RSE Stock Selling Analysis

Background: At my current job, I receive both Restricted Stock Units (RSUs) and participate in an Employee Stock Purchase Plan (ESPP). With these regular granting of stock, I often question when is the right time to sell my stock and what price should I sell them at.

For this analysis, I will attempt to create a protocol for selling my company stock in order to maximize total returns within a 2 week time frame (10 business days) sell date.

0) Import Libraries

I will use Yahoo Finance for collecting the daily stock data and Pandas-TA Library for calculations of my technical indicators. Other libraries will include pandas for data frame handling and matplot lib for any visualizations.

pip install pandas_ta

Requirement already satisfied: pandas_ta in /usr/local/lib/python3.12/dist-packages (0.4.71b0)
Requirement already satisfied: numba==0.61.2 in /usr/local/lib/python3.12/dist-packages (from pandas_ta) (0.61.2)
Requirement already satisfied: numpy>=2.2.6 in /usr/local/lib/python3.12/dist-packages (from pandas_ta) (2.2.6)
Requirement already satisfied: pandas>=2.3.2 in /usr/local/lib/python3.12/dist-packages (from pandas_ta) (2.3.3)
Requirement already satisfied: tqdm>=4.67.1 in /usr/local/lib/python3.12/dist-packages (from pandas_ta) (4.67.1)
Requirement already satisfied: llvmlite<0.45,>=0.44.0dev0 in /usr/local/lib/python3.12/dist-packages (from numba==0.61.2->pandas_ta) (0.44.0)
Requirement already satisfied: python-dateutil>=2.8.2 in /usr/local/lib/python3.12/dist-packages (from pandas>=2.3.2->pandas_ta) (2.9.0.post0)
Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.12/dist-packages (from pandas>=2.3.2->pandas_ta) (2025.2)
Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.12/dist-packages (from pandas>=2.3.2->pandas_ta) (2025.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.12/dist-packages (from python-dateutil>=2.8.2->pandas>=2.3.2->pandas_ta) (1.17.0)

%matplotlib inline

import yfinance as yf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import datetime

For my initial strategy, I will set a price at a certain percentage premium of the closing price on day 0. If the high in the 10 days is of trading (max of the high for the following 10 days) is above the premium, it is a succesful trade.

Along with just setting a premium for all trading days, I will also attempt to set different premiums based off of what momentum indicators are showing (positive, negative or neutral momentum).

For a list of momentum indicators, I used AI to get a list of the indicators and the time frames. They are:

1. Relative Strength Indicator (RSI)
2. Moving Average Convergence Divergence (MACD)
3. Stochastic Oscillator
4. Rate of Change
5. Average Directional Index (ADX)

from pandas_ta import rsi, macd, stoch, roc, adx

1) Data Preparation

#Import Yahoo Finance Data - date range corresponds to my time at Intel while withholding recent data to allow for testing if desired
startDate = datetime.datetime(2018, 8, 6)
endDate = datetime.datetime(2025, 6, 30)
company = "INTC"

data = yf.Ticker(company).history(start = startDate, end=endDate)
data = data[['Close', 'Low', 'High']]

#Quick Closing Stock Price Visualization

plt.figure(figsize=(14, 7))

plt.plot(data['Close'], label='Close Price', alpha=0.6)
plt.title('INTC Closing Stock Price')
plt.xlabel('Date')
plt.ylabel('Price (USD)')
plt.grid(True)
plt.show()

data['RSI'] = data['Close']
data[['MACD', 'MACDh', 'MACDs']] = macd(data['Close'], fast=12, slow=26, signal=9)
data[['Stoch_k', 'Stoch_d', 'Stoch_h']] = stoch(data['High'], data['Low'], data['Close'], k=14, d=3, smooth_k=3, append=True)
data['ROC'] = roc(data['Close'])
data[['ADX', 'ADX_s', 'DMP', 'DMN']] = adx(high = data['High'], low = data['Low'], close = data['Close'])
data.tail(20)

