The Lab

How we tested

• The S&P 500 monthly from 1985 and weekly from 1927, the Nasdaq-100 monthly, and Bitcoin daily from July 2010 to September 2026, all from public sources.
• The RSI(14): J. Welles Wilder's 1978 measure, from 0 to 100, of how much of the recent movement was up.
• Bitcoin's cycle tops and bottoms, by a rule fixed in advance: a top once the price has fallen 50% from it, a bottom once it has doubled from it, on monthly closes.
• The RSI's turns: a peak once the RSI has fallen 15 points from it, a low once it has risen 15 points.
• "Rolling over": the monthly RSI closing under 70 after closing at 70 or over. A weekly version and a divergence version were also tested.
• Controls:
  • random months in place of Bitcoin's real turns, 10,000 times;
  • the RSI's turns shifted in time, 110 ways.

The maths, step by step

1. The RSI. Average the recent rises and the recent falls, each smoothed so the latest move counts 1/14.

• RSI = 100 − 100 ÷ (1 + average rise ÷ average fall).
• Example (2-period, to keep it short): closes 10, 11, 10, 11. The first two moves average 0.5 up and 0.5 down, so the RSI is 50. The next rise of 1 smooths the averages to 0.75 up and 0.25 down: 100 − 100 ÷ (1 + 3) = 75.

2. When a turn is known. A peak is only a peak once the RSI has fallen 15 points from it. If it peaks at 88 in January and falls to 73 in March, the January peak was known in March. Any warning can only start then.

3. The lag. For each Bitcoin turn, find the nearest RSI turn of the same kind and count the months between them. Average the gaps, ignoring which came first.

What we found

1. The video's numbers.

• The highest monthly RSI in Bitcoin's lifetime was 87.7, in January 2018. That was the month after Bitcoin's December 2017 top, not before its run.
• The 2022 low was 42.8, in September 2022, not 62.

2. The two do turn together, looking back.

• Bitcoin's 9 cycle turns sat an average of 4.2 months from the S&P RSI's nearest turn. Seven of nine were within 3 months.
• None of 10,000 sets of random months came as close, and none of the 110 shifted RSI series did.

3. But it doesn't warn in time.

• Every RSI turn was known 3 to 23 months after Bitcoin's own turn.
  • Bitcoin topped in December 2017. The RSI's January 2018 peak was known in March 2018.
  • Bitcoin topped in October 2021. The RSI's August 2021 peak was known in February 2022.
• Two RSI swings had no Bitcoin turn at all: late 2019 to March 2020, and late 2024 to April 2025.
• They turn together because both are risk assets. Neither leads the other in a way anyone could use.

4. After a roll-over.

• Taken one signal at a time, Bitcoin did badly. Over the following 12 months the middle result was −6%, against +92% for any month.
• But the signals come in bunches, and the "any month" figure includes Bitcoin's early years. Counted as separate episodes (signals within 6 months as one), there are 6 with a year after them:

• Their middle result, +19%, was below the +63% of a typical month since 2013, but random sets of 6 months come out as low 21% of the time. That isn't a reliable warning.
• The weekly roll-over and the divergence rule gave no warning at all.

5. The three big falls, one by one.

• 2018: no monthly roll-over before Bitcoin's December 2017 top. Bitcoin fell 76%.
• 2020: the one US recession in Bitcoin's life. There were two weaker warnings in 2019, but no roll-over. Bitcoin fell 52% within March.
• 2022: a roll-over in September 2021, one month before Bitcoin's top. Bitcoin fell 73%.

6. Before Bitcoin (for context only).

• 2000: the S&P's RSI peaked 19 months before the top and kept rolling over for two years before the crash. The S&P fell 46%, the Nasdaq-100 81%.
• 2007: it rolled over 3 months before the top. The S&P fell 53%, the Nasdaq-100 50%.
• Warnings came early, often and at irregular times.

7. Gold and copper. The ratio's lows came 2,069, 395, 305, 700, 306, 1,369, 1,157 and 1,216 days apart. Only the last three are near 1,000. The gaps vary more than they would for a random series with the same moves, so the 1,000-day cycle is seen by eye, not in the data.

Update, 27 October 2026: where the idea comes from

Added after this page was published. The text above is left as it was.

• The RSI is J. Welles Wilder's, from 1978. Wilder offered 70 and 30 as lines for "overbought" and "oversold". The maths is worked through above.
• The catch: an RSI peak looks like a warning only afterwards. The RSI is a smoothed record of recent rises. It is highest just after a long run of them, which is near a top. And a peak is known only once the RSI has fallen from it. So, looking back, its peaks line up with tops, but they were known too late to act on (finding 3).
• Cycles seen by eye. Eugen Slutsky showed in 1927 that adding up random shocks makes series that look like regular cycles.
  • For example: flip a coin, add 1 for heads and take 1 for tails, and draw the running total. It rises and falls in swings that look like cycles, though no flip knows about the last.
  • That is why the gold-to-copper "1,000 days" was checked against random series with the same moves (finding 7).
• Theirs and ours. The claims are the video's. The RSI is Wilder's, the random comparison is the Monte Carlo method, and the warning about cycles is Slutsky's, all credited below. The turn rule, "when a turn was known", the random months and the shifted series are ours.

Credits

• The claims tested: The Trading Parrot, "The S&P 500 Chart That Could Change Bitcoin Forever", video, 14 October 2026.
• J. Welles Wilder Jr., New Concepts in Technical Trading Systems, Trend Research, 1978: the RSI and its smoothing.
• Nicholas Metropolis and Stanisław Ulam, "The Monte Carlo Method", Journal of the American Statistical Association, 1949: drawing random comparisons many times.
• Eugen Slutzky, "The Summation of Random Causes as the Source of Cyclic Processes", Econometrica, 1937 (first published in Russian in 1927): random shocks that look like cycles.
• The turn rule, the random months and the shifted series are our own design.
• Data: Yahoo Finance's public price history (the S&P 500, the Nasdaq-100, gold and copper); Coin Metrics' community Bitcoin series from 2010, continued with Yahoo's after it ends; FRED (US recession dates from the NBER, and a copper price as a check).
• PerpGremlin and Claude, Does the S&P 500's RSI warn of Bitcoin's falls?, TerminalGremlin, 2026: the tests, the code and the words.