Institutional Repository of Vilnius University
Prekybos strategijų finansų rinkose testavimas naudojant Monte Karlo simuliaciją ir hiperparametrų optimizaciją /
Abstract
dc:descriptionThis bachelor thesis investigates a methodology for testing trading strategies in financial markets, with the aim of optimizing strategy hyperparameters and evaluating the stability of the obtained results. The work focuses on the problem that a single historical backtest does not necessarily show whether a strategy would remain reliable under different market conditions. When many parameter combinations are tested, there is a higher risk of finding a parameter set that performs well only in a specific historical period, but does not remain stable on unseen data. The experiment uses 15-minute EURUSD candle data. A rule-based trading strategy is implemented, combining trend-following, momentum and breakout logic. The strategy uses EMA, MACD and ATR indicators, ATR-based SL and TP levels, an optional trailing stop mechanism and risk-based position sizing. Hyperparameter optimization is performed using the Optuna library, with a separate optimization study carried out for each walk-forward training window. The final historical evaluation is based on standard walk-forward testing: strategy parameters are selected only on the training data and then tested on the following unseen out-of-sample interval. In addition, Monte Carlo validation is performed using GARCH-based price scenarios. The thesis also evaluates sensitivity to execution conditions by applying spread stress tests. The experimental results showed that, in the baseline scenario, the combined out-of-sample return reached 124.84 percent, the maximum drawdown was -8.09 percent, the profit factor was 1.679 and the daily Sharpe ratio was 2.793. When the spread assumption was increased, the strategy remained profitable, but the results decreased significantly. Monte Carlo validation showed that the strategy remained profitable in most simulated scenarios, although the results depended on the chosen scenario generation method. These findings indicate that the proposed methodology can be used not only to find a profitable parameter set, but also to evaluate its stability, risk and sensitivity to execution assumptions.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bobin, Matas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308332383&prefLang=en_US