Let me state very clearly that I think bitcoin are without any real value and that I do NOT recommend anyone to buy them. The following is for instruction only and any trading that anyone does is entirely at their own risk. You have been warned.
I did an analysis of the log of bitcoin prices in $US. (logs make equal percentage changes equal)
I used CATS software which is free and available from Cycles Research Institute, see:
https://cyclesresearchinstitute.org/cats/
Shown below are graphs of various outputs from CATS.
====
Log Bitcoin prices in US$
====
The most significant cycles found with CATS The first two periods are not significant because they each have less than two cycles.
The following three graphics show the logprice wrapped to various periods. The wrap operation shows the average shape of the cycle over the entire graph assuming the exact cycle period specified. Three of the slightly longer cycles have been used. The output of CATS is the columns A to E. G and H have been calculated to show the likely accuracy of the period, and the date in years of a maximum (you can add multiples of the period to get subsequent estimates of maxima dates.
====
20.712 days wrap
====
16.934 days wrap
====
14.517 days wrap
Bitcoin Cycles

 Posts: 7
 Joined: Wed Apr 21, 2021 10:25 am
Re: Bitcoin Cycles
Hi Ray,
I'm not familiar with the outputs of CATS. Is there anywhere you can point me to that explains what I am looking at?
Marc
I'm not familiar with the outputs of CATS. Is there anywhere you can point me to that explains what I am looking at?
Marc
Re: Bitcoin Cycles
Hi Marc, you asked: "I'm not familiar with the outputs of CATS. Is there anywhere you can point me to that explains what I am looking at?"
The first graph is simply the log of bitcoin prices plotted against date and I think that is clear enough.
The table is a spreadsheet of values that is output by CATS whenever a spectrum is done. I have added the headings and the two right hand columns.
The no.of cycles is how many cycles were detected over the period of the data. In my experience, CATS usually gets this to an accuracy of about 0.1 cycle (which incidentally beats quantum mechanics by a factor of 10, but that is another story).
So the period (in this case in days) is the total number of days of data divided by the number of cycles from column 1.
The date of maximum is the number of days from a fixed point in time (near 1 BC) of one of the maxima of the cycle. The dateyears column is the year and decimal of a year of a maximum, It is the date of maximum divided by 365.2425 which is the average number of days in a year. year with .0000 is the start of the year.
The amplitude of the cycle is in units of the values for which the spectrum was done. In this case because we were using log prices, these values are the variation of the log. So the cycle of 8.346 days has an amplitude of 0.002432 which means that it varies by 0.56% above and below the mean tend (because log(1.0056)is 0.002432).
The P value of Bartell's test says how significant the cycle is. The first two lines have p=0 because the Bartell test cannot be done with less than 2 cycles. So ignore these in terms of significance. For the 8.346 day cycle it has p=0.00000118 which means such a tight cycle would not occur by chance much more than 1 time in 100,000. It is a real cycle. Often in statistics they use p<0.005, but with cycles we test many cycles, so that sort of value would let through some not real cycles. I have just listed p<0.001 and so these are probably nearly all real cycles. Yes there are lots. There are many clustered around 5.3 to 5.4 days and so something is going on there. Multiple values means that there is modulation of some underlying cycle. However the 8.346 day cycle appears to be a single stable cycle period.
The other three graphs show what I call a wrap of some cycles periods. The graph of the original data (the first graph on that page) is effectively chopped into sections of 8.346 days and assembled under each other. The the average shape is determined and plotted. Fractional days are allowed for and apportioned to the actual days either side. The result is shown at the first cycle date,so at the start of the data. You can add any multiples of the wrap period to get later dates. Or use the date of maxima from the table before.
For more info on CATS and free download, go to https://cyclesresearchinstitute.org/cats/
The first graph is simply the log of bitcoin prices plotted against date and I think that is clear enough.
The table is a spreadsheet of values that is output by CATS whenever a spectrum is done. I have added the headings and the two right hand columns.
