ACGL · Arch Capital Group Ltd.
Stock Universe · global stock marketACGL High-Precision Swing Strategy: Quantitative Analysis of a 90.41% Win Rate Algorithmic Trading Model
This quantitative study examines an algorithmic trading strategy for Arch Capital Group Ltd. (ACGL) backtested across a nearly seven-year historical period from September 2019 to September 2026. Operating on a 15-minute candle sampling timeframe, the strategy exhibits an exceptionally disciplined execution model characterized by infrequent entries, long multi-week position holdings, and a remarkably high win rate of 90.41% across 73 closed trades. Generating a total cumulative simulated return of +2967.72% with a profit factor of 2.50 and a maximum historical drawdown of 27.91%, the system achieves a DevioLab score of 71.90, ranking third among evaluated strategies for ACGL. This analysis delves into the structural relationship between its position duration, return distribution, drawdown profile, and annual stability.
Analysis as of 11.09.2026. Strategy data is updated separately.Read full analysis
Strategy profile
The evaluated quantitative strategy is an algorithmic trading system developed for the equity market asset ACGL (Arch Capital Group Ltd.). Evaluated on a 15-minute chart interval over a historical dataset spanning 6.99 years—from September 16, 2019, through September 11, 2026—the model serves as decision logic suitable for automated trading bots, trading agents, or trading robots executing systematic equities programs. Over this multi-year evaluation history, the strategy registered a total of 73 completed trades, achieving a cumulative profit sum of +2967.72% and an annualized equivalent return metric of 135.86%. The system achieved a DevioLab overall score of 71.90, placing it at rank 3 among strategies analyzed for the ACGL ticker. Rather than seeking hyper-active intraday turnover, the strategy operates as a highly selective swing trading model. It utilizes granular 15-minute price action to time market entry and exit decisions, while maintaining positions over substantial macro timeframes to capture extended directional moves in ACGL.
Trading rhythm and position duration
Despite relying on a 15-minute price interval for its underlying signal processing, the strategy maintains a patient, low-frequency execution profile. Across the 6.99-year backtest history, the model completed 73 trades, translating to an average of 10.44 trades per year or approximately 1.33 trades per active month. Position holding durations reflect a genuine multi-week swing architecture rather than short-term scalp trading. The average holding time for a completed position stands at 478.38 hours (approximately 19.9 days), while the median holding duration is 287.50 hours (roughly 11.98 days). The spacing between trade exits is similarly wide, with an average of 33.46 days between closed trades and a median interval of 27.72 days. This temporal spacing confirms that automated trading agents deploying this logic execute trades sparingly, avoiding over-trading during noisy or range-bound market regimes.
Quality of historical results
The performance metrics reveal a distinct trade profile driven by extraordinary trade selectivity and an exceptionally high hit rate. Out of 73 completed historical trades, 66 were closed in profit and 7 ended in a loss, resulting in a win rate of 90.41%. The overall profit factor of 2.50 indicates that total gross gains exceeded total gross losses by two and a half times. The average return per completed trade reached +5.26%, closely aligned with the median trade return of +4.80%. This tight alignment between average and median returns indicates that overall equity growth is not driven by isolated luck, but rather by consistent profit extraction across trades. Furthermore, the single best trade achieved a gain of +33.47%, while the top three winning trades collectively accounted for only 16.29% of total gross profit. This low profit concentration demonstrates that the system performance is broadly distributed across many winning cycles rather than dependent on a few extreme outliers.
Risk, drawdown and losing behavior
While the strategy boasts a 90.41% hit rate and an impressive longest winning streak of 20 consecutive trades, its risk profile features a notable asymmetry when losses do occur. The maximum historical drawdown experienced by the strategy was 27.91%, which corresponds exactly to its worst single trade performance of -27.91%. Across the entire 73-trade history, losing streaks were strictly limited, with the longest losing streak lasting just 1 trade. However, because a single worst loss reached -27.91%, individual stop events can offset multiple standard average wins of +5.26%. This creates a structural risk dynamic: while losing trades are rare (occurring in less than 10% of total executions), the financial impact of a rare adverse trade can be severe. Automated trading robots utilizing this strategy must be evaluated with full awareness that equity curve pullbacks occur in sharp, localized events rather than through prolonged series of losing trades.
