backtest.hpp File Reference

Relativistic Backtester — AGT-05 public API. More...

#include "srfm/types.hpp"
#include "srfm/constants.hpp"
#include <optional>
#include <span>
#include <string>
#include <vector>

Go to the source code of this file.

Classes
struct	srfm::backtest::BarData
	A single time-bar of backtester input. More...
struct	srfm::backtest::LorentzCorrectedSeries
	A complete set of relativistic corrections for one return series. More...
struct	srfm::backtest::PerformanceMetrics
	Performance metrics for a single strategy evaluation. More...
struct	srfm::backtest::BacktestComparison
	Side-by-side comparison of raw vs relativistic strategy metrics. More...
struct	srfm::backtest::BacktestConfig
	Configuration for a backtest run. More...
class	srfm::backtest::PerformanceCalculator
class	srfm::backtest::LorentzSignalAdjuster
class	srfm::backtest::Backtester
struct	srfm::backtest::BarDataEx
struct	srfm::backtest::RegimeBacktestResult
	Performance summary for all three regime strategies. More...
class	srfm::backtest::RegimeFilteredBacktester

Namespaces
namespace	srfm
namespace	srfm::backtest

Typedefs
using	srfm::backtest::ReturnSeries = std::vector< double >

Detailed Description

Relativistic Backtester — AGT-05 public API.

Module: Relativistic Backtester

Responsibility

Feed every strategy signal through Lorentz corrections (γ-weighted) before evaluation, and measure the performance lift — or cost — of relativistic adjustment versus classical raw-signal strategies.

The Core Idea

In high-velocity market regimes (high β), conventional strategy signals underweight information that is arriving "fast" relative to the market observer frame. Applying the Lorentz factor γ = 1/√(1−β²) re-weights each signal proportional to the "market speed" at the time it was generated:

adjusted_signal_t = γ(β_t) · raw_signal_t

A strategy evaluated on adjusted signals implicitly up-weights signals from fast-moving markets and down-weights signals from quiet, near-Newtonian regimes (β ≈ 0, γ ≈ 1).

Performance Metrics

Four metrics are reported for both raw and relativistic strategies:

• Sharpe ratio: (mean − r_f) / σ (annualised)
• Sortino ratio: (mean − r_f) / σ_down (annualised)
• Maximum drawdown: max peak-to-trough loss
• γ-weighted information ratio: IR_γ = (mean(rel_ret − bm_ret) · mean(γ)) / σ(rel_ret − bm_ret)

Guarantees

• Zero UB: all fallible operations return std::optional or bool
• No raw pointers: ownership by value or const-reference
• Thread-safe reads: const member functions are safe concurrently
• Returns must be finite; NaN/Inf inputs produce std::nullopt

NOT Responsible For

• Signal generation (see src/momentum/)
• Covariance estimation (see src/tensor/)
• Data loading (see src/core/)

Definition in file backtest.hpp.