Working with Strategies
1. Use ichimoku()
To create an ichimoku object from the price data.
# Simulated OHLC pricing data is assigned to data frame 'TKR':
TKR <- sample_ohlc_data
cloud <- ichimoku(TKR)2. Use strat()
To test a simple strategy with a rule condition of the form ‘long (or short) while c1 > c2’.
Here the references c1 and c2 are columns of the ichimoku object representing either candlestick values e.g. ‘close’ or cloud values e.g. ‘tenkan’.
Note: the rule condition is c1 > c2 even for short trades. Hence please take care for example when inverting ‘long while close > tenkan’ to ‘short while close < tenkan’, this should be formulated as ‘short while tenkan > close’.
strat <- strat(cloud)
print(strat[100:105], plot = FALSE)
#> open high low close cd tenkan kijun senkouA senkouB chikou
#> 2020-04-27 122.7 122.7 121.8 122.4 -1 120.50 123.35 125.625 124.85 135.6
#> 2020-04-28 122.4 122.6 121.1 121.9 -1 120.50 123.35 125.675 124.85 134.5
#> 2020-04-29 121.9 123.7 121.7 123.3 1 120.85 123.25 125.675 124.85 134.3
#> 2020-04-30 123.3 124.0 123.0 124.0 1 121.00 123.25 125.675 124.85 135.9
#> 2020-05-01 124.2 124.3 124.0 124.1 -1 121.40 123.10 125.675 124.85 135.2
#> 2020-05-03 124.1 124.1 123.4 123.9 -1 122.25 123.10 125.975 124.85 135.7
#> cloudTop cloudBase cond posn txn logret slogret
#> 2020-04-27 125.625 124.85 1 1 0 -0.0024479816 -0.0024479816
#> 2020-04-28 125.675 124.85 1 1 0 -0.0040933336 -0.0040933336
#> 2020-04-29 125.675 124.85 1 1 0 0.0114193737 0.0114193737
#> 2020-04-30 125.675 124.85 1 1 0 0.0072727593 0.0072727593
#> 2020-05-01 125.675 124.85 1 1 0 -0.0008054773 -0.0008054773
#> 2020-05-03 125.975 124.85 1 1 0 -0.0016129036 -0.0016129036
#> ret sret
#> 2020-04-27 -0.002444988 -0.002444988
#> 2020-04-28 -0.004084967 -0.004084967
#> 2020-04-29 0.011484824 0.011484824
#> 2020-04-30 0.007299270 0.007299270
#> 2020-05-01 -0.000805153 -0.000805153
#> 2020-05-03 -0.001611604 -0.001611604The use of strat() returns an augmented ichimoku object. The printout above of a slice of the object (rows 100 to 105) shows that the following columns are appended:
- cond: ‘1’ if the rule condition is met at the close of the the time period (note: if the rule involves chikou span, the comparison will be of past periods)
- posn: ‘1’ represents that a position is held during this time period (i.e. from the ‘open’ to the ‘close’)
- txn: ‘1’ represents a transaction to initiate a new position, ‘-1’ represents a transaction to square an existing position. All transactions are assumed to happen at the ‘open’ of the respective period
- logret: the log returns from the ‘open’ price of the period to the ‘open’ price of the next period. For the avoidance of doubt, ‘log returns’ means the natural logarithm of the difference between the prices
- slogret: the log returns for the strategy. It is simply logret * posn, i.e. the returns where a position is held
- ret: the discrete returns from the ‘open’ price of the period to the ‘open’ price of the next period. For the avoidance of doubt, this is calculated as e raised to the power of ‘logret’ minus 1
- sret: the discrete returns for the strategy. It is simply ret * posn, i.e. the returns where a position is held
Note: the assumption applies to all strategies that confirmation of whether a condition is satisfied is received at the ‘close’ of a particular period, and a transaction is initiated at the immediately following ‘open’. It is assumed that all transactions are conducted at opening prices. All events are aligned to their correct time periods, which are taken to be from the ‘open’ to the ‘close’ inclusive of that period. Hence ‘cond’ refers to if the condition is satisfied at the ‘close’ of the period, ‘posn’ also refers to whether a position is held at the ‘close’ of the period, and ‘txn’ refers to if there has been a transaction during that period (in this case occurring at the ‘open’).
3. View strategy summary with summary()
A summary of the strategy is saved as an attribute to the ichimoku object and can be accessed by the summary() function.
summary(strat)
#> [,1]
#> Strategy "close > tenkan"
#> --------------------- "----------"
#> Strategy cuml return % 5.01
#> Per period mean ret % 0.0218
#> Periods in market 119
#> Total trades 18
#> Average trade length 6.61
#> Trade success % 27.78
#> Worst trade ret % -2.54
#> --------------------- "----------"
#> Benchmark cuml ret % 3.38
#> Per period mean ret % 0.0148
#> Periods in market 224
#> --------------------- "----------"
#> Direction "long"
#> Start 2020-01-13
#> End 2020-09-29
#> Ticker "TKR"The summary is meant to provide a quick overview of whether a strategy is effective / desirable:
- Strategy returns are compared with the benchmark, which is simply being in the market the entire time
- Comparing returns combined with total periods in the market can help determine if the strategy is substantially different from just taking market risk
- Total trades along with average trade length can be used to filter out undesirable strategies that over-trade
- Trade success % helps to characterise the consistency of the strategy or whether it relies on a small percentage of outsize returns
- The return of the worst-performing trade provides a measure of downside risk
See strategy summary specification for details of the reported measures.
4. Visualize by calling plot().
plot(strat)
The periods where the strategy results in a market position are now shaded on the ichimoku cloud chart. To view the original chart without the strategy, simply pass the argument strat = FALSE when calling plot().
Note: iplot() interactive charting currently does not implement this additional functionality for strategies.
5. Combine strategies using stratcombine()
Strategies may be combined to form more complex strategies if desired. The function can be used on 2 ichimoku objects containing rule conditions ‘c1 > c2’ and ‘c3 > c4’ respectively, combining them into a new strategy with a rule condition of ‘c1 > c2 & c3 > c4’.
strat2 <- strat(cloud, "cloudBase", "kijun")
newstrat <- stratcombine(strat, strat2)
summary(newstrat)
#> [,1]
#> Strategy "close > tenkan & cloudBase > kijun"
#> --------------------- "----------"
#> Strategy cuml return % 10.64
#> Per period mean ret % 0.0653
#> Periods in market 32
#> Total trades 3
#> Average trade length 10.67
#> Trade success % 66.67
#> Worst trade ret % -1.4
#> --------------------- "----------"
#> Benchmark cuml ret % 0.94
#> Per period mean ret % 0.006
#> Periods in market 155
#> --------------------- "----------"
#> Direction "long"
#> Start 2020-04-02
#> End 2020-09-29
#> Ticker "TKR"6. Further backtesting and returns analysis
The strategy returns are saved within the ichimoku object in the columns ‘slogret’ (strategy log returns) and ‘sret’ (strategy discrete returns).
The returns (benchmark) for all periods are saved as ‘logret’ (log returns) and ‘ret’ (discrete returns).
As the ichimoku object inherits the ‘xts’ class, these columns may be fed directly into other econometrics or analytics packages such as ‘PerformanceAnalytics’, as per the example below.
library(PerformanceAnalytics)
# To chart performance comparison of strategy vs benchmark, daily returns and drawdowns
charts.PerformanceSummary(strat[ ,c("sret", "ret")])
# For a table detailing drawdowns, including depth, length of recovery etc.
table.Drawdowns(strat[, "sret"])