McGinley Dynamic (10, k=0.6)
An ordinary moving average treats every departure of price the same way. Making the step depend on the distance between price and the line itself lets the average adjust on its own to how far the market has strayed from it — closing the gap quickly when it has fallen behind, and resisting when price is merely oscillating around it.
Measured in forex — EUR/USD and GBP/USD · daily bars built from 15m · 1 pip spread (~0.009% per round trip) · no survivorship bias
- ✓invariance
- ✗costs
- ?placebo
- ✗benchmark
- ?out of sample
- ?multiple testing
At its worst the account was worth 0% less than its own best previous moment, and it spent 2,138 days below that peak.
What was measured
How many of those trades actually count
What it paid, before any deductions
Why it did not pass — it died here · the broker's fee
What would have happened to the money
What this result does NOT say
- One market, one universe
- Measured on 2 spot currency pairs. 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 2020-01-01 to 2026-06-26. 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 a 1.0 pip spread, crossed once. 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 710 trades, but only 270 independent market episodes: trades that overlap in time are not independent observations, and counting them as if they were inflates any result. It is the smaller number that governs the arithmetic. With 270 episodes, what the data supports is a range from -0.10% to +0.10% 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 | passed | — | — | — |
| costs | failed | -0.08 | 1.97 | 270 |
| placebo | inconclusive | 0.07 | 1.97 | 270 |
| benchmark | failed | — | — | — |
| out of sample | inconclusive | — | — | — |
| multiple testing | inconclusive | — | — | — |
- invariance710 signals across 4055 bars
- costsgross +0.004% · cost 0.008% · net -0.004% (t=-0.08)
- placeboactual -0.004% · placebo -0.008% · excess +0.004% ± 0.051% (t=0.07 against a threshold of 1.97, 270 real groups, 35,500 sham dates, draw error ±0.005%) — the status flips inside the placebo's own Monte Carlo error
- benchmarktechnique -0.00% · buy and hold (same horizon) +0.00% · excess -0.01%
- out of sampleasset half A: -0.001% (t=-0.02, 216 episodes) · asset half B: -0.007% (t=-0.13, 212 episodes) · liquid half (>= US$ 0/day): -0.004% (t=-0.08, 270 episodes) · period 1/4 (2020-01-02 a 2021-08-13): +0.054% (t=0.51, 70 episodes) · period 2/4 (2021-08-16 a 2023-04-02): -0.010% (t=-0.08, 70 episodes) · period 3/4 (2023-04-03 a 2025-01-15): +0.021% (t=0.24, 71 episodes) · period 4/4 (2025-01-16 a 2026-06-14): -0.082% (t=-1.07, 61 episodes)
- multiple testing1 variation(s) tested · t=-0.08 across 270 episodes (equivalent to t=-0.08) · p≈0.9367 · false positives expected by chance ≈ 0.94
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
- ×1.00
- worst drawdown from the peak
- −0%
- days below the previous peak
- 2,138
- signals refused for lack of capital
- 0%
- paths where the account halved (out of 12)
- 0
Reproducibility
- period
- 2020-01-01 to 2026-06-26
- assets that traded
- 2
- variations tested before this one
- 1
- gross per trade
- +0.00%
- net per trade
- −0.00%
- exit rule
- the technique itself (held until the opposite signal)
- median duration bars
- 3
- mean duration bars
- 5.7
- max duration bars
- 44
- spread pips
- 1.0
- episode days
- 7
- seed
- 20260728
- McGinley period
- 10
- McGinley k
- 0.6
Twelve Data forex (15m aggregated to 1d) · collected on 2026-07-25 · 2 assets · 4,055 bars · 2020-01-01 to 2026-06-26
Hypothesis, filed before the result
