FAILED

Volatility system (2 × average true range of 20)

A move larger than the market's normal agitation is not noise: it is the start of a direction. Measuring that agitation with the average true range and requiring price to exceed it by a margin gets you into the breakout that matters and ignores the to-and-fro that leads nowhere.

Measured in crypto — Binance spot · 540 pairs, delisted ones included · 0.2% per round trip

Net per trade

+3.26%

after fees

The fee is charged on both legs: every trade pays to open and pays to close.

Died at

invariance

the result is not physically plausible

Worst drawdown

−89%

1,926 days underwater

How far the account fell below its own best previous moment.

Timeframe

1 day

The time grid this was measured on. The same technique on a coarser or finer grid is a different measurement, and can earn a different verdict.

invariance
costs
placebo
benchmark
out of sample
multiple testing

Equity curve

Risking 1.0% per trade, marked to market every day — not only when a trade closes. The line at 1.0 is the capital it started with.

At its worst the account was worth 89% less than its own best previous moment, and it spent 1,926 days below that peak.

The technique against chance

Same number of trades, same holding time, same assets — only the dates were drawn at random.

The average trade and the typical one

The mean sits above the median: the lottery signature — the typical trade loses and a rare handful pays for everything.

What was measured

We coded the rule exactly as it is described and let it trade on its own from 2017-08-17 to 2026-07-28. It found trades in 524 coins, 4,900 in total. Each one enters and exits at the price that existed on that day — the program never sees what comes next, which is the most common mistake people make testing strategies at home. Wherever the description was ambiguous we took the reading least favourable to the technique; whatever was left open is stated in the hypothesis, at the foot of this page.

How many of those trades actually count

Only 29 of the 4,900. Cryptocurrencies rise and fall almost in unison, so the same rule fires on dozens of coins on the same day — and that is one bet repeated, not dozens of different bets. Counting them all separately is the trick that makes a bad test look impressive.

What it paid, before any deductions

+3.46% per trade. This is the number the technique claims, and the only one on this page that has not yet been through a single control.

Why it did not pass — it died here · where the result came from

We look at where the result came from: whether it is spread across the trades or packed into a handful of them. It concentrated. That is the shape of a lottery: the typical trade loses, and a rare handful pays for everything. Anyone following the rule for real would face a long run of losses before any gain, and would almost certainly quit before it arrived. To give the measure: 61% of the profit came from the top 5% of trades, and the same technique with its dates shuffled would concentrate 58%. The control fails above 50%.

What would have happened to the money

Risking 1.0% of capital per trade — the most widely taught rule — across the whole period: The capital would have ended at 0.61× what it started with. At its worst the account was worth 89% less than its own best previous moment, and it spent 1,926 days below that peak. In none of the 12 paths tested did it fall to less than half of what it started with.

The account would not fit every signal

64% of the trades had to be turned down: by the time they appeared, the money was already tied up in other positions. That matters because the average return per trade the technique claims is computed over trades that nobody could have taken all of.

What this result does NOT say

For an edge to be assertable here it would have to reach 4.54% per trade — that is the size that survives this archive's multiple-testing correction, and it rises as the archive grows. So FAILED means “we found nothing above that size”, and never “it cannot possibly work”. The instrument is far more sensitive than that: on synthetic data, with a clean effect, it separates from 0.69% upwards. The distance between the two numbers is the price of a real market and the price of publishing many claims. The difference matters, and it is the rule of this house: the card confronts the claim, never the person who made it.

What this card does not measure

Every verdict holds for the conditions it was measured under. These are this card's — and outside them the result does not apply.

