Simulación con el método de Montecarlo para verificar ganancias con márgenes negativos.

Overview

Apliación del método Monte Carlo a un ejemplo que incluye márgenes negativos.

Por Marco A. de la Cruz

Importante

La información contenida en este ejemplo, así como el planteamiento son ficticios. Este trabajo tiene como finalidad ejemplificar el uso del método Monte Carlo y cómo este puede ser replicado para algùn uso concreto.

Lo contenido en este trabajo es de acceso libre, sin embargo, pido de la ayuda del lector para dar crédito al autor original si este es usado para otros trabajos o estudios.

Planteamiento del problema

Una empresa dedicada a la venta de joyería ha optado por una estrategia agresiva, para atraer mayor mercado algunos de los precios de ciertos modelos estarán por debajo de los costos de adquisición. El equipo de mercadeo estima un aumento de sus clientes y creen que las ventas tendrán un comportamiento de una distribución normal con un promedio de 1000 y una desviación estándar de 100 ($N(\mu = 1000, \sigma = 100)$).

Por otro lado se debe considerar que esta empresa tiene una alianza comercial con una tarjeta de descuentos, el equipo de mercadeo ha indicado que consideran que hay una probabilidad del 20% de un cliente se presente con una tarjeta de descuentos plata, la cual permitiría un descuento del 5% en cualquier producto y que hay una probabilidad del 10% de que un cliente se presente con una tarjeta de descuentos oro, que permitirá adquirir un producto con 10% de descuento.

Tarjeta de descuentos % que un cliente use la tarjeta Descuento en el precio final
TDD plata 20% (-)5%
TDD oro 10% (-)10%

Por último se compartió el porcentaje de ventas de cada modelo, y se estima que esa razón se mantendrá en los meses siguientes, por lo que se puede considerar como la probabilidad de que cada modelo sea vendida.

Toda esta información está contenida en el archivo Prices.xlsx, la primera columna el artículo, la segunda columna el precio de venta, la tercera columna el costo unitario, y la última la probabilidad de que ese modelo sea vendido.

Artículo Precio Costo Probabilidad
Item #1 $ 4 600.00 $ 4 830.00 2.79%
Item #2 $ 4 100.00 $ 4 080.00 4.70%
Item #3 $ 2 500.00 $ 1 575.00 8.26%
Item #4 $ 4 400.00 $ 4 004.00 5.78%
Item #5 $ 5 000.00 $ 3 600.00 9.71%
Item #6 $ 5 000.00 $ 5 350.00 6.60%
Item #7 $ 2 700.00 $ 1 998.00 7.13%
Item #8 $ 2 200.00 $ 2 068.00 0.39%
Item #9 $ 2 600.00 $ 2 288.00 5.94%
Item #10 $ 2 800.00 $ 3 080.00 2.48%
Item #11 $ 2 500.00 $ 2 800.00 2.32%
Item #12 $ 2 300.00 $ 2 553.00 6.98%
Item #13 $ 4 000.00 $ 3 800.00 8.48%
Item #14 $ 4 800.00 $ 2 928.00 0.69%
Item #15 $ 3 900.00 $ 3 550.00 1.98%
Item #16 $ 2 700.00 $ 2 322.00 3.18%
Item #17 $ 4 400.00 $ 3 784.00 6.36%
Item #18 $ 3 300.00 $ 3 105.00 8.46%
Item #19 $ 4 700.00 $ 3 478.00 3.69%
Item #20 $ 4 800.00 $ 3 840.00 4.08%

Para este ejemplo no se considerán otros costos (por ejemplo, costos fijos) más allá que los costos unitarios, además no se considerarán los impuestos. Esto con la finalidad de simplificar este ejercicio y dar a entender el funcionamiento del método Monte Carlo.

Para dar un poco más de profesionalidad al trabajo he optado por mostrar toda esta información en un archivo Latex, ya que como ejemplo esto puede ser empleado en un fomrsto estándar (como el que se usa para tesis y trabajos similares).