A legible formula but with the source text degraded in transcription, and the chapter gives NO signal rule at all — it is an indicator without a claim. It will be measured by the direction of the line, by analogy with the rest of the chapter, and that goes declared as THE AUDITOR's choice, not McGinley's. Family prediction, filed before measuring: (1) none of the adaptive averages survives the family's Benjamini-Hochberg in crypto; (2) the control that kills the most will be COST, not the benchmark — unlike the chart-pattern census, where the benchmark was the gravedigger, because these are always-in-the-market systems and they turn over a lot; (3) each adaptive average will have HIGHER turnover than its fixed-period counterpart already in the archive, and will die more at cost than it does. The mechanism is in the source itself: Table 17.1 reports a profit factor 'even before costs' and states that success is 'inversely related to the average number of trades' (KAMA 159 trades, factor 1.53; VIDYA 443, factor 1.17). That is a cost story told as a quality story. If I am wrong and one survives cost with higher turnover, the chapter's thesis gains evidence it did not present. MEASURED IN FOREX (EUR/USD and GBP/USD, daily bars built from 15m), not in crypto. This is a pre-registered REPLICATION of the same technique in the second market — not a new discovery — and the multiple-testing count treats it as such. What changes relative to the crypto card: 2 pairs over 6.5 years against 540 over 9, a round trip costs about 20× less (a 1 pip spread crossed once, against 0.1% commission per leg), the detection floor is about 10× lower (0.092% against 0.97%) and there is NO survivorship bias, because a currency pair does not get delisted. Specific prediction: with only 2 assets, the independent episodes come from TIME and not from the variety of assets, so I expect a low effective N and a high proportion of INCONCLUSIVE — and whatever is asserted here is worth more than in crypto, because cost cannot be the gravedigger.
filed on 2026-07-31, before the number existed
The original, as it was filed
Fórmula legível mas com o texto-fonte degradado na transcrição, e o capítulo NÃO dá regra de sinal nenhuma — é indicador sem alegação. Será medida pela direção da linha, por analogia com o resto do capítulo, e isso vai declarado como escolha DO AUDITOR, não de McGinley. Previsão da família, registrada antes de medir: (1) nenhuma das adaptativas sobrevive ao Benjamini-Hochberg da família em cripto; (2) o controle que mais mata será o CUSTO, e não o benchmark — diferente do censo de padrões gráficos, onde o benchmark foi o coveiro, porque estas são sistemas sempre-no-mercado e giram muito; (3) cada adaptativa terá giro MAIOR que a sua contraparte de período fixo já no corpus, e morrerá mais no custo do que ela. O mecanismo está na própria fonte: a Tabela 17.1 relata fator de lucro 'even before costs' e afirma que o sucesso é 'inversely related to the average number of trades' (KAMA 159 operações, fator 1,53; VIDYA 443, fator 1,17). Isso é uma história de custo contada como história de qualidade. Se eu estiver errado e alguma sobreviver ao custo com giro maior, a tese do capítulo ganha uma evidência que ele não apresentou. ⚠️ MEDIDO EM FOREX (EUR/USD e GBP/USD, diário agregado de 15m), e não em cripto. Isto é uma REPLICAÇÃO pré-registrada da mesma técnica no segundo mercado — não uma descoberta nova —, e a conta de múltiplos testes a trata como tal. O que muda em relação ao card de cripto: são 2 pares em 6,5 anos contra 540 em 9, o custo do giro é ~20× menor (spread de 1 pip cruzado uma vez, contra comissão de 0,1% por lado), o piso de detecção é ~10× menor (0,092% contra 0,97%) e NÃO há viés de sobrevivência, porque par de moeda não é deslistado. Previsão específica: com apenas 2 ativos, os episódios independentes vêm do TEMPO e não da variedade de ativos, então espero N efetivo baixo e uma proporção alta de INCONCLUSIVO — e o que for afirmado aqui vale mais que em cripto, porque o custo não tem como ser o coveiro.
Pre-registration exists to keep prediction apart from rationalisation: written after the number, every hypothesis is right.