Behavior through time and yearly stability
An examination of the yearly performance breakdown shows consistent profitability across distinct calendar periods, albeit with shifting trade frequencies. In 2020, the strategy recorded 15 trades with a 93.33% win rate (14 wins, 1 loss) and a cumulative sum of +131.21%. Activity slowed in 2021, generating 3 trades with a 100% win rate and a +32.82% gain. In 2022, 7 trades yielded an 85.71% win rate and +63.26% total return, followed by 2023 with 9 trades, an 88.89% win rate, and +40.57%. Trading activity expanded in 2024 and 2025, with both years logging 16 trades and identical 87.50% win rates (14 wins, 2 losses), delivering returns of +34.49% and +49.79% respectively. Through the partial year of 2026 up to September, the system completed 7 trades, all winning (100% win rate), for a total return of +32.06%. Across all recorded years, the strategy maintained annual win rates at or above 85.71%, demonstrating remarkable multi-year stability across different market conditions.
Recent period since 2024-06-01 versus full history
Analyzing performance from June 1, 2024 onward provides crucial insight into the strategy's recent behavior. In this window, the model executed 32 completed trades, generating a cumulative return sum of +94.88%. Notably, these 32 recent trades account for nearly 43.8% of the total 73 trades completed across the entire seven-year historical dataset. This indicates a marked increase in strategy execution frequency during the recent sample compared to earlier years like 2021 or 2022. Despite this higher trade density, recent performance remained robust and consistent with the long-term baseline, delivering significant positive returns. The presence of 32 fully closed recent trades provides a substantial statistical basis to confirm that the strategy's operational characteristics have remained effective in modern market conditions.
Strengths and limitations
The primary strength of this ACGL algorithmic model lies in its exceptionally high hit rate of 90.41% combined with strong gross profit distribution. With the top three winning trades representing only 16.29% of gross gains, the system does not rely on rare market windfalls. Its profit factor of 2.50 and long winning streaks (up to 20 consecutive trades) reflect an effective edge. Conversely, the principal limitation stems from trade frequency and drawdown severity. With an average of just 10.44 trades per year, capital may remain inactive for long stretches. Furthermore, the maximum drawdown of 27.91% matching the worst single trade loss underscores that when a signal fails, the downside impact can be substantial relative to the average trade gain of +5.26%. Operators of automated trading bots must account for this payoff ratio tension when configuring risk parameters.
DevioLab analytical conclusion
With a DevioLab score of 71.90 and a rank of 3 for the ACGL ticker, this 15-minute algorithmic strategy represents a compelling example of low-frequency, high-probability swing trading. The system successfully combines high entry selectivity with extended holding times averaging nearly 20 days, allowing it to navigate short-term equity noise while capturing major trends in ACGL. While the small total sample size of 73 trades over seven years warrants disciplined statistical expectation management, the consistency of high win rates across every individual calendar year from 2020 through 2026 reinforces the validity of the backtested logic. For systematic traders and developers deploying automated trading agents, the system offers a high-conviction decision framework, provided portfolio capital can tolerate low execution frequency and infrequent but sizable single-trade drawdowns.
Data scope and methodology
All statistics presented in this analysis are derived from backtested historical simulations of closed trades for ACGL on a 15-minute timeframe between September 16, 2019, and September 11, 2026. The dataset encompasses 73 completed trade events. Recent metrics evaluated from June 1, 2024 onward represent trades with closed timestamps on or after that date. Historical start dates reflect the beginning of available backtest coverage for this specific strategy model rather than the inception date of ACGL. Returns, drawdowns, and performance metrics assume strict algorithmic execution without leverage, accounting for closed trade percentages. Simulated historical performance serves analytical evaluation purposes and does not guarantee future live execution results.