Explicación de cómo funciona el método Monte Carlo

Este es un método no determinista y se utiliza para que a través de valores probabilísticos se simule el comportamiento impredecible de la realidad con la generación de números semialeatorios.

Si bien es cierto que un mayor número de simulaciones deberían ser más parecidos a los cálculos obtenidos por métodos deterministas, es precisamente con este comportamiento con el que se pueden obtener datos extra como la variabilidad y obtener rangos de comportamiento.

Para este ejercicio específico, se utilizarán tres simulaciones en paralelo en doscientas pruebas diferentes. El primero indicará cuantas piezas se recibirán, como se mencionó anteriormente se utilizará una función beta, este primer número indicará cuantas veces se ejecutará la segunda simulación. La segunda simulación determinará qué dispositivo se ha recibido considerando un número semialeatorio continuo entre 0 y 1, y respecto a la función acumulada de los diferentes modelos indicará a qué modelo y gradación corresponde.

La tercera simulación determina a quién se venderá el equipo, la opción que mejor pague por el dispositivo tendrá una probabilidad del 60%, la segunda mejor opción tendrá una probabilidad del 30% y la tercera opción un 10%.

Se hará una operación aritmética del precio de venta de cada equipo menos el precio de adquisición y se registrará el total en una tabla.

Finalmente se tomarán todos los resultados de la tabla y podremos determinar los números estadísticos descriptivos y conocer el alcance del programa bajo dichas cantidades.

Para poder replicar el análisis en cualquier computadora con el mismo software, se utilizarán semillas en los métodos y funciones que requieran números aleatorios.


Due to the behavior of received Devices from the trade in program, it is believed that we will receive five hundred devices per month, however, in order for the simulation to consider scenarios where sales are slightly lower than planned, it is proposed two hundred simulations where a random number of pieces of the beta function $X \sim Beta(\alpha= 4, \beta = 5)$ with location 400 and scale of 200 (image 1). This distribution has the characteristic of having a mean ($\overline{X}$) of 488.03, a median ($\widetilde{X}$) of 488.89 and a standard variation ($\sigma$) of 31.43.

Image 1. Beta Function for received devices.Image 1

Analytics has determined that the equipment reception will have a distribution as indicated in table 1. For simulation issues, each received equipment will generate a semi-random number between 0 and 1 to simulate the grade of the device received, only in the event that a unit does not have a price available for grade 3, the price of grade 2 will be taken.

Table 1. Probability of receiving different grade devices.

Grading Probability of being received
Grade 1 80.00%
Grade 2 15.24%
Grade 3 4.76%

Due to the above, a file containing the probabilities of each device of being acquired by the trade-in program has been restructured, separated into three main columns: Model, grading and probability. The probability has been calculated as follows. $$ p_{model - grade}=p_{model}\cdot p_{grading} $$ An example of the above is the Galaxy Note 10 model, which in the model mix has a probability of receiving it of 0.6%, this amount would be multiplied by each of the probabilities of each grade of being received (table 1). $$ p_{Note10-Grade1}=0.6% \cdot 80.00% = 0.4800%\ p_{Note10-Grade2}=0.6% \cdot 15.24% = 0.0914%\ p_{Note10-Grade3}=0.6% \cdot 4.76% = 0.0286% $$ With the above we can affirm the following, where:

  • $p_{i \alpha}$ is the probability of each model of grade 1 to being traded.

  • $p_{i \beta}$ is the probability of each model of grade 2 to being traded.

  • $p_{i\gamma}$ Is the probability of each model of grade 3 to being traded.

  • $n$ is the quantity of models on the list.

$$ \sum_{i=1}^{n} ( p_{i \alpha} + p_{i \beta} + p_{i \gamma} ) = 1.00 $$

Since three files have been manipulated, these will be in the annexes, only details of what is contained in each column will be given.

Probabilities.xlsx (table 2) indicates the probability that a device will be acquired in the trade-in program by model and grading.

Table 2. Description in Probabilities.xlsx.

