Trading AI

Method and sources

Last updated: 26 October 2026

The strength shown with each argument comes from research on that kind of evidence, not from the AI, and the app does the arithmetic itself.

Artificial intelligence

Each analysis is written by Claude (Anthropic), set to a high level of reasoning: it thinks before it writes. This setting is the same for every analysis and every subscription. Your photo is sent to it encrypted, for this analysis only.

What it reads

It first reads the chart as printed: the asset, the timeframe and the price scale, marking each value as read or estimated. It then deduces the horizon and reads the structure, the levels where price has already reacted and the possible scenarios. It proposes a plan only when the chart supports one.

What it relies on

Each argument carries the strength that published research has measured, never one the AI gives itself. Support and resistance rest on work from the Federal Reserve Bank of New York, round numbers on a study in the Journal of Finance, and volatility regimes on the research awarded the 2003 Nobel Prize in Economics. Trends, chart patterns and indicators remain clues, never proof.

What the app checks itself

Claude reads and judges; the app does the arithmetic. It measures reward-to-risk R from the worst plausible entry, rounded down, and the win rate needed to break even, (1 + c) / (1 + R), where c is the typical cost divided by the risk. It rejects a stop that sits within ordinary candle noise, measured with the square root of time as the Basel Committee does, and, except for long-term plans, limits a plan's duration with a simulation of price moving at random. A plan that fails a check is withdrawn, and the report says why.

How to read likelihood

Likelihood is given in words, on the scale of US intelligence analysts (ICD 203): “likely” covers 55 to 80%, “roughly even chance” 45 to 55% and “unlikely” 20 to 45%. The app keeps it close to what the distances to the target and to the invalidation alone suggest. Reading confidence is a separate measure: it judges the image, not the market.

What it cannot see

A chart does not show the news, the order book or your broker's actual costs. A sharp photo, taken straight on with the price axis visible, gives the best reading. The analysis is educational and is not investment advice.

Sources

Bachelier (1900). Théorie de la spéculation. Annales scientifiques de l'École normale supérieure.
Basel Committee on Banking Supervision (1996). Amendment to the Capital Accord to Incorporate Market Risks. Bank for International Settlements.
Bollerslev (1986). Generalized Autoregressive Conditional Heteroskedasticity. Journal of Econometrics.
Engle (1982). Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation. Econometrica.
Kavajecz & Odders-White (2004). Technical Analysis and Liquidity Provision. Review of Financial Studies.
Lo, Mamaysky & Wang (2000). Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation. Journal of Finance.
Moskowitz, Ooi & Pedersen (2012). Time Series Momentum. Journal of Financial Economics.
Office of the Director of National Intelligence (2015). Analytic Standards. Intelligence Community Directive 203.
Osler (2000). Support for Resistance: Technical Analysis and Intraday Exchange Rates. FRBNY Economic Policy Review.
Osler (2003). Currency Orders and Exchange Rate Dynamics: An Explanation for the Predictive Success of Technical Analysis. Journal of Finance.
Sullivan, Timmermann & White (1999). Data-Snooping, Technical Trading Rule Performance, and the Bootstrap. Journal of Finance.

Questions people ask

Where does the strength of each argument come from?

Each argument carries the strength that published research has measured, never one the AI gives itself.

Does the analysis use Fibonacci, Elliott waves or smart money concepts?

No. Fibonacci ratios, Elliott waves, harmonic patterns, Wyckoff schematics, smart money and ICT concepts are never used as arguments.

Is the break-even win rate the plan's chance of success?

Break-even win rate, costs included: (1 + c) / (1 + R), where R is the reward-to-risk and c the costs in units of risk. It is not the plan's chance of success.