APO · Apollo Global Management Inc.
Stock Universe · global stock marketAPO · Quantitative Analysis of a High-Conviction Long-Horizon Algorithmic Strategy on 15-Minute Data
An in-depth statistical examination of the algorithmic trading strategy applied to APO stock data from September 2019 through January 2026. Operating on a 15-minute timeframe, the system exhibits an ultra-selective trade selection model, generating only 8 completed trades over 6.31 years. With a 100% historical win rate, a total simulated return of +2356.59%, an average holding period of 4263.19 hours, and zero realized closed-trade drawdown, the strategy demonstrates pristine equity curve characteristics. However, analysis reveals a heavily right-skewed profit distribution where the top 3 winning trades account for 81.78% of all gross gains, alongside significant trade clustering in recent months.
Analysis as of 18.09.2026. Strategy data is updated separately.Read full analysis
Strategy profile
The strategy under evaluation is an algorithmic trading model designed for equity markets, specifically applied to APO on a 15-minute intraday price timeframe. Spanning a backtested observation window from September 16, 2019, through January 5, 2026—a duration of 6.31 years—the algorithm serves as the decision logic for automated trading bots, trading robots, and quantitative trading agents seeking long-term trend exposure while utilizing intraday price granularity. Within the DevioLab quantitative taxonomy, this model achieves a DevioLab Score of 69.98, securing the rank of #2 for the APO symbol. The foundational attribute of this strategy is its extreme signal selectivity. Over more than six years of historical data, the system generated only 8 completed trades, achieving a perfect win rate of 100% with 8 winning outcomes and 0 losing outcomes. This flawless hit rate yielded a total cumulative simulated return of +2356.59%, which corresponds to an annualized performance metric of 189.72%. The strategy operates with a profit factor of 8.75 and an observed maximum closed-trade drawdown of 0.00%. Rather than engaging in active market timing or frequent intraday turnover, the algorithm functions as a patient, high-conviction filter designed to capture multi-month market expansions while completely abstaining from market exposure during ambiguous or unfavorable conditions.
Trading rhythm and position duration
Despite running on a 15-minute chart interval, the strategy operates with a position holding profile that aligns with macro position trading rather than intraday scalping or swing trading. The quantitative data reveals an average holding duration of 4263.19 hours, equivalent to approximately 177.6 days or nearly six months per trade. The median holding duration stands at 1897.00 hours, or roughly 79 days, indicating that even the shorter-duration trades remain active for multiple months. On an annualized basis, the strategy averages just 1.27 completed trades per year, with an active frequency of 1.00 trade per active month. Spacing between trade completions is similarly sparse, exhibiting an average of 219.86 days between exits and a median of 135.81 days. This structural combination creates a notable operational dynamic for automated trading agents executing the logic: while the 15-minute chart provides granular sub-daily price monitoring for entry and exit triggers, actual position turnover is exceptionally rare. Trading robots deployed with this logic spend the overwhelming majority of their operational life maintaining existing long positions or waiting in cash, entirely avoiding the execution friction and constant churn typical of intraday quantitative strategies.
Quality of historical results
Evaluating the payoff quality requires looking beyond the headline return of +2356.59% to examine how individual trade metrics are distributed across the dataset. The strategy produced an average trade return of +60.82%, compared against a median trade return of +20.82%. The best individual trade delivered a remarkable gain of +233.18%, whereas the worst completed trade still achieved a positive return of +11.34%. This substantial variance between average (+60.82%) and median (+20.82%) highlights a pronounced right-skewed pay-off structure. The skewness is further emphasized by concentration metrics: the top 3 winning trades generated 81.78% of the total gross profit accumulated across the entire 6.31-year evaluation period. This demonstrates that while every executed trade in the dataset finished in positive territory, the strategy's outsized cumulative performance was driven by a small cluster of exceptional outlier trades—most notably the single +233.18% position—rather than an even accumulation of modest gains across all trades.
