Hyperopt Notes

GUARDS (can have multiple)

• Define a favorable range (but don't act yet, need to wait for trigger)
• Examples:
• RSI > 20
• fisher_RSI >0.3
• To test, insert into:
•  "populate_sell_trend":
• # GUARDS AND TRENDS
• if params.get('sell-GUARD-enabled'):                                  
•                 conditions.append(dataframe['GUARD'] > params['sell-GUARD-value'])  
• "sell_indicator_space":
• Integer(75, 100, name='sell-GUARD-value'), or Real(-0.12, +1.40, name='sell-GUARD-value'),
• Categorical([True, False], name='sell-GUARD-enabled'),
• Note: Above is for sell signal. For buy signal, look at "populate_buy_trend", "indicator_space", and delete "sell-".

TRIGGER (can have only 1 single or 1 combination)

• act (buy/sell) now, provided the range (Guard) is favorable.
• examples:
• closing price cross above/below an indicator (e.g. bollinger band, parabolic sar)
• a pattern emerges (e.g. CDLHAMMER)
• if want to find the single best trigger:
• buy_strategy_generator > populate_buy_trend
• # TRIGGERS
• if 'trigger1' in params:
• if params['trigger1'] == 'bb_lower':
•                     conditions.append(dataframe['close'] < dataframe['bb_lowerband'])
• if params['trigger1'] == 'macd_cross_signal':
•                     conditions.append(qtpylib.crossed_above(dataframe['macd'], dataframe['macdsignal']))
• indicator_space
• Categorical(['bb_lower', 'macd_cross_signal'], name='trigger1')
• if want to use multiple triggers (to trigger 1 action) (and does not evaluate which trigger is better):
• (dataframe['bb_lowerband'] > dataframe['close']) & 
• (dataframe['CDLHAMMER'] == 100)
• How to pick the best-in-class within multiple triggers. Assume, for a buy signal, wanna pick the best EMA crossover (call it trigger1) and the best HA range (call it trigger2). So:
• buy_strategy_generator > populate_buy_trend
• # TRIGGERS
• if 'trigger1' in params:
•                 if params['trigger1'] == 'ema3x5':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema5']))
•                 if params['trigger1'] == 'ema3x10':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema10']))
• if 'trigger2' in params:
•                 if params['trigger2'] == 'ha_close>ema3':
•                     conditions.append(dataframe['ha_close'] > dataframe['ema3'])
•                 if params['trigger2'] == 'ha_close>ema5':
•                     conditions.append(dataframe['ha_close'] > dataframe['ema5'])
• indicator_space
• Categorical(['ema3x5', 'ema3x10'], name='trigger1'),
• Categorical(['ha_close>ema3', 'ha_close>ema5'], name='trigger2'),
• For signals which involve "OR" (compared to the usual "AND")
• focus on setting up individual conditions correctly first (using "AND")
• can be verified with a backtest
• finally, introduce the "OR" 
• cannot be verified with a backtest

How to read Categorical results

• IF sell_indicator_space :
• Categorical([True, False], name='sell-sar_reversal-enabled'),                       
• Categorical([True, False], name='sell-macd_cross_signal-enabled'),                  
• Categorical(['sell-macd_cross_signal', 'sell-sar_reversal'], name='sell-trigger3')  
• CASE1
• Hyperopt result:
• 'sell-macd_cross_signal-enabled': True,
• 'sell-sar_reversal-enabled': True,
• 'sell-trigger3': 'sell-macd_cross_signal'
• This means:
• turn on macd
• turn on sar
• so that your (verification) backtest will be accurate.
• CASE2
• Hyperopt result:
• 'sell-macd_cross_signal-enabled': False,
• 'sell-sar_reversal-enabled': True,
• 'sell-trigger3': 'sell-macd_cross_signal'
• THIS MEANS:
• turn on macd
• turn on sar
• so that your (verification) backtest will be accurate.
• CASE3
• Hyperopt result:
• 'sell-macd_cross_signal-enabled': False,
• 'sell-sar_reversal-enabled': True,
• 'sell-trigger3': 'sell-sar_reversal'
• THIS MEANS:
• turn off macd
• turn on sar
• so that your (verification) backtest will be accurate.
• CONCLUSION: 
• If an indicator is labelled either "True" or a "sell-trigger", then turn it on. Ignore any "False" or "not a sell-trigger" label that exists simultaneously for that indicator.