	Close	Low	High	RSI	MACD	MACDh	MACDs	Stoch_k	Stoch_d	Stoch_h	ROC	ADX	ADX_s	DMP	DMN
Date
2025-05-30 00:00:00-04:00	19.549999	19.309999	20.260000	19.549999	-0.204103	-0.152510	-0.051594	10.630918	14.267141	-3.636223	-9.280743	9.131414	8.821788	23.182823	33.814888
2025-06-02 00:00:00-04:00	19.740000	19.370001	19.820000	19.740000	-0.252133	-0.160431	-0.091701	9.770572	12.247860	-2.477288	-8.864266	9.811563	9.105254	22.204661	32.388125
2025-06-03 00:00:00-04:00	20.290001	19.400000	20.410000	20.290001	-0.243014	-0.121050	-0.121964	16.626224	12.342571	4.283653	-5.098215	9.609246	9.370330	25.555902	29.390697
2025-06-04 00:00:00-04:00	20.250000	20.010000	20.500000	20.250000	-0.236292	-0.091462	-0.144830	27.174448	17.857081	9.317367	-4.795489	9.299836	9.555699	25.224349	28.035119
2025-06-05 00:00:00-04:00	19.990000	19.850000	20.549999	19.990000	-0.249073	-0.083395	-0.165678	32.411208	25.403960	7.007248	-3.383281	9.212477	9.410862	23.553072	27.692046
2025-06-06 00:00:00-04:00	20.059999	20.030001	20.440001	20.059999	-0.250664	-0.067989	-0.182676	32.816684	30.800780	2.015904	-2.384427	9.131359	9.215597	22.520076	26.477522
2025-06-09 00:00:00-04:00	20.480000	20.219999	20.959999	20.480000	-0.215550	-0.026300	-0.189250	36.837914	34.021935	2.815978	2.144640	8.675930	8.944204	25.563174	24.192234
2025-06-10 00:00:00-04:00	22.080000	20.280001	22.440001	22.080000	-0.057948	0.105042	-0.162990	57.085674	42.246757	14.838917	7.445259	9.885414	9.508386	33.397883	19.779772
2025-06-11 00:00:00-04:00	20.680000	20.379999	21.830000	20.680000	-0.045490	0.094000	-0.139490	60.899804	51.607797	9.292007	1.521843	11.008507	9.842218	28.926152	17.131406
2025-06-12 00:00:00-04:00	20.770000	20.410000	20.980000	20.770000	-0.028032	0.089166	-0.117198	59.637917	59.207798	0.430119	2.567903	12.051378	10.968396	27.592162	16.341355
2025-06-13 00:00:00-04:00	20.139999	20.100000	20.600000	20.139999	-0.064291	0.042326	-0.106617	38.977641	53.171787	-14.194146	3.017904	12.500102	11.754304	26.070241	17.992072
2025-06-16 00:00:00-04:00	20.740000	20.299999	20.930000	20.740000	-0.044103	0.050011	-0.094114	39.616613	46.077390	-6.460777	5.065856	13.279985	12.665681	27.098008	16.814398
2025-06-17 00:00:00-04:00	20.799999	20.620001	21.480000	20.799999	-0.022998	0.056893	-0.079891	39.936089	39.510114	0.425974	2.513545	14.554034	13.527068	29.724594	15.616084
2025-06-18 00:00:00-04:00	21.490000	20.660000	21.600000	21.490000	0.048843	0.102987	-0.054144	54.313085	44.621929	9.691156	6.123456	15.850612	14.565298	28.412066	14.407458
2025-06-20 00:00:00-04:00	21.080000	20.879999	21.889999	21.080000	0.071865	0.100807	-0.028942	57.650887	50.633353	7.017533	5.452727	17.326880	15.940457	28.436783	13.223230
2025-06-23 00:00:00-04:00	21.190001	20.730000	21.580000	21.190001	0.097858	0.101441	-0.003582	61.410147	57.791373	3.618774	5.633106	18.399400	17.125006	26.465240	13.529934
2025-06-24 00:00:00-04:00	22.549999	21.330000	22.690001	22.549999	0.225598	0.183344	0.042254	69.884090	62.981708	6.902382	10.107421	20.342534	18.834707	31.999250	11.954776
2025-06-25 00:00:00-04:00	22.200001	22.129999	22.770000	22.200001	0.295188	0.202347	0.092841	78.143810	69.812682	8.331128	0.543482	22.205096	20.302248	31.009127	11.347743
2025-06-26 00:00:00-04:00	22.500000	22.209999	22.620001	22.500000	0.370278	0.221950	0.148328	88.565269	78.864390	9.700879	8.800772	23.934618	22.138576	29.934887	10.954627
2025-06-27 00:00:00-04:00	22.690001	22.420000	23.379999	22.690001	0.440046	0.233374	0.206672	83.196293	83.301791	-0.105498	9.244102	26.082944	24.144020	33.803214	10.093879

Finally, let's generate the output variable for our analysis, the max high in the following 10 days.

data['rolling_high'] = data['High'].rolling(window=10).max()
data['rolling_high'] = data['rolling_high'].shift(-10)
data['Rolling High %'] = data['rolling_high'] / data['Close']

2) Baseline Analysis

Initially, let's just see what the average rolling high percentage is.

