For any of the strategy to qualify, for becoming part of live trading it has to move through a series of comprehensive testing procedures.

When we embarked on creating Trading Systems, our main aim was to create mathematical models which would help in profit maximization, however we gradually moved around to the view that best trading models are those that help in risk containment. Subsequently all our testing procedures were geared towards achieving the above objective.

Each of our products constitute of separate strategy portfolios of the analyst team, which in turn are composed of hundreds of independent models, working simultaneously, analyzing the market as per their algorithm. So for managing the over-all product risk, we follow a “Bottom up “approach. First the strategies are tested on different performance parameters and then their impact is analyzed at the portfolio and product levels .

Testing Procedures

Testing Performance at the strategy level

Any strategy is inherently a mathematical model, based on technical analysis and concepts of statistics. Once the beta version is prepared we try to assess the variability in its performance based on change in multiple performance criterion by changing different input parameters. This helps us in visualizing how the model would react in different market conditions.

Apart from percentage profitability and net profit, we also review different performance parameters  like Return on account , highest Drawdown as a percentage of Net profit, profit factor, Average winning trade to Average loosing trade ratio, average time between consecutive equity peaks. The maximum number of loosing trades in a row. The time required to recover from the highest Drawdown etc. Along with the above we also try to assess whether the bulk of the profits are coming from the Long trades or the Short trades. On meeting the threshold performance level for each of the above parameters we then analyze the spatial distribution of the profits. Apart from overall profitability for the entire period, we also try to ensure that each of the quarters are profitable on their own.

We then test the efficacy of the strategy on the unseen data. The smaller the variability in the performance the more robust the strategy is likely to be.

For testing at the strategy level we use software platform of Omega Research pro-suite 2000i.

Testing Performance at the portfolio level

Having tested the performance of the strategy, we then test the impact of the system on the analyst’s existing portfolio of strategies. We try to ensure that the portfolio is balanced based on overall number of short trades and long trades, that the long and short trades separately are profitable for all months in the development period. Since the strategies are based on price data of varying timeframes, the portfolios are tested over different time intervals to maintain 100% monthly profitability, for instance if we have a portfolio of 100 strategies developed over 30 minute timeframe and another 100 based on 60 minutes timeframe, both these portfolios need to be individually profitable.

The same testing procedure is subsequently carried out over the product level also to check out the impact at the product level. We also perform Monte Carlo analysis to create a probability – drawdown matrix which might result from the addition of this strategy at the portfolio and product level.

For testing at the portfolio and product level we use standard software platform of Rina and Mcs Pro (for Monte Carlo analysis). We also use in-house proprietary software like Portfolio Risk Evaluator, which helps us determining the top drawdowns the updated portfolio/ product would encounter with the addition of this strategy, on the unseen data.

In the next step, the same strategy is tested on real time data and paper trading is done for the next six months. The reports are analyzed on the same parameters as mentioned above.


The final step is to invest a small amount on the selected portfolio and trade for about four months to test the effectiveness of the designed system. The testing and slippage identification with relation to theoretical inputs is rigorously tested and the system results are compared with the actual trading results achieved during this period.

Once the reports are there the results generated are compared to the industry benchmarks. Care is taken to understand and scientifically evaluate the results using statistical and other technical tools.