanation: https://steemit.com/bitcoin/@profitgenerator/forecasting-adventures-9-final-research
• Data: Daily BTC/USD Data (https://blockchain.info/charts/market-price?timespan=all)  + Daily Transaction Fees USD (https://blockchain.info/charts/transaction-fees-usd?timespan=all)
• Method: Transform the Price with the LN function, make the Regressor cumulative and then transform it by the LN function. Run the Model on the transformed data either backtest or forecast, and then transform back the fitted/forecasted values with the EXP() function.
• Model: ARIMAX(0,1,1){ LN(cumulative_transaction-fees-usd.csv)_1 } with constant
• Coefficients: You have to calculate them yourself from the historical data, since it varies by the data
• Estimated LN Error:               0.031060897872341 LN
• Estimated Predictive Edge:    0.012481538

Private Model:

Coming Soon, Stay tuned!

I am doing research now, later I will provide my forecasting values for sale.

There are some updates with my research, I made a tool to process the data from Blockchain.info:
* https://steemit.com/programming/@profitgenerator/forecasting-adventures-4-processing-blockchain-info-s-data

So now I have everything that I need to further analyze the data using traditional means, and then compare those results with the results the [DEMO] LINEAR CRYPTO FORECASTING TOOL 2.6 will give.

Then I will start working on my private model that I sell to people who will need forecasts. I can only do this as a hobby in my spare time so it might take a couple of weeks. Stay tuned.

The last public model is published. This is the last model I publish here for free, the rest of my research will not be published, instead if anyone wants to know the better models that I will discover in the future, just contact me and I can sell it. Or I can sell the daily or weekly forecasts to people who are interested. Just contact me.

The last model is currently the best one that I know, it has a fairly decent predictive edge, so in theory it should be profitable, it should in theory give an edge over the market, over the other traders.

Can you forecast the price of the bitcoin for the next 6 months, please? What would be the price for this forecast? I'm a physicist (not a statistician) and I'll try to learn from your article (at steemit) your methods. They seem very good, and if this could predict the price of the bitcoin in the next 6 months-one year this would be extraordiary. Mainly I am interested if the bitcoin price will crash after 10000-12000 or will continue to grow until 25000-30000 in the next two years.

I haven't ended my research yet, it's still in progress, but I doubt it can accurately forecast 6 months ahead.

I will use daily or weekly forecasts instead when I am ready. Stay tuned.