_mean = data['Rolling High %'].mean()
_median = data['Rolling High %'].median()
_std = data['Rolling High %'].std()

print(f'Mean High Percentage: {_mean:0.4f}')
print(f'Median High Percentage: {_median:0.4f}')
print(f'Standard Deviation of Highs: {_std:0.4f}')

Mean High Percentage: 1.0573
Median High Percentage: 1.0430
Standard Deviation of Highs: 0.0582

data['Rolling High %'].hist(bins=40, figsize=(10, 6))
plt.xlabel('Rolling High Percentage')
plt.ylabel('Frequency')
plt.show()

Looks like putting a sell order with a 4% premium on the close price is a good start. However, looking at the high standard deviation and the number of entries below 1 (corresponding to losses where the preference is to sell immediately), there is room for improvement.

3) Momentum Indicator Analysis

3.1) RSI

AI description of RSI "The RSI measures the speed and change of price movements. It is used to determine if an asset is overbought (RSI > 70) or oversold (RSI < 30) and to identify potential price reversals."

For this indicator, I will label Upward momentum as points with RSI < 30, downward momentum as points with RSI > 70, and neutral as those in between.

I will then see how the segmentation impacts the median, mean, and standard deviation of the rolling high percentages.

data_rsi = data[['Close', 'Rolling High %', 'RSI']]

#Quick Visualization

plt.figure(figsize=(14, 7))

plt.plot(data['RSI'], label='RSI', alpha=0.6)
plt.axhline(70, lw=1, ls='--', c='k')
plt.axhline(30, lw=1, ls='--', c='k')
plt.title('INTC 14 day RSI')
plt.xlabel('Date')
plt.ylabel('RSI')
plt.grid(True)
plt.show()

Well, there seems to be no point of Downward momentum "Over bought." Let still see about separating between neutral and upward momentum.

def assign_trend(row):
    if row['RSI'] >= 70:
        return 'Down'
    elif row['RSI'] < 30:
        return 'Up'
    else:
        return 'Neutral'

data_rsi['Signal'] = data.apply(assign_trend, axis=1)

print(len(data_rsi))
print(len(data_rsi[data_rsi['Signal']=='Up']))
print(len(data_rsi[data_rsi['Signal']=='Neutral']))
print(len(data_rsi[data_rsi['Signal']=='Down']))

1733
400
1333
0

data_up = data_rsi[data_rsi['Signal']=='Up']
data_neutral = data_rsi[data_rsi['Signal']=='Neutral']

_mean = data_up['Rolling High %'].mean()
_median = data_up['Rolling High %'].median()
_std = data_up['Rolling High %'].std()

print(f'Mean High Percentage: {_mean:0.4f}')
print(f'Median High Percentage: {_median:0.4f}')
print(f'Standard Deviation of Highs: {_std:0.4f}')

Mean High Percentage: 1.0854
Median High Percentage: 1.0674
Standard Deviation of Highs: 0.0806

_mean = data_neutral['Rolling High %'].mean()
_median = data_neutral['Rolling High %'].median()
_std = data_neutral['Rolling High %'].std()

print(f'Mean High Percentage: {_mean:0.4f}')
print(f'Median High Percentage: {_median:0.4f}')
print(f'Standard Deviation of Highs: {_std:0.4f}')

Mean High Percentage: 1.0490
Median High Percentage: 1.0384
Standard Deviation of Highs: 0.0467

data_up['Rolling High %'].hist(bins=40, figsize=(10, 6))
plt.xlabel('Rolling High Percentage')
plt.ylabel('Frequency')
plt.title('Up Momentum')
plt.show()

data_neutral['Rolling High %'].hist(bins=40, figsize=(10, 6))
plt.xlabel('Rolling High Percentage')
plt.ylabel('Frequency')
plt.title('Neutral Momentum')
plt.show()

RSI did help to differentiate, with the upward data set having a return 3% higher than the neutral. However, the standard deviations are still pretty high. Let's continue with MACD.

3.2) MACD

AI description of MACD "The MACD compares two moving averages of a security's price to signal shifts in momentum and trend direction. It is used to identify bullish or bearish momentum shifts and potential trend continuations or reversals."

