In [1]:

import os
import requests
import polars as pl
from rich import print

Let's start with something we have have experience with: the earnings endpoints.

We will collect tomorrow's earnings (Tue Nov 26th) for both afterhours and premarket and thin out the responses a bit for a more concise summary:

In [2]:

uw_token = os.environ['UW_TOKEN']
headers = {'Accept': 'application/json, text/plain', 'Authorization': uw_token}

# Collect Monday's earnings data (both afterhours and premarket)
earnings_ah_params = {
    'date': '2024-11-26'
}
earnings_ah_url = 'https://api.unusualwhales.com/api/earnings/afterhours'
earnings_ah_r = requests.get(earnings_ah_url, headers=headers, params=earnings_ah_params)
earnings_pm_params = {
    'date': '2024-11-26'
}
earnings_pm_url = 'https://api.unusualwhales.com/api/earnings/premarket'
earnings_pm_r = requests.get(earnings_pm_url, headers=headers, params=earnings_pm_params)

# Create dataframes with each response then concatenate them into a single dataframe
raw_ah_df = pl.DataFrame(earnings_ah_r.json()['data'])
raw_pm_df = pl.DataFrame(earnings_pm_r.json()['data'])
raw_earnings_df = pl.concat([raw_ah_df, raw_pm_df])
raw_earnings_df

preferred_order = [
    'symbol',
    'full_name',
    'sector',
    'is_s_p_500',
    'ending_fiscal_quarter_dt',
    'report_date_dt',
    'report_time',
    'street_mean_est',
    'pre_earnings_close',
    'expected_move_perc',
    'expected_move',
    'expected_high',
    'expected_low'
]
earnings_df = (
    raw_earnings_df
    .filter(pl.col('has_options') == True)
    .with_columns(
        pl.col('ending_fiscal_quarter').str.strptime(pl.Date, '%Y-%m-%d').alias('ending_fiscal_quarter_dt'),
        pl.col('pre_earnings_date').str.strptime(pl.Date, '%Y-%m-%d').alias('pre_earnings_date_dt'),
        pl.col('report_date').str.strptime(pl.Date, '%Y-%m-%d').alias('report_date_dt'),
        pl.col('expected_move').cast(pl.Float64),
        pl.col('expected_move_perc').cast(pl.Float64),
        pl.col('post_earnings_close').cast(pl.Float64),
        pl.col('pre_earnings_close').cast(pl.Float64),
        pl.col('reaction').cast(pl.Float64),
        pl.col('street_mean_est').cast(pl.Float64)
    )
    .with_columns(
        (pl.col('pre_earnings_close') + pl.col('expected_move')).alias('expected_high'),
        (pl.col('pre_earnings_close') - pl.col('expected_move')).alias('expected_low')
    )
    .select(preferred_order)
    .sort('expected_move_perc', descending=True)
)
earnings_df

Out[2]:

shape: (32, 13)

symbol	full_name	sector	is_s_p_500	ending_fiscal_quarter_dt	report_date_dt	report_time	street_mean_est	pre_earnings_close	expected_move_perc	expected_move	expected_high	expected_low
str	str	str	bool	date	date	str	f64	f64	f64	f64	f64	f64
"ZH"	"ZHIHU"	"Communication …	false	2024-09-30	2024-11-26	"premarket"	null	3.68	0.44788	1.65	5.33	2.03
"DDD"	"3D SYSTEMS"	"Technology"	false	2024-09-30	2024-11-26	"postmarket"	-0.09	3.43	0.167289	0.57	4.0	2.86
"EMBC"	"EMBECTA"	"Healthcare"	false	2024-09-30	2024-11-26	"premarket"	0.4	14.41	0.157259	2.27	16.68	12.14
"MDWD"	"MEDIWOUND"	"Healthcare"	false	2024-09-30	2024-11-26	"premarket"	-0.48	17.33	0.152562	2.64	19.97	14.69
"TITN"	"TITAN MACHINER…	"Industrials"	false	2024-10-31	2024-11-26	"premarket"	0.07	15.42	0.143327	2.21	17.63	13.21
…	…	…	…	…	…	…	…	…	…	…	…	…
"ADSK"	"AUTODESK"	"Technology"	true	2024-10-31	2024-11-26	"postmarket"	2.11	319.39	0.05546	17.71	337.1	301.68
"SJM"	"J M SMUCKER"	"Consumer Defen…	true	2024-10-31	2024-11-26	"premarket"	2.51	113.62	0.054674	6.21	119.83	107.41
"M"	"MACY'S"	"Consumer Cycli…	false	2024-10-31	2024-11-26	"premarket"	-0.01	15.94	0.054649	0.87	16.81	15.07
"MANU"	"MANCHESTER UNI…	"Consumer Cycli…	false	2024-09-30	2024-11-26	"premarket"	-0.29	17.21	0.047792	0.82	18.03	16.39
"ADI"	"ANALOG DEVICES…	"Technology"	true	2024-10-31	2024-11-26	"premarket"	1.64	223.58	0.043165	9.65	233.23	213.93

