资源加载中... loading...

Dynamic Strategy Analytics Tool

Author: ChaoZhang, Date: 2023-10-13 15:54:35
Tags:

Overview

The main idea of this strategy is to simulate real-time trading, collect weekly trading data, and present the statistics in a table for more intuitive review of the strategy’s performance. It can help us quickly evaluate the profit and loss of the strategy, identify periods of poor performance, and optimize the strategy accordingly.

Strategy Logic

  1. Set the start and end time for the calculation period.

  2. Set the precision of statistics and the number of weeks in each group.

  3. Simulate RSI strategy for entries and exits.

  4. Define variables for the statistics table.

  5. Calculate the result for the current period.

  6. If period changes and trading is enabled, record the time and result for this period.

  7. If it’s the last bar and trading is enabled, record the time and result for the current period.

  8. If period changes and trading is disabled, record the time and result for the previous period.

  9. Find highest and lowest period results.

  10. Render the statistics table.

  • Calculate total number of statistical periods first

  • Iterate through each period, render headers, time and results

  • Cumulatively calculate the result for each group

  • Color code positive and negative results

Advantage Analysis

  • Can observe weekly results in real time for quick strategy evaluation

  • Intuitive presentation of results for clear insights

  • Help identify periods of poor performance for strategy adjustment

  • Convenient to track cumulative gains for long-term strategies

  • Can compare trading styles across different time periods

  • Customizable precision and groups to meet different needs

  • Simple and clear code, easy to understand and extend

Risk Analysis

  • The strategy is based on RSI, which has inherent trend following limitations

  • Trading costs can significantly impact actual results

  • Backtest data may not reflect actual market conditions

  • Default capital in backtest may not match real account size

  • Avoid overfitting by blindly tuning parameters based on statistics

Can incorporate more indicators for trend and optimize entries and exits to improve basic RSI strategy. Use actual trading costs in live trading. Add randomness to capital size in backtest. Maintain skepticism instead of over-tuning based on statistics.

Optimization Directions

  • Consider adding stop loss to limit downside

  • Optimize RSI parameters like overbought and oversold levels

  • Try different trading frequencies like intraday vs. monthly holding

  • Incorporate more indicators for trend and timing

  • Add profit taking logic

  • Optimize statistical parameter settings

  • Expand to track multiple assets

Stops can better manage risk/reward. RSI tuning improves win rate. More indicators and frequencies make strategy robust. Statistical tuning highlights important data. Expanding to multiple assets gives complete view.

Summary

The goal is to collect periodic results for intuitive statistical visualization to quickly judge performance across time. This provides data to optimize strategies. Strengths include real-time weekly results, clarity and extensibility. Be wary of over-reliance and curve-fitting with statistical outputs. Use rationally along with core strategy logic for insights, not as basis for changes. Overall, convenient way to assess performance and crucial for optimization.


/*backtest
start: 2023-09-12 00:00:00
end: 2023-10-12 00:00:00
period: 3h
basePeriod: 15m
exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}]
*/

//@version=5
// strategy('Strategy weekly results as numbers v1', overlay=true, default_qty_type=strategy.percent_of_equity, default_qty_value=25, commission_type=strategy.commission.percent, commission_value=0.04)

after = input(title='Trade after', defval=timestamp('01 Jan 2019 00:00 UTC'), tooltip="Strategy will be executed after this timestamp. The statistic table will include only periods after this date.")
before = input(title='Trade before', defval=timestamp('31 Dec 2024 23:59 UTC'), tooltip="Strategy will be executes before this timestamp. The statistic table will include only periods before this date.")

statisticPrecision = input.int(title='Statistic precision', group='Statistic visualisation', defval=1, tooltip="Defines how many digits should be rendered in every statistic cell.")
statisticGroupSize = input.int(title='Statistic group size', group='Statistic visualisation', defval=12, tooltip="Defines how many cells should be in one group inside the statistic table.")

// determinet whether the starategy should be traded between the period
isTradeEnabled = true


// *******************************************************************************************
// Core strategy simulation logic
// *******************************************************************************************
// calculate rsi strategy emulation data
rsiEmulationData = ta.rsi(close, 7)
rsiEmulationCrossover = ta.crossover(rsiEmulationData, 70)
rsiEmulationCrossunder = ta.crossunder(rsiEmulationData, 30)

// entry loogic based on the rsi calculations
if (isTradeEnabled and rsiEmulationCrossover)
    strategy.entry('Long', strategy.long)
if (isTradeEnabled and rsiEmulationCrossunder)
    strategy.entry('Short', strategy.short)