For this indicator, I will use the MACD histogram to distinguish between neutral, upward, and downward. Unfortunately, there is no clear guidance on what separates neutral and upward/downward (only that upward is above 0 and downward is below 0). I will look at a histogram of the values to see if there is visually a good separation point.

data_MACD = data[['Close', 'Rolling High %', 'MACD', 'MACDh', 'MACDs']]

#Quick Visualization

plt.figure(figsize=(14, 7))

plt.plot(data_MACD[['MACD', 'MACDh', 'MACDs']], alpha=0.6)
plt.legend(['MACD', 'MACD Signal', 'MACD Histogram'])
plt.axhline(0, lw=1, ls='--', c='k')
plt.title('INTC 26-14-9 MACD')
plt.xlabel('Date')
plt.ylabel('MACD')
plt.grid(True)
plt.show()

data_MACD['MACDh'].hist(bins=40, figsize=(10, 6))
plt.xlabel('MACD Histogram')
plt.ylabel('Frequency')
plt.title('MACD Histogram')
plt.show()

data_MACD['MACDh'].describe()

	MACDh
count	1700.000000
mean	0.002253
std	0.333680
min	-1.565794
25%	-0.161131
50%	0.021869
75%	0.209578
max	1.128676

dtype: float64

Let's use the standard deviation as the cut off point. Neutral momentum will be any point between -0.33 and 0.33.

def assign_trend(row):
    if row['MACDh'] <= -0.33:
        return 'Down'
    elif row['MACDh'] >= 0.33:
        return 'Up'
    else:
        return 'Neutral'

data_MACD['Signal'] = data.apply(assign_trend, axis=1)
data_MACD.head(100)

	Close	Rolling High %	MACD	MACDh	MACDs	Signal
Date
2018-08-06 00:00:00-04:00	42.163277	1.026369	NaN	NaN	NaN	Neutral
2018-08-07 00:00:00-04:00	42.505390	1.018108	NaN	NaN	NaN	Neutral
2018-08-08 00:00:00-04:00	42.727737	1.012810	NaN	NaN	NaN	Neutral
2018-08-09 00:00:00-04:00	42.881680	0.982050	NaN	NaN	NaN	Neutral
2018-08-10 00:00:00-04:00	41.778419	1.007984	NaN	NaN	NaN	Neutral
...	...	...	...	...	...	...
2018-12-20 00:00:00-05:00	39.192696	1.054018	-0.170389	-0.262606	0.092217	Neutral
2018-12-21 00:00:00-05:00	38.590263	1.071142	-0.313561	-0.324622	0.011061	Neutral
2018-12-24 00:00:00-05:00	37.514473	1.114247	-0.507978	-0.415231	-0.092747	Down
2018-12-26 00:00:00-05:00	39.752098	1.053258	-0.476010	-0.306610	-0.169399	Neutral
2018-12-27 00:00:00-05:00	39.898396	1.063201	-0.433868	-0.211575	-0.222293	Neutral

100 rows × 6 columns

print(len(data_MACD))
data_up = data_MACD[data_MACD['Signal']=='Up']
data_neutral = data_MACD[data_MACD['Signal']=='Neutral']
data_down = data_MACD[data_MACD['Signal']=='Down']
print(len(data_up))
print(len(data_neutral))
print(len(data_down))

1733
213
1294
226

_mean = data_up['Rolling High %'].mean()
_median = data_up['Rolling High %'].median()
_std = data_up['Rolling High %'].std()

print(f'Upward Mean High Percentage: {_mean:0.4f}')
print(f'Upward Median High Percentage: {_median:0.4f}')
print(f'Upward Standard Deviation of Highs: {_std:0.4f}')

Upward Mean High Percentage: 1.0505
Upward Median High Percentage: 1.0361
Upward Standard Deviation of Highs: 0.0507

_mean = data_neutral['Rolling High %'].mean()
_median = data_neutral['Rolling High %'].median()
_std = data_neutral['Rolling High %'].std()

print(f'Neutral Mean High Percentage: {_mean:0.4f}')
print(f'Neutral Median High Percentage: {_median:0.4f}')
print(f'Neutral Standard Deviation of Highs: {_std:0.4f}')

Neutral Mean High Percentage: 1.0587
Neutral Median High Percentage: 1.0443
Neutral Standard Deviation of Highs: 0.0579

_mean = data_down['Rolling High %'].mean()
_median = data_down['Rolling High %'].median()
_std = data_down['Rolling High %'].std()

print(f'Downward Mean High Percentage: {_mean:0.4f}')
print(f'Downward Median High Percentage: {_median:0.4f}')
print(f'Downward Standard Deviation of Highs: {_std:0.4f}')

Downward Mean High Percentage: 1.0557
Downward Median High Percentage: 1.0430
Downward Standard Deviation of Highs: 0.0656

MACD Histogram did not help separate the performance.

3.3) Stochastic Oscillator

AI description of Stoch "This indicator compares a security's closing price to its price range over a specific period. It is used to forecast potential trend reversals and identify overbought and oversold conditions."