Column name Description
Device name Model name.
Model Mix Probability of receiving this model in the trade-in program.
Grade Model grade.
Probability Probability of receiving this model with the corresponding grading in the trade-in program.

Buying price.xlsx (table 3) indicates the purchase prices by model and grading.

Table 3. Description in Buying price.xlsx.

Column name Description
Device name Model name.
Grading Model grade.
Buying price Acquisition price of the model and corresponding grading.

Aggregators prices.xlsx (table 4) indicates the prices at which the equipment will be sold, in this specific case it will be considered that 60% of the collected equipment will be sold to the best option, 30% to the second highest bidder and 10 % to the highest bidder.

Table 4. Aggregators prices.xlsx.

Column name Description
Device name Model name.
Grade Model grade.
Selling price 1 Maximum sale price to aggregators for the model and grading indicated.
Selling price 2 Second highest selling price to aggregators for the model and grading indicated.
Selling price 3 Third highest selling price to aggregators for the designated model and grading.

Explanation how Monte Carlo method works

This is a non-deterministic method and it is used so that through probabilistic values the unpredictable behavior of reality is simulated with the generation of semi-random numbers.

Although it is true that a greater number of simulations should be more similar to the calculations obtained by deterministic methods, it is precisely with this behavior that extra data such as variability can be obtained and behavior ranges obtained.

For this specific exercise, three simulations will be used in parallel in two hundred different tests. The first will indicate how many pieces will be received, as previously mentioned a beta function will be used, this first number will indicate how many times the second simulation will be executed. The second simulation will determine which device has been received considering a continuous semi-random number between 0 and 1, and with respect to the accumulated function of the different models it will indicate which model and grading it corresponds to.

The third simulation determines to whom the equipment will be sold, the option that pays the best for the device will have a probability of 60%, the second best option will have a probability of 30% and the third option a 10%.

An arithmetic operation will be made of the sale price of each equipment minus the acquisition price and the total will be recorded in a table.

Finally, all the results of the table will be taken and we will be able to determine the descriptive statistics numbers and know the scope of the program under said quantities.

In order to be able to replicate the analysis on any computer with the same software, seeds will be used in the methods and functions that require random numbers.

Analysis using Python 3.X

Step 1

Annexes

Annex 1 Creation of image 1 using python 3.X.

import numpy as np
import scipy.stats as st
import matplotlib.pyplot as plt
from matplotlib import ticker

x1 = np.arange(400, 601)
y1 = st.beta(4, 5, loc=400, scale=200).pdf(x1)

fig, ax = plt.subplots(figsize=(6,4))
plt.style.use('bmh')
ax.bar(x, y*100, color='grey', width=3)
ax.set_xlabel('Devices that can be received')
ax.set_ylabel('Probability of receiving those devices')
ax.yaxis.set_major_formatter(ticker.PercentFormatter())

![image-20211209231942028](../../../../../Library/Application Support/typora-user-images/image-20211209231942028.png)

Telegram bot for downloading covid-19 vaccine certificate

cowin-certificate-bot This is the source code of @cowincertbot, A telegram bot inspired by the whatsapp bot implementation of indian government for co

ArUn Pt 30 Oct 07, 2022
This is a python wrapper for "the best api in the world"

This is a python wrapper for my api api_url = "https://api.dhravya.me/" This wrapper now has async support, its basically the same except it uses asyn

Dhravya Shah 3 Dec 21, 2021
WhatSender is a python package that allows you to send Whatsapp messages at a given time.

WhatSender is a python package that allows you to send Whatsapp messages at a given time.

IdoBarel 0 Apr 21, 2022
Kodi script for proper Australian weather data

Kodi Oz Weather weather.ozweather Script for Kodi for high quality Australian weather data sourced directly from the BOM. Available from the Kodi offi

Jeremy Daalder 5 Nov 24, 2022
A telegram bot to track whales activities on multiple blockchains.

