Boom and Crash spikes have no memory: a test on about 98,000 one-minute candles per index

18 November 2026, 18:31
Jerson Ernesto Chura Pacci
0
7

A common idea behind Boom and Crash tools is that a spike becomes "due": the longer the index goes without one, the closer the next one should be. Before building any spike tool, I tested that idea on real one-minute data. This post shows the method, the numbers and their limits.

Data and method

  • Six indices: Boom 300, Boom 500, Boom 1000, Crash 300, Crash 500 and Crash 1000.
  • One-minute candles from 31 August to 8 November 2026 (UTC), taken from a broker's demo server: between 98,089 and 98,097 candles per index.
  • Spike definition: a candle whose body in the spike direction (up for Boom, down for Crash) is larger than 8 times the median absolute candle body of the series.
  • Measured value: the probability of at least one spike within the next 5 candles. First for all candles (the base rate), then split by the time elapsed since the last spike, in bands relative to the average gap: 0-0.5x, 0.5-1x, 1-2x, 2-3x and more than 3x the average gap.

If spikes became "due", the probability should rise from the first band to the last one.

Results

IndexSpikesAverage gap (candles)Gap CVBase rate0-0.5x0.5-1x1-2x2-3x>3x
Boom 10003,94824.80.9818.7%18.9%18.2%19.0%18.2%17.9%
Crash 10004,01224.50.9918.8%19.0%18.4%18.9%19.8%17.4%
Boom 5004,77220.50.9722.0%22.3%21.5%21.8%22.5%22.9%
Crash 5004,78920.50.9822.2%21.8%22.4%22.4%22.8%21.7%
Boom 3003,00532.61.2214.2%15.8%15.9%13.4%11.9%8.3%
Crash 3003,04332.21.1114.6%15.6%15.4%14.3%12.9%10.3%

Gap CV is the coefficient of variation of the gaps between spikes (standard deviation divided by the mean). Base rate and band columns show the probability of a spike within the next 5 candles.

What the numbers say

  • 500 and 1000 indices: the probability stays within 1.4 percentage points of the base rate in every band, with deviations in both directions and no consistent upward trend, and the gap CV is close to 1. That is what a memoryless process produces: the time already waited tells you nothing about the time still to wait.
  • 300 indices: the probability does change, but in the opposite direction: it falls as the wait grows (Boom 300: from 15.8% to 8.3%), and the gap CV is above 1. Spikes there tend to come in clusters rather than becoming "due".
  • In none of the six indices did the probability of a spike rise in a meaningful way with the time since the last one: the largest gain from the first band to the last one is 0.6 points (Boom 500).

A second test: three candles against the spike direction

Another popular filter is to wait for a pullback before a spike. I measured the probability of a spike within 5 candles right after 3 consecutive candles against the spike direction. Result versus the base rate: Boom 1000 18.7% vs 18.7%, Crash 1000 18.7% vs 18.8%, Boom 500 22.1% vs 22.0%, Crash 500 22.1% vs 22.2%, Boom 300 14.3% vs 14.2%, Crash 300 14.8% vs 14.6%. The filter makes no measurable difference.

Practical takeaways

  • A "spike is due" signal based on elapsed time is not supported by this data.
  • If an arrow indicator shows a high hit rate on history, check whether its arrows are drawn only after the spike candle has closed, or moved later (repainting).
  • Since timing gives no edge here, position size and loss limits are the part you can control.

Limitations

One period of about ten weeks, one data feed, one spike threshold (8 times the median body) and one 5-candle window. Other thresholds, windows or periods may give different numbers. Past statistics do not describe future results.

Measure it on your own chart

I built Chakana Spike Stats (free) to show these statistics live instead of predictions. It marks every spike candle when it closes and never repaints, lets you set what counts as a spike (body larger than X times the median body of the previous N candles), and shows candles since the last spike, spikes today and in the last 24 hours, the average and median gap, and the largest spike of the day. It can watch several symbols in one table, alert you when a spike candle closes, and export the spikes to CSV.

Trading synthetic indices involves a high risk of loss. This post is a statistical study, not trading advice.