Optimizing Bollinger bands (confirmed works? yes)

• strategy file
• populate_indicators
• bollinger2 = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=2)
•         dataframe['bb_lowerband2'] = bollinger2['lower']
•         dataframe['bb_middleband2'] = bollinger2['mid']
•         dataframe['bb_upperband2'] = bollinger2['upper']
• bollinger3 = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=3)
•         dataframe['bb_lowerband3'] = bollinger3['lower']
•         dataframe['bb_middleband3'] = bollinger3['mid']
•         dataframe['bb_upperband3'] = bollinger3['upper']
• bollinger4 = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=4)
•         dataframe['bb_lowerband4'] = bollinger4['lower']
•         dataframe['bb_middleband4'] = bollinger4['mid']
•         dataframe['bb_upperband4'] = bollinger4['upper']
• populate_buy_trend (if required by strategy)
• (dataframe['close'] < dataframe['bb_lowerband2'])
• populate_sell_trend (if required by strategy)
• (dataframe['close'] > dataframe['bb_lowerband1'])
• hyperopt file
• buy side (if required by strategy)
• buy_strategy_generator > populate_buy_trend
• # TRIGGERS
• if 'trigger' in params:
•                 if params['trigger'] == 'bb_lower2':
•                     conditions.append(dataframe['close'] < dataframe['bb_lowerband2'])
•                 if params['trigger'] == 'bb_lower3':
•                     conditions.append(dataframe['close'] < dataframe['bb_lowerband3'])
•                 if params['trigger'] == 'bb_lower4':
•                     conditions.append(dataframe['close'] < dataframe['bb_lowerband4'])
• indicator_space
• Categorical(['bb_lower2', 'bb_lower3', 'bb_lower4'], name='trigger')
• sell side (if required by strategy)
• sell_strategy_generator > populate_sell_trend
• (similar to populate_buy_trend)
• sell_indicator_space
• (similar to indicator_space)

Optimising RSI (confirmed works? yes)

• strategy file
• populate_indicators
• dataframe['rsi'] = ta.RSI(dataframe)
• populate_buy_trend (if required by strategy)
• (dataframe['rsi'] < 20) & (use guard if inequality?)
• populate_sell_trend (if required by strategy)
• (qtpylib.crossed_above(dataframe['rsi'], 98)) & (use trigger if cross?)
• hyperopt file
• buy side (if required by strategy)
• buy_strategy_generator > populate_buy_trend
• # GUARDS AND TRENDS
• if params.get('rsi-enabled'):
•                 conditions.append(dataframe['rsi'] < params['rsi-value'])
• indicator_space
• Integer(5, 50, name='rsi-value'),
• Categorical([True, False], name='rsi-enabled'),
• sell side (if required by strategy)
• sell_strategy_generator > populate_sell_trend
• #TRIGGERS
• if params.get('sell-rsi-enabled'):
•                 conditions.append(dataframe['rsi'] > params['sell-rsi-value'])
• sell_indicator_space
• Integer(60, 100, name='sell-rsi-value'),
• Categorical([True, False], name='sell-rsi-enabled'),

Optimising Stochastic Indicator's "SlowK"  (confirmed works? yes)

