Understanding Our Signal Metrics

Our Commitment to Transparency

At SYGNAL, we believe that providing clear, transparent, and reproducible data is essential to our mission.

The performance numbers we show — such as yearly return (return pa), volatility, and max drawdown — come from an automated simulation using historical market data to estimate how each model might have performed in the past.

This document explains in simple terms how our system tests how signal models would have performed in the past. It covers the main ideas behind the test, how we calculate returns, and how we summarise these into our main performance numbers.

Please note that all figures presented are hypothetical, derived solely from model simulations, and do not reflect live trading results. They exclusively represent simulated historical model behaviour under standardised conditions, rather than any real-world trading performance.

Key Assumptions (Summary)

Category	Assumption
Data	Historical exchange prices (e.g., daily close data). Adjusted for splits/dividends where applicable.
Time handling	Each data point uses a UNIX timestamp in UTC. Returns are grouped by calendar day.
Execution timing	The signal at the start of each period determines the simulated position for that period.
Leverage	A constant leverage factor is applied to all positions. The system does not currently use adaptive volatility targeting.
Signal bounds	Signals range from -1.0 (strong short) to +1.0 (strong long). Zero indicates a neutral position.
Transaction cost (slippage)	A standardised relative slippage rate applies whenever the signal changes. The cost equals \|ΔSignal\| × L × Slippage Rate. The initial entry also incurs a one-time cost of \|S₀\| × L × Slippage Rate.
Costs excluded	No brokerage commissions, borrow fees, funding rates, taxes, or cash drag are included.
Shorting	Permitted in simulation.
Data latency	Prices are assumed available at the close; no look-ahead bias.
Survivorship bias	Historical datasets exclude delisted instruments where possible, but full elimination cannot be guaranteed.

Parameters (current defaults)

L (leverage factor): 1.0 (default when not explicitly provided by the caller)
Slippage Rate (relative): 0
Trading days per year used for scaling: 252
Daily aggregation window: 00:00–23:59 UTC
Initial entry cost applied: Yes (|S₀| × L × Slippage Rate)

Step 1 – Signal Generation

Every analysis begins with a signal series, produced by the model under evaluation:

Input Data: Historical market data (e.g., prices) for a given instrument.
Model Application: The model processes the data point by point.
Signal Output: Each point produces a signal value in the range -1.0 to +1.0, where the sign indicates direction and magnitude indicates conviction.
Series Creation: These values are recorded sequentially, forming a complete historical signal time series.

Step 2 – Return Simulation (Back-testing Simulation)

Using the historical signal series, the system simulates hypothetical profits and losses (P&L) for the purpose of analysis and evaluation.

For each period between one signal and the next, the return is calculated as:

Period Return = (Signal_t × L × Price Change_t) − (|ΔSignal_t| × L × Slippage Rate)

where:

Signal_t = signal value fixed at the start of the period,
L = leverage factor,
Price Change_t = percentage price movement of the instrument during that period,
Slippage Rate = standardised transaction cost per 100% signal change.

The first signal incurs a one-time cost:

|S₀| × L × Slippage Rate.

This way, every change in the signal adds a real trading cost, making the simulation careful and repeatable.

Step 3 – Data Aggregation

The back-test may generate multiple signal changes within a single day. Before we determine the main performance numbers, we add up all returns for each day to get a single number.

All simulated returns that occur within a single calendar day (00:00 UTC – 23:59 UTC) are summed into one daily return.
Trading activity data is aggregated in the same way.

This gives a clear and steady set of daily returns, which we use to make all our main statistics.

Step 4 – Key Performance Indicator (KPI) Calculation

We use these daily numbers to determine the main statistics you see on the SYGNAL website.

Annualised Return (Return p.a.)

The geometric annualised return (CAGR) represents the constant annual growth rate of simulated equity:

CAGR = (Ending NAV)^{(1 / N_years)} − 1

All metrics are computed from the aggregated daily return series in UTC.

Note: We display the compound annual growth rate (CAGR) rather than the simple arithmetic average return.

Annualised Volatility (Volatility p.a.)

Annualised volatility measures variability:

Volatility p.a. = StdDev(Daily Returns) × √252

252 corresponds to the approximate number of trading days in a year.

Maximum Drawdown (Max DD)

This represents the largest drop from a past high to the subsequent lowest point, based on the hypothetical account value.

The system constructs a cumulative equity path (starting at 1.0) and tracks each peak-to-trough ratio. The Max DD equals:

1 − 1 / (max peak-to-trough factor).

Additional Internal Metrics

Signal Turnover per Year

We calculate the total absolute changes in the signal and divide by the number of years to determine trading intensity. This tells how often the model changed its position historically. This continuous measure differs from counting individual trades or ‘round turns’.

Interpretation and Limitations

These results measure historical model behaviour with a standard method. What happens in real markets can differ due to timing, costs, trading limits, and human factors. Other data, frequencies, or parameters yield different results. These simulated metrics are intended to help compare models under identical assumptions, not to forecast investment outcomes.

Compliance & Disclosures

Important Information about Hypothetical / Backtested Results

The performance metrics displayed on SYGNAL are entirely hypothetical, generated through the back-testing methodology described above, and do not represent actual trading.

Backtested data has inherent limitations, including:

Reliance on historical information and standardised assumptions.
It does not include costs such as commissions, borrowing, funding, taxes, and cash not invested, and it assumes that trading is always possible.
It also does not account for how real investors behave or any operational limitations.

Results are for research and evaluation only. Past simulated performance does not predict future results. Different assumptions would produce materially different outcomes.

The information herein is for research illustration. SYGNAL does not provide investment advice or personalised recommendations.

Regulatory Notices

United States (SEC RIA Disclosure)
SYGNAL is a brand of Sanostro AG, a Swiss company registered with the U.S. SEC as an Investment Adviser. Registration does not imply a certain level of skill or training. Under SEC Rule 206(4)-1, hypothetical or backtested performance information may only be presented when accompanied by appropriate disclosures and policies for audience suitability. Such information should not be relied upon as advice or an indication of future performance.

Switzerland
This document is for informational purposes only and does not constitute an offer, recommendation, or solicitation to invest. It is intended only for professional or institutional investors who understand the methodological limitations of such model-based analyses.

For full risk disclosures, refer to our Regulatory & Risk Disclosures.

Version control

This methodology reflects the production system (as deployed). All values and formulas correspond exactly to the current codebase parameters and logic.