My original request was for the creation of a standalone Python script that would perform Monte Carlo simulations for
Agile estimation. I provided the developer with two articles that outline what I am trying to accomplish.
[login to view URL]
This article describes using Monte Carlo simulation to determine how many days it might take to complete a Scrum
product backlog based on story points. In this article the author describes how he groups the run times from a hundred
simulations into 10 buckets, each covering one-tenth of the time between the fastest and slowest. This aspect of the
script is not working as described here. The calculations in the script are based on grouping things into 11 buckets,
which I don’t understand and is not what I asked for.
[login to view URL]
This article is what the script is based on. It describes using historical sprint velocity as a baseline then applying Monte
Carlo simulation to determine the likelihood of completing X number of story points within a 10-day sprint.
The script as-is, does not perform the way I expected. Based on using the same methods described in the two articles, I
would expect the script to produce similar results, but it's not. When I execute the same scenarios as described in this
article, I get very different results. The developer chose to use gaussian distribution to accomplish the task, and it
doesn't seem to be producing the results I expected.
What I really need is someone who has experience with Monte Carlo simulations in Python to look at the script and
recommend what changes need to be made, then make the changes.
I'm happy to answer any questions you have. When you execute the script, it will prompt you for values to be entered.
Here are some sample values:
Sprint velocity: 114, 143, 116, 109, 127, 153, 120
Working days in sprint: 10
Forecasted sprint backlog points: 125 Number
of simulations: 1000
For reference, here are the two scenarios in the second article I am trying to reproduce.
Scenario #1
Team's velocity for 6 sprints is: 114, 143, 116, 109, 127, 153. Number of story points being forecasted for the sprint is
125. 1000 simulations are ran to assess the likelihood of 125 points actually being completed in their sprint of 10
working days. Result is 543 out of 1000 runs completed in 10.00 days or less, which is a 54.3% chance of the proposed
125-point backlog being completed within the sprint time-box.
Scenario #2
Product Owner agrees to drop a 3-point user story from the sprint backlog, so the new forecasted sprint backlog is 122.
After 1000 simulations are ran a total of 994 runs now completed in 10 days or less. That results in a 99.4% likelihood of
completing the work in the sprint backlog.
i too am working on monte carlo simulation in agile,
do kindly share across the exact requirements and also do lets discuss on chat to understand more on the requirements
Hi,
I graduated from IIT Kharagpur, one of the top engineering school in India. I have 8+ years of work experience in analytics and worked as actuary with companies like AIG and Swiss Re.
I am well versed with machine learning algorithms and programming languages like R, Python, SQL, VBA etc. I analyzed your problem and seem to understand what is happening. Gaussian distribution may not be a right choice for this. 6 data points you give do not reflect this. I can suggest some remedies. Would be happy to discuss more.
Best Regards
Chetan