OK, 32 records, let's take a closer look at the tickers and full names to see if anything looks familiar or interesting.

In [3]:

earnings_tickers = earnings_df['symbol'].to_list()
full_names = earnings_df['full_name'].to_list()
tickers_and_names = {ticker: name for ticker, name in zip(earnings_tickers, full_names)}
print(tickers_and_names)

{
    'ZH': 'ZHIHU',
    'DDD': '3D SYSTEMS',
    'EMBC': 'EMBECTA',
    'MDWD': 'MEDIWOUND',
    'TITN': 'TITAN MACHINERY',
    'NOAH': 'NOAH HOLDINGS',
    'PD': 'PAGERDUTY',
    'ARWR': 'ARROWHEAD PHARMACEUTICALS',
    'IREN': 'IRIS ENERGY',
    'GES': 'GUESS',
    'ANF': 'ABERCROMBIE & FITCH',
    'KSS': 'KOHLS',
    'AMBA': 'AMBARELLA',
    'NTNX': 'NUTANIX',
    'DKS': "DICK'S SPORTING GOODS",
    'AMWD': 'AMERICAN WOODMARK',
    'URBN': 'URBAN OUTFITTERS',
    'HTHT': 'H WORLD GROUP',
    'DELL': 'DELL TECHNOLOGIES',
    'WDAY': 'WORKDAY',
    'JWN': 'NORDSTROM',
    'TEN': 'Tsakos Energy',
    'BURL': 'BURLINGTON STORES',
    'CRWD': 'CROWDSTRIKE HOLDINGS',
    'BBY': 'BEST BUY CO',
    'HPQ': 'HP',
    'YY': 'JOYY',
    'ADSK': 'AUTODESK',
    'SJM': 'J M SMUCKER',
    'M': "MACY'S",
    'MANU': 'MANCHESTER UNITED',
    'ADI': 'ANALOG DEVICES'
}

DELL has seen quite a journey since early August, from $88 to $144 in under 4 months.

Let's experiment with a new endpoint, from the Institutional collection:

https://api.unusualwhales.com/docs#/operations/PublicApi.InstitutionController.ownership

We pass in a ticker and get back institutions that have a footprint there, with some variations all applying to the latest quarter (ending Sep 30th):

Closing a previously-existing position
Opening a new position
Adding to an existing position
Holding an existing position steady

Let's bust a move:

In [4]:

# Secret tip from the guy who manually identified key institutional players:
# this list of tags will limit the enormous universe of institutions down
# to a punchy group of interesting firms.
tags = [
    '13d_activist', 'activist', 'value_investor', 'small_cap', 'biotech',
    'credit', 'technology', 'tiger_club', 'energy', 'esg', 'event'
]

dfs = []
tickers_with_no_io = []
for idx, ticker in enumerate(earnings_tickers):
    io_url = f'https://api.unusualwhales.com/api/institution/{ticker}/ownership'
    io_params = {
        'tags[]': tags,
    }
    io_r = requests.get(io_url, headers=headers, params=io_params)
    if len(io_r.json()['data']) > 0:
        raw_df = pl.DataFrame(io_r.json()['data'])
        df = (
            raw_df
            .with_columns(
                pl.lit(ticker).alias('ticker'),
                pl.col('shares_outstanding').cast(pl.String).fill_null('')
            )
        )
        dfs.append(df)
        schema = str(df.schema)
    else:
        tickers_with_no_io.append(ticker)

io_df = pl.concat(dfs)

OK, after a little fancy footwork to account for some international stocks where UW does not have outstanding share count (yet), we have combined all the responses into a single polars dataframe named io_dfs.

