Course Name: Value Strategy in Forex, Section No: 4, Unit No: 4, Unit type: Notebook
Question 1
In this notebook, the following code is illustrated to compute the equity curves of the performance for 8 currencies:
currency_returns = currency_price.pct_change()
strategy_returns = currency_returns * signal.shift(11)
. . .
# Plot the strategy returns
daily_ret = leverage * strategy_returns.sum(axis=1) / 8 # there are 8 currencies
daily_ret.cumsum().plot(figsize=[15, 7]) # plots 8 equity curves for 8 currencies
daily_ret = leverage * strategy_returns.sum(axis=1) / 8
daily_ret.cumsum().plot(figsize=[15, 7])
Should we not use cumulative products to evaluate returns like so?
(1 + daily_ret).cumprod()
I suppose the cumulative sum would be correct only if we dealt with log returns.
Question 2
The strategy as presented in the notebook uses 500% leverage, which is very aggressive. In fact, using no leverage (leverage =1), I obtain the following results:
maximum_drawdown: -2.70%
Sharpe Ratio: 0.58
CAGR: 2.1
This is very bad, particularly since we have not included any transaction costs. I do wonder whether and how the REER strategy can be improved. I think it would not survive live trading in its current form when market frictions and funding costs were included.
Hi Benjamin
To answer your queries-
Ans1:You are absolutely correct in your understanding. Since the notebook uses pct_change() , the returns computed are simple returns. In such a case, the mathematically correct way to construct the equity curve is by compounding as you mentioned correctly.
Using cumsum() would be strictly appropriate only if we were working with log returns. While the difference may be small when daily returns are relatively small as the case here, but your observation is conceptually accurate and important from a theoretical standpoint.
To avoid any confusion for learners, we will update the notebook accordingly.
We really appreciate you pointing this out, questions like yours help us improve the course quality for everyone.
Ans2: Your observation is absolutely valid. The 500% leverage in the notebook is mainly for demonstration, to make the return profile more visible. At 1x leverage, the raw returns are modest, and incorporating transaction and funding costs would likely reduce them further.
REER-based value strategies are typically slow-moving and are often used within diversified, multi-factor portfolios rather than as standalone strategies. Their performance can potentially be improved by adding volatility-based position sizing, combining value with momentum or carry filters, applying cross-sectional ranking, and explicitly modelling transaction costs.
This notebook is meant to illustrate the concept and implementation logic, not a fully production-ready trading system.
We appreciate your critical thinking and this is exactly the approach we encourage when evaluating strategies.