Be wary of retrospective/backwards-looking analyses because they're easily very manipulated when there are too few data points/cycles

TL;DR:

Past performance is no guarantee of future results. Correlation does not imply causation

Most crypto predictions are retroactive studies that are cherry-picked to look good with past data but have no accuracy in predicting the future. They're complete guesswork and have a high chance of eventually failing in future cycles.

Retrospective study: A backwards-looking study that considers past data

Prospective study: A forwards-looking study that establishes a hypothesis first and then checks whether future data supports the hypothesis.

The problem with retrospective studies

In general, prospective studies are always considered to be more trustworthy than retrospective studies of similar sample sizes. They are heavily affected by selection bias. There are exceptions for large-scale retroactive studies that have 10000+ data points that span decades.

In addition, many retrospective studies (especially in crypto) are completely manipulated because are not properly set up with a hypothesis ahead of time. Instead, they look for interesting patterns in past data until they find one. And then they establish a hypothesis AFTERWARDS to match that data. This is a huge taboo in statistics, and yet it happens all the time in crypto investing.

Using retrospective data to explain correlation without having a strong hypothesis created at the start of the study ends up looking like this XKCD comic:

https://xkcd.com/882/

Only the results that have an interesting narrative are published while the uninteresting ones are quietly discarded.

Too few crypto cycles for data points

Correlation does not imply causation, and past performance is no guarantee of future results.

We really only have 3 past price cycles (since the 1st epoch didn't have stable markets):

• Epoch 2: 2012-2016
• Epoch 3: 2016-2020
• Epoch 4: 2020-2024
• Epoch 5: 2024-

A line can fit any 2 data points perfectly while a curve can fit any 3 data points perfectly.

That means any logarithmic, exponential, or 2nd-degree polynomial can perfectly-fit 3 crypto cycles. It's so easy to fit any chart type because there are so few crypto cycles for data points.

Prior notable predictions that failed

• Rainbow Chart was plotted to perfect-fit previous cycles, and then it immediately failed before it lasted a full cycle and had to be revised.
• CBBI to perfect-fit previous cycles, and then it failed to last a single cycle because it didn't predict a double-peak
• It was said that no one who held BTC for over 4 years would have a loss. And then it was moved to 5 years, and then 6 years when each of those narratives failed in subsequent cycles.
• Bitcoin bull run starts 6 months BEFORE the halving is a false narrative that never happened
• Bitcoin bull run starts 6 months AFTER the halving is a false narrative that that only happened in Epoch 2
• Cycle dates are unpredictable:
  • BTC price hits a new ATH 5 days (Epoch 2), 145 days (Epoch 3), 85 days (Epoch 4) after the halving
  • BTC price hits a cycle peak 365 days (Epoch 2), 525 days (Epoch 3), 341/524 days (Epoch 4) after the halving
  • BTC large bull run starts -270 days and 225 days (Epoch 2, there are 2 pumps), -400 days (Epoch 3), 150 days (Epoch 4) after the halving

Just like before, many narratives will fail, and some will succeed. It's mostly just pure luck and guesses.

If you really want to establish a strong study, create a prospective study with a strong hypothesis, and then test it over 5+ epochs. Retroactive studies with few data points are just not trustworthy.