In a recent idea, I attempted to predict the approximate date of this cycle's top by comparing the length of the prior two cycles, and using the rate of expansion to extrapolate the date for the current cycle. That method has two major flaws:
(1) It assumes cycles are lengthening at a linear rate (since I only compared the lengths of two cycles)
(2) Where I chose to start measuring "the bull run" is subjective
I've come up with a new model that attempts to resolve or mitigate these flaws.
To solve the first problem, I simply included a third cycle in the calculation (2011). This gives us a value for the rate of the rate of change. We're no longer assuming that this cycle's length will increase by a factor of x simply because the prior cycle's length increased by a factor of x. Instead, we can assume that this cycle's length will increase by some fraction of x, based on the rate that x has diminished over the last 3 cycles. It's the same concept as diminishing returns with price, but using cycle length as the variable. I also used this same method for predicting the correction/consolidation period after the current cycle's top.
The second problem is a bit trickier to solve. You could eliminate the subjectivity by simply measuring "the bull run" from cycle bottom to top. But bottoms are usually an overreaction to the prior cycle's blow off top, so the exact location of the bottom isn't necessarily the best point to measure from. Instead, I've drawn logarithmic regression bands through the cycle peaks and troughs, and measured the length from where price first touches the bottom band to where it touches the top band. Note that these bands were drawn manually with the curve drawing tool, so they're only a rough estimate (I couldn't figure out an easy way to plot a logarithmic regression using PineScripts, not sure if it's possible but may look into it more when I have the time).
Long story short, the new model projects 7/11/22 as the date that this cycle will top (which is a few months earlier than predicted by my previous model). We can also use the regression bands to extrapolate the price on 7/11/22 to approximately $95k. This is a bit lower than I had previously predicted, but that method shares the same flaw as #1 above (only comparing the prior two cycles).
While I think that a 129 week retracement/consolidation period is reasonable before next touching the bottom band, I think it's also quite likely that we will see a lower low during that 129 week period (i.e. $47.5k will NOT be the true bottom). How low will the true bottom be? I think my previous bottom prediction is still a pretty good bet, but that model also failed to account for problem #1 above. I'll update my prior models when I have some time (I'm actually quite curious how those top/bottom values will compare with the values from this logarithmic regression model).
It's also worth noting that these bands are squeezing more tightly together now, and will eventually break (probably sooner rather than later). As always, these predictions are only meant to be a rough estimate based on prior data. We need to continually evaluate technicals/fundamentals etc to fine-tune these values as more data become available.
(1) It assumes cycles are lengthening at a linear rate (since I only compared the lengths of two cycles)
(2) Where I chose to start measuring "the bull run" is subjective
I've come up with a new model that attempts to resolve or mitigate these flaws.
To solve the first problem, I simply included a third cycle in the calculation (2011). This gives us a value for the rate of the rate of change. We're no longer assuming that this cycle's length will increase by a factor of x simply because the prior cycle's length increased by a factor of x. Instead, we can assume that this cycle's length will increase by some fraction of x, based on the rate that x has diminished over the last 3 cycles. It's the same concept as diminishing returns with price, but using cycle length as the variable. I also used this same method for predicting the correction/consolidation period after the current cycle's top.
The second problem is a bit trickier to solve. You could eliminate the subjectivity by simply measuring "the bull run" from cycle bottom to top. But bottoms are usually an overreaction to the prior cycle's blow off top, so the exact location of the bottom isn't necessarily the best point to measure from. Instead, I've drawn logarithmic regression bands through the cycle peaks and troughs, and measured the length from where price first touches the bottom band to where it touches the top band. Note that these bands were drawn manually with the curve drawing tool, so they're only a rough estimate (I couldn't figure out an easy way to plot a logarithmic regression using PineScripts, not sure if it's possible but may look into it more when I have the time).
Long story short, the new model projects 7/11/22 as the date that this cycle will top (which is a few months earlier than predicted by my previous model). We can also use the regression bands to extrapolate the price on 7/11/22 to approximately $95k. This is a bit lower than I had previously predicted, but that method shares the same flaw as #1 above (only comparing the prior two cycles).
While I think that a 129 week retracement/consolidation period is reasonable before next touching the bottom band, I think it's also quite likely that we will see a lower low during that 129 week period (i.e. $47.5k will NOT be the true bottom). How low will the true bottom be? I think my previous bottom prediction is still a pretty good bet, but that model also failed to account for problem #1 above. I'll update my prior models when I have some time (I'm actually quite curious how those top/bottom values will compare with the values from this logarithmic regression model).
It's also worth noting that these bands are squeezing more tightly together now, and will eventually break (probably sooner rather than later). As always, these predictions are only meant to be a rough estimate based on prior data. We need to continually evaluate technicals/fundamentals etc to fine-tune these values as more data become available.