One market, one universe: Measured on 524 spot cryptocurrency pairs, delisted ones included. It says nothing about futures, equities or indices, nor about how the same technique behaves in another market.
One window of time, not every window: The measured period runs from 2017-08-17 to 2026-07-28. A market moves through regimes, and a technique can work in one and fail in another — the card measures the regimes that fit inside this window, not the ones still to come.
One cost structure: The cost charged is 0.10% per leg, in and out. Anyone paying more than that gets a worse result, and anyone paying less gets a better one — the verdict holds for this fee.
One exit rule: The trade was closed by: the technique itself (held until the opposite signal). The same entry measured with a different exit is a different strategy, and can earn a different verdict — it happens in this archive.
The number of trades is not the sample size: There are 4,900 trades, but only 29 independent market episodes: a single move fires the technique across dozens of assets at once, and counting those as separate observations inflates any result. It is the smaller number that governs the arithmetic. With 29 episodes, what the data supports is a range from -12.41% to +19.34% per trade — the published average is the centre of it, not the exact measurement.
Daily bars: Measured at the daily close. Nothing here measures what happens inside the day, and an intraday technique is not auditable with this data.

Numbers and reproducibility

The six controls

control	status	t	threshold	episodes
invariance	failed	—	—	—
costs	inconclusive	0.42	2.05	29
placebo	inconclusive	0.03	2.05	29
benchmark	failed	—	—	—
out of sample	inconclusive	—	—	—
multiple testing	failed	—	—	—

invariance61% of the gross profit comes from 245 trades (5% of the total) — lotteryconcentration: +61.26% of the profit sits in the top 5% of trades — with the dates shuffled, +57.62% (fails above +50.00%)
costsgross +3.461% · cost 0.200% · net +3.261% (t=0.42) · the range runs from -12.614% to +19.135%
placeboactual +3.261% · placebo +2.996% · excess +0.265% ± 8.132% (t=0.03 against a threshold of 2.05, 29 real groups, 245,000 sham dates, draw error ±0.537%) — the status flips inside the placebo's own Monte Carlo error
benchmarktechnique +3.26% · buy and hold (same horizon) +13.76% · excess -10.50%
out of sampleasset half A: +4.017% (t=0.57, 29 episodes) · asset half B: +2.432% (t=0.25, 28 episodes) · liquid half (>= US$ 2,066,613/day): +12.400% (t=1.20, 29 episodes) · illiquid half: -6.221% (t=-0.90, 27 episodes) · period 1/4 (2017-09-14 a 2022-01-18): +23.474% (t=1.61, 15 episodes) · period 2/4 (2022-01-21 a 2023-06-08): -12.732% (5 episodes — too small, does not count) · period 3/4 (2023-06-10 a 2024-12-08): +2.919% (6 episodes — too small, does not count) · period 4/4 (2024-12-09 a 2026-07-25): -0.563% (6 episodes — too small, does not count)
multiple testing18 variation(s) tested · t=0.42 across 29 episodes (equivalent to t=0.40) · p≈0.6872 · false positives expected by chance ≈ 12.37

Equity — outside the six controls, and here is why

The t of the trade series is invariant to bet size: 0.5%, 1% and 3% agree to the sixth decimal. Nothing here moves the verdict — it moves what the account would have lived through.

risking 1.0% per trade: ×0.61
worst drawdown from the peak: −89%
days below the previous peak: 1,926
signals refused for lack of capital: 64%
paths where the account halved (out of 12): 0

Reproducibility

period: 2017-08-17 to 2026-07-28
assets that traded: 524
variations tested before this one: 18
gross per trade: +3.46%
net per trade: +3.26%
exit rule: the technique itself (held until the opposite signal)
median duration bars: 94
mean duration bars: 127.7
max duration bars: 1402
fee per leg: 0.001
episode days: 112
seed: 20260728
Bookstaber k: 2.0
Bookstaber ATR: 20

Binance spot klines (delisted pairs included) · collected from 2026-07-27 23:31 to 2026-08-03 10:49 · 540 assets · 750,934 bars · 2017-08-17 to 2026-07-28