Before we move on forward, let's do a quick check on to see if any of the tickers did not have institutional ownership:

In [5]:

print(tickers_with_no_io)

['ZH']

Nothing of concern there with the single ticker showing no institutional ownership, ZH.

Let's pick DELL for the rest of this example: we want to see which institutions closed their DELL positions, which institutions opened new DELL positions, and which ones added to existing positions... and maybe at the end we can look at "static" positions too.

In [6]:

closed_dell_df = (
    io_df
    .filter(
        (pl.col('ticker') == 'DELL') &
        (pl.col('units') == 0)
    )
    .sort('units_changed', descending=False)
)
closed_dell_df

Out[6]:

shape: (5, 15)

avg_price	first_buy	historical_units	inst_share_value	inst_value	name	people	report_date	shares_outstanding	short_name	tags	units	units_changed	value	ticker
str	str	list[i64]	str	str	str	list[str]	str	str	str	list[str]	i64	i64	i64	str
null	"2022-09-30"	[0, 2388359, … 4508189]	"27759423751"	"39343262001"	"POINT72 ASSET …	["Steven Cohen"]	"2024-09-30"	"307762769"	"Point72 Asset …	["activist"]	0	-2388359	0	"DELL"
null	"2024-06-30"	[0, 126279, … 0]	"404051553"	"422582142"	"LIGHT STREET C…	["Glen Kacher"]	"2024-09-30"	"307762769"	"Light Street C…	["activist", "tiger_club", "technology"]	0	-126279	0	"DELL"
null	"2024-06-30"	[0, 34720, … 0]	"14055650749"	"17658794959"	"BRIDGEWATER AS…	["Ray Dalio"]	"2024-09-30"	"307762769"	"Bridgewater As…	["13d_activist"]	0	-34720	0	"DELL"
null	"2024-06-30"	[0, 19703, … 0]	"3810553458"	"4617003629"	"MAGNETAR FINAN…	["Alec Litowitz", "Ross Laser"]	"2024-09-30"	"307762769"	"Magnetar Finan…	["13d_activist", "energy", … "credit"]	0	-19703	0	"DELL"
null	"2024-06-30"	[0, 3780, … 0]	"447103807"	"486683812"	"CANNELL CAPITA…	["Carlo Cannell"]	"2024-09-30"	"307762769"	"Cannell Capita…	["activist", "small_cap"]	0	-3780	0	"DELL"

Who closed their existing DELL positions as of EOQ Sep 30th?

Point72 (Steven Cohen)
Light Street ("Tiger Cub" Glen Kacher)
Bridgewater (Ray Dalio, has since retired)
Magnetar (Alec Litowitz and Ross Laser)
Cannell (Carlo Cannell)

We can see that Point72 has closed a large (notionally-speaking anyway) holding: the firm sold approximately 2.39 million shares this past quarter on a position they opened in the quarter ending 2022-09-30... impressive trade and worth noting that they closed ALL their holdings.

Let's look at the other side of the coin, firms that opened new positions this past quarter:

In [7]:

new_dell_df = (
    io_df
    .filter(
        (pl.col('ticker') == 'DELL') &
        (pl.col('units') > 0) &
        (pl.col('first_buy') == '2024-09-30')
    )
)
new_dell_df

Out[7]:

shape: (1, 15)

avg_price	first_buy	historical_units	inst_share_value	inst_value	name	people	report_date	shares_outstanding	short_name	tags	units	units_changed	value	ticker
str	str	list[i64]	str	str	str	list[str]	str	str	str	list[str]	i64	i64	i64	str
"115.53"	"2024-09-30"	[176200, 0, … 0]	"56601576463.0"	"56635136796.0"	"ADAGE CAPITAL …	["Robert Atchinson", "Phillip Gross"]	"2024-09-30"	"307762769"	"Adage Capital …	["13d_activist"]	176200	176200	20886748	"DELL"

Who opened new longs in DELL as of EOQ Sep 30th?

