Jenkins & Jones Open live

Free research report · 15 pages · PDF

What Doesn’t Work

5,192 strategy variants tested on 5-minute Bitcoin prediction markets

By Zion Boggan, Jenkins & Jones · October 30, 2026 · Version 1.0 · 15 pages

The bar rises with every idea testedOne-sided Bonferroni t-statistic needed at a 5% family-wise error rate, rising from 1.64 for one test to 2.39 for 6, 2.94 for 30, 3.92 for 1,152 and 4.27 for all 5,192 variants.2.03.04.01101001k10kt = 2, the usual “significant” line6 lanes · 2.3930 · 2.94one sweep, 1,152 · 3.92all 5,192 · t ≥ 4.27Variants tested (log scale)t needed
Figure 1. One-sided Bonferroni threshold at a 5% family-wise error rate. A t of 2 is the usual line; after 5,192 tries the bar is 4.27.
distinct variants logged in one registry
5,192
the bar after that many tries
t ≥ 4.27
clear it on data they were not designed on
0

How we tested them, which families failed and why, the simulator bugs that nearly fooled us, and the two things that still hold up. Written for people who research, run or build data for these markets.

What is inside

  1. The market. How a 5-minute round settles, and what the taker fee does to break-even at each price.
  2. Method. Pre-registration, one global multiple-testing bar, forward scoring on unseen rounds, realistic fills.
  3. The registry. Where the 5,192 variants went, and what happened to each.
  4. What failed, and why. Pattern searches, dip and fade rules, momentum, pricing models, cross-venue lead, exits and sizing.
  5. The bugs that nearly fooled us. Nine simulator mistakes, the number each produced, and the fix.
  6. What holds up. Market calibration, and a closing-minute lock that is still small and still unproven.
  7. What this means for researchers, venues and data teams.

Who it is for

Quant and market-microstructure researchers, teams running or listing event contracts, and anyone buying or building prediction-market data who wants to know which backtest results to distrust.

What it is not

It is not a trading system and it does not sell signals. Every strategy result in the report is hypothetical, from simulated trading on recorded or historical data, and carries the CFTC Rule 4.41 notice. Losses get the same space as gains.

Who wrote it

Jenkins & Jones is the prediction-market data and research brand of ZCC Media Agency, founded and run by Zion Boggan. We build market-data recorders, integrity audits and research pipelines. To talk about your data, book a 20-minute call.