Guides

Technical indicators

On this page7

data.indicator() supports TA-Lib and pandas_ta with hundreds of technical indicators. One line calculates thousands of stocks over long time periods, for individual stock analysis or machine learning. The result is a date-by-ticker table, like data.get(), so you can use it directly in a stock-selection rule.

Install TA-Lib

!pip install ta-lib

For more parameters and functions, see the TA-Lib official documentation.

Compute RSI for every stock

Pass the indicator's name and parameters to data.indicator(), for example RSI (timeperiod=14):

from finlab import data

data.set_market('us')
rsi = data.indicator('RSI', timeperiod=14)
rsi[['AAPL', 'MSFT', 'NVDA']].tail()

It computes every stock in the active market. The result is a DataFrame indexed by date with one column per ticker.
The first 14 days are NaN because there is not enough data yet.
By default the indicator uses unadjusted prices; pass adjust_price=True for adjusted prices, or resample='W' for weekly bars.

Compute a two-line indicator: Stochastic (KD)

data.indicator('STOCH') computes the Stochastic Oscillator (KD values) and returns two DataFrames (K and D):

from finlab import data

data.set_market('us')
k, d = data.indicator('STOCH')
k

Pick stocks where K is above D

List stocks where K > D on the most recent day:

signal = (k > d).iloc[-1]
signal[signal].index  # tickers with K above D on the latest date

(k > d).iloc[-1] is a boolean Series indexed by ticker for the latest date.

Indicators beyond the classic set

data.indicator() looks up TA-Lib first, by upper-case name. Newer TA-Lib releases include indicators such as SuperTrend, which returns the trend line and its direction (1 up, -1 down):

from finlab import data

data.set_market('us')
supertrend, direction = data.indicator('SUPERTREND')

When TA-Lib has no function of that name, a lower-case name is computed with pandas_ta if it is installed (it may be slower).

Build your own indicators with Pandas

Build custom indicators directly with Pandas operations, no TA-Lib required.

Momentum, moving-average deviation and revenue growth

from finlab import data

data.set_market('us')
close = data.get('price:adj_close')
volume = data.get('price:volume')
revenue = data.get('us_income_statement:revenue')  # quarterly

# Momentum: N-day return
momentum_20 = close / close.shift(20) - 1

# Moving average deviation
ma60 = close.average(60)
bias = (close - ma60) / ma60

# Price-volume divergence: price at high but volume shrinking
price_high = close == close.rolling(60).max()
volume_low = volume < volume.average(20)
divergence = price_high & volume_low

# Quarterly revenue growth
rev_yoy = revenue / revenue.shift(4) - 1  # year-over-year
rev_qoq = revenue / revenue.shift(1) - 1  # quarter-over-quarter
rev_trend = (
    revenue.rolling(2).mean() / revenue.rolling(4).mean() - 1
)  # last 2 vs last 4 quarters

Combine factors into one score (.cs)

.cs computes across all stocks on each date. Standardize each factor, then add them with equal weight:

eps = data.get('us_income_statement:eps_diluted')
net_income = data.get('us_income_statement:net_income')
equity = data.get('us_balance_sheet:total_stockholders_equity')

raw_close = data.get('price:close')  # unadjusted price for valuation
pe = raw_close / eps.rolling(4).sum()  # trailing-12-month P/E
roe = net_income / equity[equity > 0]  # quarterly ROE, positive equity only

# Multi-factor composite: standardize then equal-weight sum
score = roe.cs.zscore() + (-pe).cs.zscore()  # higher ROE, lower PE = better

# Select top 10% by composite score
position = score.cs.rank() > 0.9

Neutralize by sector (.sector)

Selecting on raw ROE overweights sectors where ROE is high anyway. Standardizing within each sector picks the standouts of each sector instead:

# Problem: raw ROE selection overweights high-ROE sectors (e.g., technology)
# Solution: within-sector standardization selects "sector standouts"
profile = data.get('us_company_profile')
sectors = dict(zip(profile['symbol'], profile['sector']))

roe_neutral = roe.sector(sectors).zscore()  # sector-neutralized
rev_neutral = rev_yoy.sector(sectors).zscore()  # revenue growth also neutralized

# Composite factor
score = roe_neutral + rev_neutral
position = score.cs.rank() > 0.8

# Or use within-sector ranking directly
position = roe.sector(sectors).rank() > 0.7  # top 30% ROE within each sector

Without a mapping, .sector uses the Taiwan industry classification; for U.S. stocks always pass one.

Set weights and backtest (.weight)

Once you have a selection, .weight sets each holding's weight; then hand the weights to sim():

from finlab.backtest import sim

position = score.cs.rank() > 0.8

# Refined weights: inverse volatility + turnover limit
weights = (
    position.weight.inverse_volatility(window=60, price=close)
    .rebalance('QE')  # keep rebalance dates only, so turnover is per rebalance
    .weight.limit_turnover(max_turnover=0.3)
)

report = sim(weights, resample='Q')

inverse_volatility() reads Taiwan prices unless you pass price. weight.cap_industry() caps Taiwan industry weights and is Taiwan market only.

See the finlab.dataframe reference for all .cs, .sector, and .weight methods.

Next steps

Technical indicator selection A complete U.S. strategy built on technical indicators.
Write better rules Take the top names by rank, combine conditions, set entries and exits.
Backtesting Rebalance frequency, stops, fees and the trade log.
finlab.dataframe .cs, .sector, .weight and other FinlabDataFrame methods.

Was this page helpful?