For this, I will use 80 as the downward cutoff and 20 as the upward cutoff.

data_stoch = data[['Close', 'Rolling High %', 'Stoch_k', 'Stoch_d', 'Stoch_h']]

#Quick Visualization

plt.figure(figsize=(14, 7))

plt.plot(data_stoch[['Stoch_k', 'Stoch_d']], alpha=0.6)
plt.legend(['K', 'D'])
plt.axhline(80, lw=1, ls='--', c='k')
plt.axhline(20, lw=1, ls='--', c='k')
plt.title('INTC Stochastic Oscillator')
plt.xlabel('Date')
plt.ylabel('Stoch')
plt.grid(True)
plt.show()

def assign_trend(row):
    if row['Stoch_k'] >= 80:
        return 'Down'
    elif row['Stoch_k'] <= 20:
        return 'Up'
    else:
        return 'Neutral'

data_stoch['Signal'] = data.apply(assign_trend, axis=1)

data_up = data_stoch[data_stoch['Signal']=='Up']
data_neutral = data_stoch[data_stoch['Signal']=='Neutral']
data_down = data_stoch[data_stoch['Signal']=='Down']
print(len(data_up))
print(len(data_neutral))
print(len(data_down))

356
1041
336

_mean = data_up['Rolling High %'].mean()
_median = data_up['Rolling High %'].median()
_std = data_up['Rolling High %'].std()

print(f'Upward Mean High Percentage: {_mean:0.4f}')
print(f'Upward Median High Percentage: {_median:0.4f}')
print(f'Upward Standard Deviation of Highs: {_std:0.4f}')

Upward Mean High Percentage: 1.0623
Upward Median High Percentage: 1.0453
Upward Standard Deviation of Highs: 0.0727

_mean = data_neutral['Rolling High %'].mean()
_median = data_neutral['Rolling High %'].median()
_std = data_neutral['Rolling High %'].std()

print(f'Neutral Mean High Percentage: {_mean:0.4f}')
print(f'Neutral Median High Percentage: {_median:0.4f}')
print(f'Neutral Standard Deviation of Highs: {_std:0.4f}')

Neutral Mean High Percentage: 1.0592
Neutral Median High Percentage: 1.0454
Neutral Standard Deviation of Highs: 0.0548

_mean = data_down['Rolling High %'].mean()
_median = data_down['Rolling High %'].median()
_std = data_down['Rolling High %'].std()

print(f'Downward Mean High Percentage: {_mean:0.4f}')
print(f'Downward Median High Percentage: {_median:0.4f}')
print(f'Downward Standard Deviation of Highs: {_std:0.4f}')

Downward Mean High Percentage: 1.0459
Downward Median High Percentage: 1.0333
Downward Standard Deviation of Highs: 0.0491

While the downward signal help to identify the lower performers, the upward signal was not useful.

3.4) Rate of Change

The ROC measures the percentage change in price between the current price and a price from a certain number of periods ago. It can confirm trends, identify price surges or drops, and signal overbought/oversold conditions.

data_roc = data[['Close', 'Rolling High %', 'ROC']]

#Quick Visualization

plt.figure(figsize=(14, 7))

plt.plot(data_roc['ROC'], alpha=0.6)
plt.title('INTC ROC')
plt.xlabel('Date')
plt.ylabel('ROC')
plt.grid(True)
plt.show()

data['ROC'].describe()

	ROC
count	1723.000000
mean	-0.044809
std	8.065740
min	-39.817908
25%	-4.173308
50%	0.228599
75%	4.339471
max	41.331272

dtype: float64

I will use a cutoff of 4 to distinguish between upward, downward, and neutral positions

def assign_trend(row):
    if row['ROC'] <= -4:
        return 'Down'
    elif row['ROC'] >= 4:
        return 'Up'
    else:
        return 'Neutral'

data_roc['Signal'] = data.apply(assign_trend, axis=1)

data_up = data_roc[data_roc['Signal']=='Up']
data_neutral = data_roc[data_roc['Signal']=='Neutral']
data_down = data_roc[data_roc['Signal']=='Down']
print(len(data_up))
print(len(data_neutral))
print(len(data_down))

454
836
443

_mean = data_up['Rolling High %'].mean()
_median = data_up['Rolling High %'].median()
_std = data_up['Rolling High %'].std()

print(f'Upward Mean High Percentage: {_mean:0.4f}')
print(f'Upward Median High Percentage: {_median:0.4f}')
print(f'Upward Standard Deviation of Highs: {_std:0.4f}')

Upward Mean High Percentage: 1.0529
Upward Median High Percentage: 1.0361
Upward Standard Deviation of Highs: 0.0538

