Research

Methodology

How our algorithmic-trading research frameworks work — what we measure, how we test, and what we publish.

What We Publish

WolfSignal is a research-and-frameworks platform, not a brokerage or advisory service. We publish three things:

• Source code — the Python implementations of every strategy and combiner, under a personal-use license. Subscribers can read, run, and study every line.
• Research notes — daily and long-form analysis of how each framework's signals are reading the current market regime.
• Backtests & live results — full statistical disclosure of historical and live performance across multiple market regimes (bull, bear, COVID, post-COVID, current).

Research Frameworks

WhiteWolf — Equities (TQQQ / SQQQ / BIL)

A regime-detection engine combining seven indicator signals into a composite score. Three published combiner versions:

• V1 — original momentum-weighted combiner, ~31.94% CAGR over 3.5yr backtest
• V2 (production) — volatility-adaptive position sizing, ATR-based trailing stops on SQQQ, asymmetric risk profile. Sharpe 1.06 over 6yr backtest.
• V3 — composite-gated entries, lowest drawdown profile, designed for risk-averse compounding

Live on Alpaca with real capital since 2024.

BlackWolf — Crypto (BTC)

MarketCipher v2-based ensemble with five entry strategies (sustained-signal, range-bound VWAP scalping, fast-momentum dip-buying, and two compositional variants). Vol-targeted position sizing, runs 24/7. Backtest Sharpe 1.97.

MoonWolf — Prediction Markets (Kalshi)

Weather and prediction-market signal research using NWS / Open-Meteo / ECMWF model outputs to score Kalshi event contracts. Currently in research / paper-trading mode. Long-form methodology to follow.

Backtesting Approach

All frameworks are backtested under consistent rules:

• Period: 2011-02-11 through 2026-02-22 (15 years), using TQQQ since inception with synthetic 3× pre-inception fills where required (clearly disclosed).
• Data: daily OHLCV from official sources, including BIL for cash-equivalent return during flat periods.
• Slippage and fees: conservative assumptions; per-trade transaction cost included.
• Survivorship bias: none — TQQQ has been continuously listed throughout the period.
• Out-of-sample testing: strategies were defined before deployment; live trading since 2024 is the out-of-sample window.
• Monte Carlo: 10,000-run resampling with replacement on daily returns to test return-distribution stability.

Performance Metrics We Report

• CAGR — compound annual growth rate over the backtest period
• Max Drawdown — largest peak-to-trough decline
• Sharpe Ratio — risk-adjusted return (excess return / volatility), risk-free rate matched to period
• Sortino Ratio — downside-deviation-adjusted return
• Calmar Ratio — CAGR / max drawdown
• Profit Factor — gross profit / gross loss
• Win Rate & Trade Frequency — for context, not as primary measures

All metrics are published at multiple time horizons (1yr, 2yr, 3.5yr, 5yr, 6yr, 15yr) and across regime windows (COVID, 2022 bear, 2023-24 bull) so subscribers can judge framework behavior under different conditions.

What We Don't Do

• We don't manage subscriber money — you trade your own account.
• We don't guarantee performance — past results, including live results, do not predict future results.
• We don't tailor research to individual circumstances — content is general-audience educational.
• We don't place trades on subscribers' behalf.
• We don't accept commissions or referral fees from brokers.

Open Code

Foundational strategy code is published openly on GitHub under the MIT License. Paid-tier source code (combiners, BlackWolf, MoonWolf, the Research Dashboard) is delivered to Wolf Pack and Apex subscribers under a personal-use license — read, run, modify, and fork for your own trading; no redistribution.

For full legal disclaimers, see /disclaimers.html. For terms of service, see /terms.html.