Adage (Robert Atchinson and Philip Gross)

Interesting, we can see Adage bought approximately 176,000 shares of DELL this past quarter. Adage is incredibly secretive, but their track record is known industry-wide: when the manager of Harvard University's takes a new position, it's noteworthy.

Let's see if any institutions added to existing positions in DELL this past quarter:

(after all, there was a significant pullback into the early August correlation one sell-off)

In [8]:

add_dell_df = (
    io_df
    .filter(
        (pl.col('ticker') == 'DELL') &
        (pl.col('units_changed') > 0) &
        (pl.col('first_buy') != '2024-09-30')
    )
)
add_dell_df

Out[8]:

shape: (6, 15)

avg_price	first_buy	historical_units	inst_share_value	inst_value	name	people	report_date	shares_outstanding	short_name	tags	units	units_changed	value	ticker
str	str	list[i64]	str	str	str	list[str]	str	str	str	list[str]	i64	i64	i64	str
"82.67"	"2023-06-30"	[2146777, 1447573, … 0]	"17794627077"	"58348942078"	"ALKEON CAPITAL…	["Panayotis “Takis” Sparaggis"]	"2024-09-30"	"307762769"	""	["technology"]	2146777	699204	254478946	"DELL"
"93"	"2022-09-30"	[646196, 342054, … 146624]	"66460868115.0"	"116492369102.0…	"D. E. SHAW & C…	["David Shaw"]	"2024-09-30"	"307762769"	"D.E. Shaw"	["13d_activist"]	646196	304142	76600074	"DELL"
"107.39"	"2024-03-31"	[312870, 256325, … 0]	"126914340970"	"127163204011"	"CLEARBRIDGE IN…	["Terrence Murphy"]	"2024-09-30"	"307762769"	"Clearbridge In…	["activist"]	312870	56545	37087610	"DELL"
"123.29"	"2024-03-31"	[33966, 15712, … 0]	"5128678280"	"5128708760"	"MAVERICK CAPIT…	["Lee Ainslie"]	"2024-09-30"	"307762769"	"Maverick Capit…	["activist", "tiger_club"]	33966	18254	4026330	"DELL"
null	"2021-06-30"	[23391, 21696, … 846571]	"36463931599"	"61306204232"	"BALYASNY ASSET…	["Dmitry Balyasny", "Scott Schroeder", "Taylor O'Malley"]	"2024-09-30"	"307762769"	"Balyasny Asset…	["activist"]	23391	1695	2772769	"DELL"
"106.1"	"2023-12-31"	[1128, 266, … 0]	"79316910"	"252652870"	"SACHETTA, LLC"	[]	"2024-09-30"	"307762769"	"Sachetta"	["activist"]	1128	862	133714	"DELL"

Who added to existing DELL positions as of EOQ Sep 30th?

Alkeon (Panoyotis 'Takis' Sparaggis)
D.E. Shaw (David Shaw)
Clearbridge (Terrence Murphy)
Maverick ("Tiger Cub" Lee Ainslie)
Balyasny (Dmitry Balyasny)

I'm quite interested in Alkeon's add here as a Technology-specialist. They have built up their size to nearly Point72 in nominal terms, and it's an interesting contrast to see them increasing their position by roughly +50%.

Let's round this out by identifying any institutions who continue to hold DELL with no change to the overall position:

In [9]:

flat_dell_df = (
    io_df
    .filter(
        (pl.col('ticker') == 'DELL') &
        (pl.col('units_changed') == 0) &
        (pl.col('first_buy') != '2024-09-30')
    )
)
flat_dell_df

Out[9]:

shape: (0, 15)

avg_price	first_buy	historical_units	inst_share_value	inst_value	name	people	report_date	shares_outstanding	short_name	tags	units	units_changed	value	ticker
str	str	list[i64]	str	str	str	list[str]	str	str	str	list[str]	i64	i64	i64	str

Empty!

Turns out DELL was a good pick for this review, since it's clearly a polarizing ticker: there was not a single fund in our set of activists and specialists with a static holding in DELL.