• strategy file
• populate_indicators
• stoch = ta.STOCH(dataframe)
• dataframe['slowd'] = stoch['slowd']
• dataframe['slowk'] = stoch['slowk']
• populate_buy_trend (if required by strategy)
• (dataframe['slowk'] < 20) &
• hyperopt file
• buy side (if required by strategy)
• buy_strategy_generator > populate_buy_trend
• # GUARDS AND TRENDS
• if params.get('slowk-enabled'):
•                 conditions.append(dataframe['slowk'] < params['slowk-value'])
• indicator_space
• Integer(0, 50, name='slowk-value'),
• Categorical([True, False], name='slowk-enabled'),

Optimising CDLHAMMER (CanDLestick Hammer) (confirmed works? yes)

Ref1: https://www.ig.com/uk/trading-strategies/16-candlestick-patterns-every-trader-should-know-180615

Ref2: https://quantlabs.net/blog/2018/06/reliable-candlestick-patterns-for-ta-lib-pattern-recognition-acts-as-a-bullish-reversal/

• strategy file
• populate_indicators
• dataframe['CDLHAMMER'] = ta.CDLHAMMER(dataframe)
• populate_buy_trend 
• (dataframe['CDLHAMMER'] == 100) &
• hyperopt file
• buy side (if required by strategy)
• buy_strategy_generator > populate_buy_trend
• #GUARDS
• if params.get('CDLHAMMER-enabled'):
•                 conditions.append(dataframe['CDLHAMMER'] == 100)
• indicator_space
• Categorical([True, False], name='CDLHAMMER-enabled'),

Optimising Parabolic SAR (stop and reverse) (confirmed works? yes)

• strategy file
• populate_indicators
• dataframe['sar'] = ta.SAR(dataframe)
• populate_sell_trend
• (qtpylib.crossed_above(dataframe['sar'], dataframe['close'])) &
• hyperopt file
• buy side (if required by strategy) (untested)
• buy_strategy_generator > populate_buy_trend
• # TRIGGERS
• if 'trigger' in params:
•                 if params['trigger'] == 'sar_reversal':
•                     conditions.append(qtpylib.crossed_above(
•                         dataframe['close'], dataframe['sar']
•                     ))
• sell side (if required by strategy) (confirmed works? yes)
• sell_strategy_generator > populate_sell_trend
• #GUARDS
• if params.get('sell-sar_reversal-enabled'):
•                 conditions.append(qtpylib.crossed_above(dataframe['sar'], dataframe['close']))
• sell_indicator_space
• Categorical([True, False], name='sell-sar_reversal-enabled'),

Optimising Fisher RSI (confirmed works? yes)

• strategy file
• populate_indicators
• adj_rsi = 0.1 * (dataframe['rsi'] - 50)
• dataframe['fisher_rsi'] = (np.exp(2 * adj_rsi) - 1) / (np.exp(2 * adj_rsi) + 1)
• populate_sell_trend 
• (dataframe['fisher_rsi'] > 0.3) &
• hyperopt file
• sell side (if required by strategy) (confirmed works? yes)
• sell_strategy_generator > populate_sell_trend
• # GUARDS AND TRENDS
• if params.get('sell-fisher_rsi-enabled'):
•                 conditions.append(dataframe['fisher_rsi'] > params['sell-fisher_rsi-value'])
• sell_indicator_space (confirmed works? yes)
• Real(-0.99990920426, +0.99990920426, name='sell-fisher_rsi-value'),
• Categorical([True, False], name='sell-fisher_rsi-enabled'),
• After hyperopt process is completed: (1) Note the optimized "rsi" value in the hyperopt results (e.g. 'rsi-value': 30,); then (2) manually calculate the optimized "fisher_rsi" (i.e. fisher_rsi=-0.9640275801) (use https://www.desmos.com/scientific); then (3) update the value in "populate_sell_trend" (i.e. (dataframe['fisher_rsi'] > -0.9640275801) &)

Optimising EMA (exponential moving average) (confirmed works? yes)

