Research / Claude preview

Interpret. Question. Validate.

Explore how Anthropic’s Claude API is planned to help researchers explain performance, understand risk and form testable hypotheses.

Strategy sample

Synthetic demonstration data

Prices, trades and returns are invented examples dated in 2024, not recorded backtests or historical market data. Explore the workflow; these numbers do not demonstrate profitability.

BTC / USD1 day2024 · Invented datesStarting sample equity $10,000.00

01Backtest results
02Claude analysis
03Risk explanation
04Research hypotheses
05Further validation

The engine supplies the evidence

A public synthetic example of the structured input planned for Claude.

Strategy context: Trend research
Trade count: 18
Net sample return: +3.03%
Cost assumptions: Fees + slippage
Data limitations: Explicitly supplied

The planned API receives analytical results and context, not exchange keys or trading authority.

Inspect planned input format

{
  "data_kind": "synthetic",
  "sample": "Trend research",
  "trade_count": 18,
  "net_return_pct": 3.0267,
  "max_drawdown_pct": -1.0387,
  "fees_usd": 102.9041,
  "slippage_usd": 30.8713,
  "assumptions": {
    "feeRate": 0.001,
    "slippageRate": 0.0003,
    "exposure": 0.28,
    "priceBasis": "Invented prices; no exchange data",
    "equityBasis": "Daily marked-to-market equity after estimated liquidation costs",
    "execution": "Predefined sample trades; no strategy execution",
    "currency": "USD"
  },
  "limitations": [
    "Invented prices and trades",
    "No out-of-sample validation",
    "Not evidence of profitability"
  ]
}

Illustrative research report

Planned integration

Written example template using synthetic metrics. Not a Claude API response.

01 / Performance context

A result is a starting point.

The sample ends at +3.03% after costs across 18 predefined trades. A positive sample result does not establish an edge.

02 / Risk interpretation

Read the path, not just the endpoint.

Daily drawdown reaches -1.04%. Fees and slippage total $133.78. The sample cannot tell us how a strategy behaves with different data, costs or execution assumptions.

03 / Research hypothesis

Test whether the observation holds.

A candidate question: does the same approach remain stable in an untouched sample with higher trading costs? Define the experiment first; the quantitative engine, not the explanation, must determine its outcome.

Research assistant concept

Choose a question to explore a written example. No AI request is sent.

What could a headline return hide?

This synthetic path reaches a -1.04% daily drawdown. A terminal return of +3.03% says little about losses along the way, trade concentration or a different market regime. A real report would need more evidence before drawing conclusions.

Illustrative answer

Further validation

0/3 noted

A local research checklist, not an executed test or verdict.

Define an untouched evaluation windowStress fees and slippage assumptionsReview concentration and losing trades

Checking an item only records your choice during this visit. This demo never executes validation or trades.

Return to the evidence