Ranking and Position Sizing

In this notebook, we will learn about the features of PyBroker that enable you to rank ticker symbols and set position sizes for a group of symbols in your trading strategy. With these features, you can easily optimize your strategy and manage risk more effectively.

import pybroker
from pybroker import Strategy, StrategyConfig, YFinance

pybroker.enable_data_source_cache('ranking_and_pos_sizing')

<diskcache.core.Cache at 0x7fd17427dd60>

Ranking Ticker Symbols

In this section, we will learn about how to rank ticker symbols when placing buy orders. Let’s begin with an example of how to rank ticker symbols based on volume when placing buy orders.

def buy_highest_volume(ctx):
    # If there are no long positions across all tickers being traded:
    if not tuple(ctx.long_positions()):
        ctx.buy_shares = ctx.calc_target_shares(1)
        ctx.hold_bars = 2
        ctx.score = ctx.volume[-1]

The buy_highest_volume function ranks ticker symbols by their most recent trading volume and allocates 100% of the portfolio for 2 bars. The ctx.score is set to ctx.volume[-1], which is the most recent trading volume.

config = StrategyConfig(max_long_positions=1)
strategy = Strategy(YFinance(), '6/1/2021', '6/1/2022', config)
strategy.add_execution(buy_highest_volume, ['T', 'F', 'GM', 'PFE'])

To limit the number of long positions that can be held at any time to 1, we set max_long_positions to 1 in the StrategyConfig. In this example, we add the buy_highest_volume function to the Strategy object and specify the ticker symbols to trade: ['T', 'F', 'GM', 'PFE'].

result = strategy.backtest()
result.trades

Backtesting: 2021-06-01 00:00:00 to 2022-06-01 00:00:00

Loading bar data...
[*********************100%***********************]  4 of 4 completed
Loaded bar data: 0:00:00

Test split: 2021-06-01 00:00:00 to 2022-05-31 00:00:00

100% (253 of 253) |######################| Elapsed Time: 0:00:00 Time:  0:00:00

Finished backtest: 0:00:02

	type	symbol	entry_date	exit_date	entry	exit	shares	pnl	return_pct	agg_pnl	bars	pnl_per_bar	stop
id
1	long	F	2021-06-02	2021-06-04	14.85	16.13	6734	8619.52	8.62	8619.52	2	4309.76	bar
2	long	F	2021-06-07	2021-06-09	15.93	15.51	6801	-2856.42	-2.64	5763.10	2	-1428.21	bar
3	long	F	2021-06-10	2021-06-14	15.43	15.06	6832	-2527.84	-2.40	3235.26	2	-1263.92	bar
4	long	F	2021-06-15	2021-06-17	14.96	14.99	6900	207.00	0.20	3442.26	2	103.50	bar
5	long	F	2021-06-18	2021-06-22	14.61	14.96	7003	2451.05	2.40	5893.31	2	1225.53	bar
...	...	...	...	...	...	...	...	...	...	...	...	...	...
80	long	F	2022-05-10	2022-05-12	13.43	12.47	7263	-6972.48	-7.15	-9423.13	2	-3486.24	bar
81	long	F	2022-05-13	2022-05-17	13.25	13.34	6835	615.15	0.68	-8807.98	2	307.58	bar
82	long	F	2022-05-18	2022-05-20	13.03	12.59	6739	-2965.16	-3.38	-11773.14	2	-1482.58	bar
83	long	F	2022-05-23	2022-05-25	12.72	12.57	6936	-1040.40	-1.18	-12813.54	2	-520.20	bar
84	long	F	2022-05-26	2022-05-31	12.99	13.59	6711	4026.60	4.62	-8786.94	2	2013.30	bar

84 rows × 13 columns

Setting Position Sizes

In PyBroker, you can set position sizes based on multiple tickers. To illustrate this, let’s take a simple buy and hold strategy that starts trading after 100 days and holds positions for 30 days:

def buy_and_hold(ctx):
    if not ctx.long_pos() and ctx.bars >= 100:
        ctx.buy_shares = 100
        ctx.hold_bars = 30

strategy = Strategy(YFinance(), '6/1/2021', '6/1/2022')
strategy.add_execution(buy_and_hold, ['T', 'F', 'GM', 'PFE'])

This will buy 100 shares in each of ['T', 'F', 'GM', 'PFE']. But what if you don’t want to use equal position sizing? For example, you may want to size positions so that more shares are allocated to tickers with lower volatility to decrease the portfolio’s overall volatility.