Telegram Bot : Whale Watcher A straightforward telegram bot written in python to track whales activity on multiple blockchains, using whale-alert API

Laurenz Bougan 1 Dec 10, 2021
Telegram Voice Chat Music Player UserBot Written with Pyrogram Smart Plugin and tgcalls

Telegram Voice Chat UserBot A Telegram UserBot to Play Audio in Voice Chats. This is also the source code of the userbot which is being used for playi

Dash Eclipse 7 May 21, 2022
Use Seaborn to visualize interpret the byte layout of Solana account types

solana-account-vis Use Seaborn to visually interpret the byte layout of Solana account types Usage from account_visualization import generate_account_

Jarry Xiao 15 Aug 25, 2022
A Telegram Message Manager Bot by @AbirHasan2005

Messages-Manager-Bot A Telegram Message Manager Bot by @AbirHasan2005. This Bot can delete specific type of messages from Group. I specially use for @

Abir Hasan 32 Nov 12, 2022
A wrapper for the Discord Python Pixels API.

DPYPX A simple wrapper around Python Discord Pixels. Requires Python 3.7+ (3.x where x = 7). Requires pillow and aiohttp from pip. Example import dpy

Artemis 3 Oct 01, 2022
Man-Userbot adalah userbot Telegram modular yang berjalan di Python3 dengan database sqlalchemy

Man-Userbot Telegram Man-Userbot adalah userbot Telegram modular yang berjalan di Python3 dengan database sqlalchemy. Berbasis Paperplane dan ProjectB

DzLyz 1 Feb 12, 2022
An integrated information collection tool

infoscaner 环境配置 目前infoscaner仅支持在linux上运行,建议运行在最新版本的kali中 infoscaner是基于python3版本实现的,运行之前首先安装python库 如果同时存在python2和python3,请输入以下命令 pip3 install -r requi

CMACCKK 74 Sep 13, 2021
Ulaavi for nuke, helps to keep our stocl elements organised.

Ulaavi Ulaavi for nuke, helps to keep our stock elements organised. Installation Downlaod ffmpeg from ffmpeg.org linux : https://johnvansickle.com/ffm

Arun Subramaniyam 17 Aug 24, 2022
A Telegram Video Watermark Adder Bot in Pyrogram by @AbirHasan2005

Watermark-Bot A Telegram Video Watermark Adder Bot by @AbirHasan2005 Features: Save Custom Watermark Image. Auto Resize Watermark According to Video q

Abir Hasan 95 Nov 20, 2022
Powerful spammer bots for telegram made with python and telethon.

Powerful spammer bots for telegram made with python and telethon. We can deploy upto 70 bots at a time.

32 Dec 15, 2022
A Python library for the Buildkite API

PyBuildkite A Python library and client for the Buildkite API. Usage To get the package, execute: pip install pybuildkite Then set up an instance of

Peter Yasi 29 Nov 30, 2022
You cant check for conflicts until course enrolment actually opens. I wanted to do it earlier.

AcornICS I noticed that Acorn it does not let you check if a timetable is valid based on the enrollment cart, it also does not let you visualize it ea

Isidor Kaplan 2 Sep 16, 2021
This project, search all entities related to A2P in twilio

Mirror A2P Twilio This project, search all entities related to A2P in twilio (phone numbers, messaging services, campaign, A2P brand information and P

Iván Cárdenas 2 Nov 03, 2022
Multi-Branch CI/CD Pipeline using CDK Pipelines.

Using AWS CDK Pipelines and AWS Lambda for multi-branch pipeline management and infrastructure deployment. This project shows how to use the AWS CDK P

AWS Samples 36 Dec 23, 2022
A simple bot to upload file to various cloud servers.

Cloudsy Bot A simple bot to upload file to various cloud servers. Variables API_HASH Your API Hash from my.telegram.org API_ID Your API ID from my.tel

Flying Santas 8 Oct 31, 2022
An advanced QR Code telegram bot with more features.

QR Code Bot A telegram qr code encode and decode bot Advanced Features 1. Database ( MongoDB ) Support 2. Broadcast Support 3. Status Command 4. Setti

Fayas Noushad 16 Nov 12, 2022