Every idea is rebuilt exactly as taught, judged against coin-flip and placebo controls, and run against market bars with real, dated costs. The failures are published so everyone learns what is real.
How we count: 198 strategies rated in our July 2026 matrix (179 rated 1–2 stars: dead, falsified, noise-grade or unvalidated), plus 53 strategy studies logged Jul–Oct 2026, each with a written FAIL verdict and each counted as ONE strategy even when it held thousands of variants. Studies still being checked privately are not counted, so the real totals are higher.
Every family below has at least one study with a frozen plan and a written verdict. The red bar is the number of failed studies; tap a family to filter the log.
Newest first. "Tested" is what was rebuilt, "scale" is how much, and the verdict is the plain-English result. Counts of tests and verdicts are current; any performance chart on this site ends at least three months back.
Loading the log…
Every dot is one failed strategy run from our public weekly dataset — 195 runs across 77 strategies and 7 instruments. The median run was green in 36% of its weeks; 70 were green in 40% or more, and 6 won more weeks than they lost. All of them bled out anyway. Hover or tap a dot.
Historical simulation in R (multiples of risk per trade), costs included · ISO weeks 2020-W53 → 2026-W29 · education, not advice
Cumulative weekly R for the six deepest losers in the dataset. Green weeks, red weeks — and the line still goes to the floor.
290 ISO weeks, 2020-W53 → 2026-W29 · education, not advice · simulations
Most backtests fail in the method, not the market. These are the house checks. Studies from October 2026 on went through all of them; earlier studies went through the ones that existed at the time, and several of these rules were born from mistakes those studies exposed.
Securities Transaction Tax on Indian index futures (sell side). Every trade in a test pays the rate of its own date.
| From | Futures STT |
|---|
Source: Union Finance Bills and NSE STT tables.
Markets changed after 2020. New verdicts are judged on 2021-onward data, 2020 is kept as a separate stress test, and older years are used only for labelled context. Fewer years means more lucky passes, so the pass bars were not lowered.
Honest testing means hunting your own bugs. Each of these once distorted a result, until a replica or an audit caught it.
Cleaned, de-duplicated, session-aligned and integrity-checked daily. Bar counts are upper bounds read from the stores and grow every night.
| Store | Coverage |
|---|---|
| Audited intraday store — M bars, 599 instruments | 1/5/15/30/60-minute bars |
| NIFTY, BANKNIFTY, India VIX + 19 sector indices — spot | Jan 2015 → today |
| NIFTY & BANKNIFTY futures — real volume | 2020 → Oct 2024 · May 2026 → today |
| 531 NSE stocks — real volume | 2015 → today |
| Gold, silver, copper, crude, natural gas — global, 24h | 2013 → today |
| 23 crypto pairs | BTC & ETH from Aug 2017 → today |
| Index ETFs, currency futures, MCX gold & silver minis | various → today |
| Option minutes with open interest — 111M rows | 2020 → Oct 2024 · mid-2026 → today |
| NSE F&O daily files, every contract incl. stock options | 3,915 sessions · 2011 → Oct 2026 |
| Every NSE stock daily, incl. delisted · point-in-time index membership | 2008 → 2026 |
| Daily histories | stocks from 1996 · commodities from 2000 · indices from 2007 |
| Non-price data: participant open interest, delivery %, futures positioning (COT), crypto taker flow, news tone | 2020 → today (varies) |
Iron condors, ORB, candlestick classics, VWAP, day-of-week folklore — rebuilt exactly as taught, every one negative out-of-sample.
Browse the case studies →Names, logic and sources of publicly documented strategies, searchable by category. Catalogued here, tested honestly, and — usually — buried in the graveyard.
Open the encyclopedia →