Backtesting Event-Driven Strategies with a Portfolio NAV
Summary
The document considers how to evaluate strategies whose trades begin and end at irregular event times. It distinguishes measuring the average return around an event from measuring the cumulative performance of a strategy. For event analysis, the question proposes aligning observations by event date and averaging returns across events to examine how performance develops during a holding window.
The response recommends simulating a portfolio for cumulative results, allowing positions to have different entry and exit times and to be stopped before a planned horizon. It also argues that a return calculation requires starting capital: simulate daily net asset values, calculate daily portfolio returns, and link them through time. The response does not specify detailed rules for sizing, overlapping trades, transaction costs, or capital constraints, so those choices still need to be defined for a complete backtest.
Key ideas
- Event-aligned average returns can describe how performance evolves after a signal.
- Cumulative strategy performance is better represented by simulating a portfolio through time.
- Irregular trades and early exits can be modeled using daily portfolio net asset values.
- A cumulative return requires an initial capital base and linked daily returns.
- Position sizing, overlapping trades, costs, and capital limits need explicit treatment.
Tags
Full text
# How to backtest a strategy with irregular in-out signal? # How to backtest a strategy with irregular in-out signal? Hi I'm currently backtesting an event-driven strategies. Unlike factor strategy which has a regular rebalancing interval, event-driven strategy is conducted whenever there is an event. Since we do not know how many signals would pop up in the future in the past timepoint, I defined unit position size per one in-out betting. I think there can be 2 possible ways to backtest such strategy. - (I think this would be a common case) Align every event-driven signal into one single timeline and check the mean return of the event. For example, if I want to test buyback event driven strategy, regardless of the announcement of the buyback event, set announcement date as 0 and last holding moment as 20 (1-month holding strategy, for example). And average the returns per day and check how long does buyback alpha persist from the announcement date. - (This is the case where I get confused) I want to calculate cumulative return of this strategy. (Final amount of money I can earn by conducting this strategy) However, since there is no fixed seed money at the very first time of the backtest, how can I calculate the return? Currently, I'm tracking - unrealized return(evaluated return) - realized return - unit position price(number of currently held position * unit position size) - unit position price averaged and final return is calculated as (unrealized return + realized return) / unit position price averaged. But is it right way to calculate cumulative return of such strategy? How can I backtest cumulative return of a strategy without fixed seed money but flexible in-out position? ## Answer by Ilya Voytov (score 1) https://quant.stackexchange.com/a/71411 - Typically events like those last a quarter or some variable time period. Then your position may get stop lossed half way through. So instead of forcing an alignment, just simulate a portfolio. - Your strategy must start with some seed money, otherwise how can you calculate a return or do a trade? You start with some capital, and then simulate daily NAVs for your strategy and then link the daily returns to create a backyesr
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.