Risk, drawdown and losing behavior
From a closed-trade risk perspective, the strategy records exemplary metrics, featuring a 0.00% maximum drawdown and a longest losing streak of 0 trades across its 8 completed executions. With zero losing trades recorded, the historical profit factor stands at 8.75, and the worst trade outcome was a positive +11.34% return. However, a rigorous quantitative assessment requires contextualizing what a 0.00% closed-equity drawdown truly represents. Because drawdown in this dataset is computed at trade exit, and every position closed with a positive net return, no realized equity retracement was registered upon trade completion. Nevertheless, given an average position duration of 4263.19 hours (~177 days), open positions were exposed to substantial intermediate market fluctuations over multi-month holding periods. For automated trading bots executing this model, the primary risk profile is not sudden closed-capital impairment, but rather unrealized open-position equity drawdowns and prolonged exposure to market volatility while waiting for the algorithm's exit criteria to trigger.
Behavior through time and yearly stability
An analysis of the yearly breakdown reveals significant multi-year asymmetry and trade concentration across the historical window. Out of the seven calendar years covered by the dataset (2019 through 2026), three full years—2019, 2020, and 2023—registered zero completed trades. When trading activity did manifest, trade frequency remained capped at one or two completed positions per calendar year. In 2021, the system executed 1 trade (100% win rate), capturing a massive +233.18% gain that served as the primary anchor for early strategy compounding. In 2022, performance moderated to 2 trades (100% win rate) yielding a combined return of +32.61%. Following a completely silent 2023, activity resumed in 2024 with 2 trades (100% win rate) producing +164.76% in sum. The system continued into 2025 with 2 trades (100% win rate) totaling +36.77%, and closed 1 trade in early 2026 (100% win rate) adding +19.27%. This temporal behavior underscores that performance is non-linear and highly episodic; performance gains arrive in concentrated surges separated by extended periods of inactivity.
Recent period since 2024-06-01 versus full history
Examining performance since June 1, 2024, provides vital context regarding the strategy's recent operational pattern relative to its multi-year baseline. Out of the 8 total trades completed across the entire 6.31-year backtest, 5 trades closed after June 1, 2024. These 5 recent trades produced a combined return sum of +220.80%, accounting for over 62% of all historical trade completions and a massive proportion of cumulative portfolio growth. This indicates a dramatic acceleration in trade completion frequency during the final 1.5 years of the test window compared to the 2019–2023 era, which saw only 3 completed trades in total. For quantitative analysts and automated trading agents evaluating recent performance, the post-June 2024 window demonstrates that the algorithm's exit logic has been repeatedly triggered by recent market trends in APO, confirming that the underlying signal filters remained highly active and aligned with recent market structure.
Strengths and limitations
The primary strength of this strategy lies in its historical precision and equity curve stability, characterized by a 100% win rate, a 0.00% closed-trade drawdown, a strong profit factor of 8.75, and a total return of +2356.59%. By maintaining extreme selectivity and holding positions for an average of 4263.19 hours, the model effectively eliminates overtrading, commission drag, and short-term market noise. Conversely, the strategy's primary limitation stems from its small statistical sample size of just 8 completed trades over 6.31 years. A sample of 8 trades, while historically flawless, carries inherent statistical uncertainty regarding future performance degradation. Furthermore, profit concentration is high, with 81.78% of gross gains generated by just 3 trades. Automated trading systems utilizing this logic must also tolerate multi-year periods of complete trading inactivity alongside open-position equity swings during six-month holding periods.
DevioLab analytical conclusion
With a DevioLab Score of 69.98 and a ranking of #2 for APO, this algorithmic model represents a highly specialized, low-frequency trend execution model. The strategy successfully demonstrates how 15-minute price data can be leveraged not for rapid scalping, but for building ultra-selective decision logic tailored for long-horizon position trading. For automated trading bots and trading agents, the logic offers a blueprint for capturing high-conviction macro movements while maintaining rigorous capital preservation on closed equity. Market participants evaluating this model should weigh its flawless historical hit rate and zero closed drawdown against its sparse trade volume and high winner concentration, recognizing that future realized performance will depend heavily on the occurrence of sustained macro trends in APO.
Data scope and methodology
This analysis is based strictly on historical backtested data for the strategy applied to APO on a 15-minute chart interval across the period from September 16, 2019, to January 5, 2026. All reported performance metrics, including the total return of +2356.59%, win rate of 100%, holding times, and drawdown figures, represent simulated closed-trade statistics. The dataset reflects completed trades where position exits occurred within the specified timeframe. These results are compiled for quantitative research and software development purposes for automated trading systems and do not account for live broker execution variables such as slippage, variable order execution fees, or overnight margin rates, nor do they guarantee future financial performance.