I've got your back

I've got your back Why backtesting should play a bigger role in trading
Monday, 8. March 2021 08.03.2021 von Dr. Jochen Tackenberg 0 Comment

Backtesting is a standard tool in the development of trading strategies on the financial market, however, it is often underutilized in energy trading. There are good reasons for this, and yet backtesting should not be completely ignored. Why? A brief assessment.

What is backtesting?

Backtesting is carried out to evaluate and test new ideas and strategies. Wikipedia says: "Backtesting is the process of evaluating a strategy, theory, or model by applying the strategy, theory, or model to historical data."

That is to say, the strategy is provided with the data it needs from what was already available/known at a point in the past. From there, data is added for the future with respect to how it may occur in reality. From this, the strategy decides in the same way as it would have done in the corresponding historical situation. For example, orders can be executed directly or own orders can be set. The aim is to reproduce the real conditions as realistically as possible. The results of the strategy actions are then evaluated; in this way, for example, one can compare financial benchmarks with the results of other strategies.

What should be considered?

Backtesting is based on the assumption that the future will behave similarly to the past. For individual days, this statement is obviously wrong because every day is different. Therefore, it is important to simulate as large a time period as possible when backtesting. In contrast to the financial market, the electricity market changes very quickly - be it due to changes in regulations and market conditions (examples are the introduction of xbid/sidc, sdat or the standard labour market) or due to changes in the market/the market participants (increase in renewables, shutdown of nuclear power plants, changes in strategies/digitisation). Therefore, older data quickly becomes unrepresentative. It is not always easy to find a suitable trade-off here.

Especially for the marketing of EEG quantities, it is important that the traded quantities match the corresponding day that is simulated. Since the forecast deviations are similar for all marketers, we speak of market-correlated quantities: if you have to sell, everyone else has to sell too - and vice versa. So, to test one's strategy, real forecast deviations should be simulated with the matching days.

Limitations

Backtesting is a simulation with "outdated" data. This leads to the following difficulties:

  • As a rule, historical market participants are static:
    • The other market participants do not react to one's own orders: One is not clicked away, even if one stands directly in front of the other order book page.
    • Other orders on one's own side do not react to one's own orders. For example, there is no build-up due to one's own behaviour.
    • The order book is not "refilled"; if quantities are executed, this has no influence on the prices or the quantities shown.
  • The continuous intraday market for electricity is only liquid to a limited extent. Illiquid markets are particularly difficult to replicate, as individual actions may have a large impact on the market.
  • Every day is different, which is why there is a danger of over-optimisation (overfitting): You may find a strategy with which you are
  • Special situations are (often) deliberately excluded. However, this also means that these must be intercepted by other mechanisms so that no economic damage can occur.
  • Every test is only as good as the data with which it is made. Both exchange data and position data must be historically complete and correct.
    One could put more or less effort into countering the individual points. The most important thing is to be aware of these difficulties and their implications. In this way, the right conclusions can be drawn from backtesting, and the method has added value and supports the search for the right strategy.

Application tips

So, what do I need to consider in order to benefit from the advantages of backtesting in electricity trading?

  • Due to the many restrictions in the electricity market, the results of backtesting runs and of strategies actually executed in the market are not comparable. To avoid false incentives, one should be aware that backtesting can both overestimate success, for example because the influence on the market was disregarded, or underestimate it, for example because the spread is always taken and orders are never clicked away.
  • Backtesting strategies that are to be compared with each other should follow similar approaches. In addition, one should be aware of which aspects are favoured and which are disadvantaged by the restrictions.
  • Backtesting is suitable for comparing different values for individual parameters.
  • According to parameter studies, the parameters that lie on a plateau are to be preferred: Similar input parameters lead to only small deviations in the result. For these values, the probability is high that small changes in the market will also lead to only small deviations in the result.
  • The basic strategy development should take place in the electricity market with intelligence and not with the mere comparison of backtesting results.
  • Also important: do not blindly trust backtesting and understand the productive phase as a test phase as well. If a strategy has proven itself in backtesting, it does not necessarily have to work in reality as well. Therefore, it is best to start with small quantities and perhaps not directly manage the entire portfolio with the new strategy. Instead, you should constantly monitor the results and thus be able to react quickly and try out new things.

Conclusion

Backtesting never delivers the one truth, but it can help to validate and improve one's own strategies. At the same time, the entry hurdle is high, because in addition to the data, software is needed to synchronously merge the information and simulate the market accordingly. Therefore, in iTrade we have tried to offer our users the easiest possible access to backtesting while maintaining the necessary flexibility. Backtesting in iTrade can be fully controlled via the UI. In doing so, we offer the following parameters to ensure the most comprehensive, simple and flexible handling possible:

  • Existing strategies or new strategies created in the UI can be tested.
  • New strategies can be added for testing via csv files.
  • The time range to be tested can be conveniently selected and adjusted
  • The quarter hours and hours to be simulated can be selected.

Those who do not want to use the UI here can create the strategies and jobs directly in the database. This provides a higher level of automation and the user can even set up automated parameter tuning. And to counteract the limitations mentioned above, iTrade also recognises iceberg orders in the order book in backtesting and is able to convert own trades into orders. In this way, one's own influence on the market is better taken into account.

write comment

+49 241 400 230

call now

Use of cookies

This site uses cookies for Matomo Analytics. You can accept or dismiss these cookies and change your decision any time.

Further information can be found in the privacy policy

Dismiss Accept
Cookie settings History

History

delete History close