// *******************************************************************************************
// Weekly statistics table
// *******************************************************************************************
// define statistic variables
var statisticTable = table(na)
var statisticPeriodTime = array.new_int(0)
var statisticPeriodResult = array.new_float(0)
var statisticIsLatestCalculated = bool(na)
var statisticResultHighest = float(na)
var statisticResultLowest = float(na)
var statisticColorGray = color.new(color.gray, transp = 60)
var statisticColorGreen = color.new(color.green, transp = 60)
var statisticColorRed = color.new(color.red, transp = 60)

// claculate current period result
barResult = not na(strategy.equity[1])
             ? (strategy.equity / strategy.equity[1] - 1) : 0
isPeriodChanged = not na(time[1]) and weekofyear(time) != weekofyear(time[1])
currentPeriodResult = 0.0
currentPeriodResult := not na(currentPeriodResult[1]) and not isPeriodChanged
                       ? ((1 + currentPeriodResult[1]) * (1 + barResult) - 1) : 0.0

// initialise highest and lowest results variables
statisticResultHighest := na(statisticResultHighest) ? currentPeriodResult : statisticResultHighest
statisticResultLowest := na(statisticResultLowest) ? currentPeriodResult : statisticResultLowest

// search for highest and lowest results
statisticResultHighest := currentPeriodResult > statisticResultHighest ? currentPeriodResult : statisticResultHighest
statisticResultLowest := currentPeriodResult < statisticResultLowest ? currentPeriodResult : statisticResultLowest

// new week while trade is active
if isPeriodChanged and isTradeEnabled
    timeCalculated = time - 1000 * 60 * 60 * 24 * 7
    resultCalculated = currentPeriodResult[1]
    statisticIsLatestCalculated := false

    array.push(statisticPeriodTime, timeCalculated)
    array.push(statisticPeriodResult, resultCalculated)

// latest bar while trade is active
if barstate.islast and isTradeEnabled
    timeCalculated = time - 1000 * 60 * 60 * 24 * (dayofweek(time) - 2)
    resultCalculated = currentPeriodResult

    array.push(statisticPeriodTime, timeCalculated)
    array.push(statisticPeriodResult, resultCalculated)

// new week after trade disabled
if isPeriodChanged and not isTradeEnabled and not na(statisticIsLatestCalculated) and not statisticIsLatestCalculated
    timeCalculated = time - 1000 * 60 * 60 * 24 * (dayofweek(time) + 5)
    resultCalculated = currentPeriodResult[1]
    statisticIsLatestCalculated := true

    array.push(statisticPeriodTime, timeCalculated)
    array.push(statisticPeriodResult, resultCalculated)

// render statistics table
if barstate.islast
    statisticLength = array.size(statisticPeriodResult)
    statisticTableSteps = math.floor(statisticLength / statisticGroupSize) + (statisticLength % statisticGroupSize != 0 ? 1 : 0)
    statisticTable := table.new(position.bottom_right, columns = statisticGroupSize + 2, rows = statisticTableSteps + 1, border_width = 1)

    // render headers
    for i = 0 to (statisticGroupSize - 1)
        statisticHeaderContent = str.tostring(i + 1)
        table.cell(statisticTable, 1 + i, 0, statisticHeaderContent, bgcolor = statisticColorGray)

    // render time points
    for i = 0 to (statisticTableSteps - 1)
        statisticPointContent = str.format("{0,date,medium}", array.get(statisticPeriodTime, i * statisticGroupSize))
        table.cell(statisticTable, 0, 1 + i, statisticPointContent, bgcolor = statisticColorGray)

    // render the result
    statisticResultCummulative = 0.0
    for i = 0 to (array.size(statisticPeriodTime) - 1)
        statisticColumn = 1 + i % statisticGroupSize
        statisticRow = 1 + math.floor(i / statisticGroupSize)

        statisticResult = array.get(statisticPeriodResult, i)
        statisticResultCummulative := (i % statisticGroupSize == 0) ? 0.0 : statisticResultCummulative
        statisticResultCummulative := (1 + statisticResultCummulative) * (1 + statisticResult) - 1

        statisticResultColor = statisticResult > 0 ? statisticColorGreen : statisticColorRed
        table.cell(statisticTable, statisticColumn, statisticRow, str.tostring(math.round(statisticResult * 100, statisticPrecision)), bgcolor = statisticResultColor)

        // if it is the last item of the row or data array
        isStatisticLastOfTheRow = ((i + 1) % statisticGroupSize) == 0
        isStatisticLastOfTheData = i == (statisticLength - 1)
        if (isStatisticLastOfTheRow or isStatisticLastOfTheData)
            resultsTableCummulativeCellColor = statisticResultCummulative > 0 ? statisticColorGreen : statisticColorRed
            resultsTableCummulativeCellContent = str.tostring(math.round(statisticResultCummulative * 100, statisticPrecision))
            table.cell(statisticTable, 1 + statisticGroupSize, statisticRow, resultsTableCummulativeCellContent, bgcolor = resultsTableCummulativeCellColor)

More