• strategy file
• populate_indicators
• dataframe['ema3'] = ta.EMA(dataframe, timeperiod=3)
• dataframe['ema5'] = ta.EMA(dataframe, timeperiod=5)
• dataframe['ema10'] = ta.EMA(dataframe, timeperiod=10)
• dataframe['ema20'] = ta.EMA(dataframe, timeperiod=20)
• dataframe['ema50'] = ta.EMA(dataframe, timeperiod=50)
• dataframe['ema100'] = ta.EMA(dataframe, timeperiod=100)
• populate_buy_trend (if required by strategy)
• qtpylib.crossed_above(dataframe['ema20'], dataframe['ema50']) &
• populate_sell_trend (if required by strategy)
• qtpylib.crossed_above(dataframe['ema50'], dataframe['ema100']) &
• hyperopt file
• buy side (if required by strategy)
• buy_strategy_generator > populate_buy_trend
• # TRIGGERS
• if 'trigger' in params:
•                 if params['trigger'] == 'ema3x5':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema5']))
•                 if params['trigger'] == 'ema3x10':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema10']))
•                 if params['trigger'] == 'ema3x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema20']))
•                 if params['trigger'] == 'ema3x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema50']))
•                 if params['trigger'] == 'ema3x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema100']))
•                 if params['trigger'] == 'ema5x10':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema10']))
•                 if params['trigger'] == 'ema5x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema20']))
•                 if params['trigger'] == 'ema5x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema50']))
•                 if params['trigger'] == 'ema5x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema100']))
•                 if params['trigger'] == 'ema10x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema20']))
•                 if params['trigger'] == 'ema10x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema50']))
•                 if params['trigger'] == 'ema10x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema100']))
•                    if params['trigger'] == 'ema20x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema20'], dataframe['ema50']))
•                    if params['trigger'] == 'ema20x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema20'], dataframe['ema100']))
•                    if params['trigger'] == 'ema50x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema50'], dataframe['ema100']))
• indicator_space
• Categorical(['ema3x5', 'ema3x10', 'ema3x20', 'ema3x50', 'ema3x100', 'ema5x10', 'ema5x20', 'ema5x50', 'ema5x100', 'ema10x20', 'ema10x50', 'ema10x100', 'ema20x50', 'ema20x100', 'ema50x100'], name='trigger')
• sell side (if required by strategy)
• sell_strategy_generator > populate_sell_trend
• # TRIGGERS
• if 'sell-trigger' in params:
•                 if params['sell-trigger'] == 'ema3x5':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema5']))
•                 if params['sell-trigger'] == 'ema3x10':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema10']))
•                 if params['sell-trigger'] == 'ema3x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema20']))
•                 if params['sell-trigger'] == 'ema3x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema50']))
•                 if params['sell-trigger'] == 'ema3x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema3'], dataframe['ema100']))
•                 if params['sell-trigger'] == 'ema5x10':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema10']))
•                 if params['sell-trigger'] == 'ema5x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema20']))
•                 if params['sell-trigger'] == 'ema5x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema50']))
•                 if params['sell-trigger'] == 'ema5x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema5'], dataframe['ema100']))
•                 if params['sell-trigger'] == 'ema10x20':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema20']))
•                 if params['sell-trigger'] == 'ema10x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema50']))
•                 if params['sell-trigger'] == 'ema10x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema10'], dataframe['ema100']))
•                    if params['sell-trigger'] == 'ema20x50':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema20'], dataframe['ema50']))
•                    if params['sell-trigger'] == 'ema20x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema20'], dataframe['ema100']))
•                    if params['sell-trigger'] == 'ema50x100':
•                     conditions.append(qtpylib.crossed_above(dataframe['ema50'], dataframe['ema100']))
• sell_indicator_space
• Categorical(['ema3x5', 'ema3x10', 'ema3x20', 'ema3x50', 'ema3x100', 'ema5x10', 'ema5x20', 'ema5x50', 'ema5x100', 'ema10x20', 'ema10x50', 'ema10x100', 'ema20x50', 'ema20x100', 'ema50x100'], name='sell-trigger')