Hypothesis, filed before the result

The chapter's only complete system, pp. 860–861: «buy if the next close rises more than k × ATR_t(n) relative to the current close; sell if it falls more than k × ATR_t(n)». The source classifies it as a «volatility breakout» and gives no exit rule — it reverses on the opposite signal and stays positioned after the first, which is the reading of its own mechanics. BOTH PARAMETERS ARE OPEN: k is declared open by the source («approximately 3, but it can vary up or down, to make signals less or more frequent») and the ATR's n IS NOT GIVEN anywhere in the section. Both enter as a SWEEP. The wording invites lookahead and the technique has none: «the NEXT close against TODAY's ATR» is implemented with the previous bar's ATR, which is what makes the rule causal. Family-level prediction, recorded before measuring: (1) NONE survives the corpus-wide Benjamini-Hochberg — a filter that only decides WHEN NOT TO ENTER does not create an effect, it redistributes one; (2) the gravedigger will be the BENCHMARK, and not the placebo, and this is where this family parts from `fibonacci`: there the claim was that one specific level matters, and the placebo is what asks that; here the claim is that REMOVING TRADES IMPROVES THE RESULT, and what asks that is the control that compares against doing nothing — the protocol already records that a filter almost always cuts `n` without separating anything; (3) the LOW volatility filter and the HIGH one will give results IN THE SAME DIRECTION, despite the source claiming the low one is superior with two numbers and no error bar (information ratio 1.143 against 0.867) — if both improve, what improves is cutting trades, not picking a regime; (4) the RESET rule will have the family's best `t`, being the only layer that gives exposure back instead of taking it away, and even so it will not pass the BH; (5) the BASE will do better than any layer built on top of it.

filed on 2026-08-04, before the number existed

The original, as it was filed

O único sistema completo do capítulo, p. 860–861: «compra se o próximo fechamento subir mais que k × ATR_t(n) em relação ao fechamento corrente; vende se cair mais que k × ATR_t(n)». A fonte o classifica como «rompimento de volatilidade» e não dá regra de saída — reverte no sinal contrário e fica sempre posicionado depois do primeiro, que é a leitura da própria mecânica. ⚠️ OS DOIS PARÂMETROS SÃO ABERTOS: o k é declarado aberto pela fonte («aproximadamente 3, mas pode variar para mais ou para menos, para tornar os sinais menos ou mais frequentes») e o n do ATR NÃO É DADO em lugar nenhum da seção. Os dois entram como VARREDURA. ⚠️ A redação convida ao lookahead e a técnica não tem: «o PRÓXIMO fechamento contra o ATR de HOJE» se implementa com o ATR da vela anterior, que é o que torna a regra causal. Previsão da família, registrada antes de medir: (1) NENHUMA sobrevive ao Benjamini-Hochberg do corpus — um filtro que só decide QUANDO NÃO ENTRAR não cria efeito, redistribui; (2) o coveiro será o BENCHMARK, e não o placebo, e é aqui que esta família se separa de `fibonacci`: lá a alegação era que um nível específico importa, e o placebo é quem pergunta isso; aqui a alegação é que REMOVER OPERAÇÕES MELHORA O RESULTADO, e quem pergunta isso é o controle que compara com não fazer nada — o protocolo já registra que filtro quase sempre corta `n` sem separar nada; (3) o filtro de volatilidade BAIXA e o de ALTA darão resultados NA MESMA DIREÇÃO, apesar de a fonte alegar superioridade do baixo com dois números e nenhuma barra de erro (razão de informação 1,143 contra 0,867) — se os dois melhoram, o que melhora é cortar operação, não escolher regime; (4) a regra de RESET terá o melhor `t` da família, por ser a única camada que devolve exposição em vez de tirar, e ainda assim não passará no BH; (5) a BASE irá melhor que qualquer camada sobre ela.

Quotations from the source were translated from the Portuguese record and back into English — they are not the author's exact words.

Pre-registration exists to keep prediction apart from rationalisation: written after the number, every hypothesis is right.

This claim has been audited once — there is no history to compare against.