To customize position sizing for each ticker, we can define a pos_size_handler function that calculates the position size for each ticker:

import numpy as np

def pos_size_handler(ctx):
    # Fetch all buy signals.
    signals = tuple(ctx.signals("buy"))
    # Return if there are no buy signals (i.e. there are only sell signals).
    if not signals:
        return
    # Calculates the inverse volatility, where volatility is defined as the
    # standard deviation of close prices for the last 100 days.
    get_inverse_volatility = lambda signal: 1 / np.std(signal.bar_data.close[-100:])
    # Sums the inverse volatilities for all of the buy signals.
    total_inverse_volatility = sum(map(get_inverse_volatility, signals))
    for signal in signals:
        size = get_inverse_volatility(signal) / total_inverse_volatility
        # Calculate the number of shares given the latest close price.
        shares = ctx.calc_target_shares(size, signal.bar_data.close[-1], cash=95_000)
        ctx.set_shares(signal, shares)

strategy.set_pos_size_handler(pos_size_handler)

The handler runs on every bar that generates a buy or sell signal when buy_shares or sell_shares is set on the ExecContext:

result = strategy.backtest()

Backtesting: 2021-06-01 00:00:00 to 2022-06-01 00:00:00

Loaded cached bar data.

Test split: 2021-06-01 00:00:00 to 2022-05-31 00:00:00

100% (253 of 253) |######################| Elapsed Time: 0:00:00 Time:  0:00:00

Finished backtest: 0:00:00

result.trades

	type	symbol	entry_date	exit_date	entry	exit	shares	pnl	return_pct	agg_pnl	bars	pnl_per_bar	stop
id
1	long	T	2021-10-21	2021-12-03	19.60	17.54	2248	-4630.88	-10.51	-4630.88	30	-154.36	bar
2	long	F	2021-10-21	2021-12-03	16.41	19.66	2028	6591.00	19.80	1960.12	30	219.70	bar
3	long	GM	2021-10-21	2021-12-03	58.20	60.28	129	268.32	3.57	2228.44	30	8.94	bar
4	long	PFE	2021-10-21	2021-12-03	42.76	53.75	235	2582.65	25.70	4811.09	30	86.09	bar
5	long	T	2021-12-06	2022-01-19	17.81	20.49	2665	7142.20	15.05	11953.29	30	238.07	bar
6	long	F	2021-12-06	2022-01-19	19.05	23.66	1075	4955.75	24.20	16909.04	30	165.19	bar
7	long	GM	2021-12-06	2022-01-19	59.72	57.99	209	-361.57	-2.90	16547.47	30	-12.05	bar
8	long	PFE	2021-12-06	2022-01-19	52.57	53.97	290	406.00	2.66	16953.47	30	13.53	bar
9	long	T	2022-01-20	2022-03-04	20.53	17.87	2625	-6982.50	-12.96	9970.97	30	-232.75	bar
10	long	F	2022-01-20	2022-03-04	22.22	17.01	789	-4110.69	-23.45	5860.28	30	-137.02	bar
11	long	GM	2022-01-20	2022-03-04	55.88	43.08	255	-3264.00	-22.91	2596.28	30	-108.80	bar
12	long	PFE	2022-01-20	2022-03-04	53.80	48.09	192	-1096.32	-10.61	1499.96	30	-36.54	bar
13	long	T	2022-03-07	2022-04-19	17.88	19.51	3132	5105.16	9.12	6605.12	30	170.17	bar
14	long	F	2022-03-07	2022-04-19	16.43	15.98	1303	-586.35	-2.74	6018.77	30	-19.55	bar
15	long	GM	2022-03-07	2022-04-19	41.09	41.50	220	90.20	1.00	6108.97	30	3.01	bar
16	long	PFE	2022-03-07	2022-04-19	48.18	50.62	200	488.00	5.06	6596.97	30	16.27	bar

Using this method allows for a lot of possibilities, such as using Mean-Variance Optimization to determine portfolio allocations.

In the next notebook, we will discuss how to implement custom indicators in PyBroker.