_mean = data_neutral['Rolling High %'].mean()
_median = data_neutral['Rolling High %'].median()
_std = data_neutral['Rolling High %'].std()

print(f'Neutral Mean High Percentage: {_mean:0.4f}')
print(f'Neutral Median High Percentage: {_median:0.4f}')
print(f'Neutral Standard Deviation of Highs: {_std:0.4f}')

Neutral Mean High Percentage: 1.0550
Neutral Median High Percentage: 1.0427
Neutral Standard Deviation of Highs: 0.0509

_mean = data_down['Rolling High %'].mean()
_median = data_down['Rolling High %'].median()
_std = data_down['Rolling High %'].std()

print(f'Downward Mean High Percentage: {_mean:0.4f}')
print(f'Downward Median High Percentage: {_median:0.4f}')
print(f'Downward Standard Deviation of Highs: {_std:0.4f}')

Downward Mean High Percentage: 1.0660
Downward Median High Percentage: 1.0485
Downward Standard Deviation of Highs: 0.0728

No good classification for the ROC

3.5) ADX

The ADX line rises when a trend is gaining strength and falls when momentum is weakening. To determine whether the momentum is bullish or bearish, the ADX is used in conjunction with the two directional movement indicator (DMI) lines: the Positive Directional Indicator (+DI) and the Negative Directional Indicator (-DI).

data_adx = data[['Close', 'Rolling High %', 'ADX', 'DMP', 'DMN']]

#Quick Visualization

plt.figure(figsize=(14, 7))

plt.plot(data_adx[['ADX']], alpha=0.6)
plt.title('INTC ADX')
plt.xlabel('Date')
plt.ylabel('ADX')
plt.grid(True)
plt.show()

def assign_trend(row):
    if row['ADX'] >= 25:
      if row['DMP'] >= row['DMN']:
        return 'Up'
      else:
        return 'Down'
    else:
        return 'Neutral'

data_adx['Signal'] = data.apply(assign_trend, axis=1)

data_up = data_roc[data_adx['Signal']=='Up']
data_neutral = data_adx[data_adx['Signal']=='Neutral']
data_down = data_adx[data_adx['Signal']=='Down']
print(len(data_up))
print(len(data_neutral))
print(len(data_down))

294
1153
286

_mean = data_up['Rolling High %'].mean()
_median = data_up['Rolling High %'].median()
_std = data_up['Rolling High %'].std()

print(f'Upward Mean High Percentage: {_mean:0.4f}')
print(f'Upward Median High Percentage: {_median:0.4f}')
print(f'Upward Standard Deviation of Highs: {_std:0.4f}')

Upward Mean High Percentage: 1.0490
Upward Median High Percentage: 1.0313
Upward Standard Deviation of Highs: 0.0513

_mean = data_neutral['Rolling High %'].mean()
_median = data_neutral['Rolling High %'].median()
_std = data_neutral['Rolling High %'].std()

print(f'Neutral Mean High Percentage: {_mean:0.4f}')
print(f'Neutral Median High Percentage: {_median:0.4f}')
print(f'Neutral Standard Deviation of Highs: {_std:0.4f}')

Neutral Mean High Percentage: 1.0606
Neutral Median High Percentage: 1.0458
Neutral Standard Deviation of Highs: 0.0600

_mean = data_down['Rolling High %'].mean()
_median = data_down['Rolling High %'].median()
_std = data_down['Rolling High %'].std()

print(f'Downward Mean High Percentage: {_mean:0.4f}')
print(f'Downward Median High Percentage: {_median:0.4f}')
print(f'Downward Standard Deviation of Highs: {_std:0.4f}')

Downward Mean High Percentage: 1.0524
Downward Median High Percentage: 1.0422
Downward Standard Deviation of Highs: 0.0566

ADX doesn't provide any separation of the returns

Of the five technical indicators, the best performing seem to be the RSI and the stochastic oscillator. Since each one performed well on one side (RSI for upward trends and stoch for downward trends), I will see if combining them will improve the performance.