The no.of cycles is how many cycles were detected over the period of the data. In my experience, CATS usually gets this to an accuracy of about 0.1 cycle (which incidentally beats quantum mechanics by a factor of 10, but that is another story).
So the period (in this case in days) is the total number of days of data divided by the number of cycles from column 1.
The date of maximum is the number of days from a fixed point in time (near 1 BC) of one of the maxima of the cycle. The dateyears column is the year and decimal of a year of a maximum, It is the date of maximum divided by 365.2425 which is the average number of days in a year. year with .0000 is the start of the year.
The amplitude of the cycle is in units of the values for which the spectrum was done. In this case because we were using log prices, these values are the variation of the log. So the cycle of 8.346 days has an amplitude of 0.002432 which means that it varies by 0.56% above and below the mean tend (because log(1.0056)is 0.002432).
The P value of Bartell's test says how significant the cycle is. The first two lines have p=0 because the Bartell test cannot be done with less than 2 cycles. So ignore these in terms of significance. For the 8.346 day cycle it has p=0.00000118 which means such a tight cycle would not occur by chance much more than 1 time in 100,000. It is a real cycle. Often in statistics they use p<0.005, but with cycles we test many cycles, so that sort of value would let through some not real cycles. I have just listed p<0.001 and so these are probably nearly all real cycles. Yes there are lots. There are many clustered around 5.3 to 5.4 days and so something is going on there. Multiple values means that there is modulation of some underlying cycle. However the 8.346 day cycle appears to be a single stable cycle period.
The other three graphs show what I call a wrap of some cycles periods. The graph of the original data (the first graph on that page) is effectively chopped into sections of 8.346 days and assembled under each other. The the average shape is determined and plotted. Fractional days are allowed for and apportioned to the actual days either side. The result is shown at the first cycle date,so at the start of the data. You can add any multiples of the wrap period to get later dates. Or use the date of maxima from the table before.
For more info on CATS and free download, go to https://cyclesresearchinstitute.org/cats/

 Posts: 10
 Joined: Fri Jul 02, 2021 9:32 pm
Re: Bitcoin Cycles
Thanks for your work on this Ray.. and appreciate your ongoing contributions to FSC and to the study of Cycles. Similar to Marc i imagine, (as i have also little experience with CATS) I was just wondering how on how we can take this data and turn it into something which has forecasting ability. I would love to figure out how to take this work and take it a step further. I realize you have your own cycles software that you're familiar and comfortable with. I was wondering your thoughts on the FSC's Cycle Finder APP or ways which we could better make use of what you have done? Thanks, Mark
Re: Bitcoin Cycles
Hi Mark
Thanks for your kind comments.
CATS is designed for studying cycles, and it has the ability to do forecasts, but its ease of use for that is not as great as the FSC app. OTOH I think it would be ultimately possible to get better results because it has greater accuracy of cycles length determinations. If you take the various short cycles that I did the "wrap" for, you can take those graphs and add multiples of the cycle length to get projections and add them all up. You can automate this within CATS once you get familiar with it.
However for short term projections, Lars has demonstrated that using a short period to project from is a reasonable approach. I recommend watching his recent video on bitcoin forecasting and see if it is useful for you. https://www.youtube.com/watch?v=GMMz52t4U2M
Regards
Ray
Thanks for your kind comments.
CATS is designed for studying cycles, and it has the ability to do forecasts, but its ease of use for that is not as great as the FSC app. OTOH I think it would be ultimately possible to get better results because it has greater accuracy of cycles length determinations. If you take the various short cycles that I did the "wrap" for, you can take those graphs and add multiples of the cycle length to get projections and add them all up. You can automate this within CATS once you get familiar with it.
However for short term projections, Lars has demonstrated that using a short period to project from is a reasonable approach. I recommend watching his recent video on bitcoin forecasting and see if it is useful for you. https://www.youtube.com/watch?v=GMMz52t4U2M
Regards
Ray