3.6) RSI and Stochastic Oscillator

data_combo = data_rsi

data_combo['Signal_stoch'] = data_stoch['Signal']

table = pd.crosstab(data_combo['Signal'], data_combo['Signal_stoch'])
print("Crosstab (Count):")
print(table)

Crosstab (Count):
Signal_stoch  Down  Neutral   Up
Signal                          
Neutral        289      800  244
Up              47      241  112

table = pd.crosstab(data_combo['Signal'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='mean')
print("Crosstab (Average 10 day max high):")
print(table)

Crosstab (Average 10 day max high):
Signal_stoch      Down   Neutral        Up
Signal                                    
Neutral       1.045966  1.051684  1.044017
Up            1.045774  1.085120  1.101993

table = pd.crosstab(data_combo['Signal'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='median')
print("Crosstab (Median 10 day max high):")
print(table)

Crosstab (Median 10 day max high):
Signal_stoch      Down   Neutral        Up
Signal                                    
Neutral       1.032861  1.040138  1.039599
Up            1.035666  1.069826  1.080496

table = pd.crosstab(data_combo['Signal'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='std')
print("Crosstab (Standard Deviation 10 day max high):")
print(table)

Crosstab (Standard Deviation 10 day max high):
Signal_stoch      Down   Neutral        Up
Signal                                    
Neutral       0.049436  0.047560  0.039448
Up            0.047551  0.068584  0.105602

The combindation of the two indicatiors seems to have help. If the RSI is showing an upward trend and the Stochastic Oscillator is not showing downward (upward or neutral), there seems to be higher possible returns than for the remaining combinations. This has potential to test it with the unseen data in the future (I will wait until the end of the year to give it about 126 data points (half a year).

4) Machine Learning

Let's see if any further insight can be gained from machine learning algorithms. In the end, I would like to still have some interpretable results. Therefore, I will stick with decision tree, random forest, and XGBoost algorithms to test. Decision tree will have a final visualization we can look at while the random forest and XGBoost we will mostly use to see the relative order of feature importance.

4.1) Decision Tree

#Library Import

from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor
import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.tree import plot_tree

data_clean = data.dropna()
features = ['RSI', 'MACDh', 'Stoch_k', 'Stoch_d', 'Stoch_h', 'ROC', 'ADX', 'ADX_s', 'DMP', 'DMN']
objective = ['Rolling High %']

X = data_clean[features]
y = data_clean[objective]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

regressor = DecisionTreeRegressor(max_depth=3, random_state=42)
regressor.fit(X_train, y_train)

DecisionTreeRegressor(max_depth=3, random_state=42)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

y_pred = regressor.predict(X_test)

mse = mean_squared_error(y_test, y_pred)
print(f"Mean Squared Error: {mse}")
r2 = r2_score(y_test, y_pred)
print(f"R2: {r2}")

Mean Squared Error: 0.00270067498311426
R2: 0.2676761976268732

#Decision Tree Visualization

plt.figure(figsize=(12, 8)) # Adjust figure size for better readability
plot_tree(regressor, filled=True, feature_names=features, rounded=True)
plt.title("Decision Tree Regressor Visualization")
plt.show()

Well, for the top performing groups, the decision tree also separated them using the Stochastic and RSI (although they switched the cutoff between the two to roughly 30 and 20 respectively. On the low performing side, it looks like a MACD below 0 signals the downward trend section.

4.2) Random Forest

rf_regressor = RandomForestRegressor(n_estimators=100, random_state=42)
rf_regressor.fit(X_train, y_train)

# Make predictions on the test set
y_pred = rf_regressor.predict(X_test)

# Evaluate the model
mse = mean_squared_error(y_test, y_pred)
print(f"Mean Squared Error: {mse}")

/usr/local/lib/python3.12/dist-packages/sklearn/base.py:1389: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples,), for example using ravel().
  return fit_method(estimator, *args, **kwargs)

Mean Squared Error: 0.0018319961475055085

importances = rf_regressor.feature_importances_
feature_names = X.columns
sorted_indices = np.argsort(importances)[::-1] # Sort in descending order

print("Feature Importances:")
for i in sorted_indices:
    print(f"{feature_names[i]}: {importances[i]:.4f}")

Feature Importances:
RSI: 0.3146
DMP: 0.1097
DMN: 0.0993
ROC: 0.0846
MACDh: 0.0803
Stoch_h: 0.0719
Stoch_d: 0.0700
Stoch_k: 0.0590
ADX: 0.0562
ADX_s: 0.0545

4.3) XGBoost

reg = xgb.XGBRegressor(n_estimators=5000, early_stopping_rounds=50, learning_rate = 0.01)
reg.fit(X_train, y_train,
        eval_set=[(X_train, y_train), (X_test, y_test)],
        verbose=100)

[0]	validation_0-rmse:0.05766	validation_1-rmse:0.06076
[100]	validation_0-rmse:0.04303	validation_1-rmse:0.05035
[200]	validation_0-rmse:0.03610	validation_1-rmse:0.04768
[300]	validation_0-rmse:0.03189	validation_1-rmse:0.04640
[400]	validation_0-rmse:0.02812	validation_1-rmse:0.04513
[500]	validation_0-rmse:0.02600	validation_1-rmse:0.04472
[600]	validation_0-rmse:0.02427	validation_1-rmse:0.04435
[700]	validation_0-rmse:0.02240	validation_1-rmse:0.04412
[800]	validation_0-rmse:0.02072	validation_1-rmse:0.04405
[900]	validation_0-rmse:0.01871	validation_1-rmse:0.04378
[1000]	validation_0-rmse:0.01706	validation_1-rmse:0.04346
[1100]	validation_0-rmse:0.01590	validation_1-rmse:0.04341
[1200]	validation_0-rmse:0.01427	validation_1-rmse:0.04318
[1300]	validation_0-rmse:0.01321	validation_1-rmse:0.04309
[1372]	validation_0-rmse:0.01268	validation_1-rmse:0.04306

XGBRegressor(base_score=None, booster=None, callbacks=None,
             colsample_bylevel=None, colsample_bynode=None,
             colsample_bytree=None, device=None, early_stopping_rounds=50,
             enable_categorical=False, eval_metric=None, feature_types=None,
             feature_weights=None, gamma=None, grow_policy=None,
             importance_type=None, interaction_constraints=None,
             learning_rate=0.01, max_bin=None, max_cat_threshold=None,
             max_cat_to_onehot=None, max_delta_step=None, max_depth=None,
             max_leaves=None, min_child_weight=None, missing=nan,
             monotone_constraints=None, multi_strategy=None, n_estimators=5000,
             n_jobs=None, num_parallel_tree=None, ...)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

fi = pd.DataFrame(reg.feature_importances_,
             index = reg.feature_names_in_,
             columns=['importance']).sort_values('importance')

print(fi)

         importance
Stoch_h    0.062621
ADX        0.088518
Stoch_k    0.089050
MACDh      0.089672
ADX_s      0.090524
ROC        0.095911
Stoch_d    0.097947
DMN        0.104419
DMP        0.120347
RSI        0.160990

In both instances, RSI was the most important with DMP and DMN in the next three.

5) RSI/Stoch Redo

Based on the decision tree result above, I wanted to redo the RSI/Stochastic Oscillator result with the new cutoff or oversold/upward momentum.

def assign_trend_stoch(row):
    if row['Stoch_k'] >= 80:
        return 'Down'
    elif row['Stoch_k'] <= 30: #New cutoff (originally 20)
        return 'Up'
    else:
        return 'Neutral'

def assign_trend_rsi(row):
    if row['RSI'] >= 70:
        return 'Down'
    elif row['RSI'] < 20: #New cutoff (originally 30)
        return 'Up'
    else:
        return 'Neutral'

data_combo['Signal_stoch'] = data.apply(assign_trend_stoch, axis=1)
data_combo['Signal_rsi'] = data.apply(assign_trend_rsi, axis=1)

table = pd.crosstab(data_combo['Signal_rsi'], data_combo['Signal_stoch'])
print("Crosstab (Count):")
print(table)

Crosstab (Count):
Signal_stoch  Down  Neutral   Up
Signal_rsi                      
Neutral        336      863  482
Up               0       15   37

table = pd.crosstab(data_combo['Signal_rsi'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='mean')
print("Crosstab (Average 10 day max high):")
print(table)

Crosstab (Average 10 day max high):
Signal_stoch     Down   Neutral        Up
Signal_rsi                               
Neutral       1.04594  1.056494  1.053056
Up                NaN  1.124915  1.205262

table = pd.crosstab(data_combo['Signal_rsi'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='median')
print("Crosstab (Median 10 day max high):")
print(table)

Crosstab (Median 10 day max high):
Signal_stoch      Down   Neutral        Up
Signal_rsi                                
Neutral       1.033294  1.043275  1.043035
Up                 NaN  1.127831  1.163918

table = pd.crosstab(data_combo['Signal_rsi'], data_combo['Signal_stoch'], values=data_combo['Rolling High %'], aggfunc='std')
print("Crosstab (Standard Deviation 10 day max high):")
print(table)

Crosstab (Standard Deviation 10 day max high):
Signal_stoch      Down   Neutral        Up
Signal_rsi                                
Neutral       0.049116  0.052023  0.051720
Up                 NaN  0.029711  0.116514

Looking at the count crosstab, it looks like the improvement in separation is due to the higher cutoff for RSI upward momentum.

6) Conclusion

After performing this analysis, I will start using the RSI indicator in deciding when I should sell my stock. The ADX DMP and DMI indicators also look like they might be useful in determining the direction of movement, although the ADX indicator didn't provide much additional information according to the random forest and XGBoost results. I want to return to this after the end of 2025 so I can test some of the strategies proposed here and see how they would perform in real life conditions.