Devoxx BE '23 Top Rated Talks
Dom Hodgson : Embracing Imposter Syndrome

Devoxx BE '23 Top Rated Talks

As usual we'll share the top-100 talks for our 20th edition of Devoxx Belgium.

This year was a bit different, as we now also have insights into the number of connected devices present in each room during each talk ??????

Of course one attendee can be connected via multiple devices, so the Max Attendees metric should be taken with a grain of salt!

Unique devices per room

This year, we successfully tracked the number of devices connected to each access point at 15-minute intervals. By aggregating this information for each room and pinpointing the peak value during the length of each talk, we were able to extract the maximum number of devices in each room for each specific talk ??

Each room had the following wireless stats!

Wireless log files per room

The "Unique Clients Trend Over Time - Client Count" was the most interesting CSV to calculate the number of unique users per room. Extract of the CSV:

Unique users per interval and room

By correlating the start/end times with the Devoxx Belgium 2023 schedule, we can pin point what the total connected users were for any given talk ??

Hey ChatGPT. What are the best formulas to calculate top rated talks based on total ratings, average rating, number of attendees and maximum capacity of room?
The suggested formulas by ChatGPT

Typically, the most time-consuming part of the process is aggregating the data and integrating it with additional resources. Once you've compiled the final CSV, ChatGPT can handle the rest effortlessly through its 'Advanced Data Analysis' plugin.

The awesome part about multiple statistical perspectives is that you can cherry-pick the one that makes your talk look the best ??

Let's have a closer look...

The Popularity List

I did add one extra "formula" which was the popularity view, basically the maximum number of users/devices linked to the room capacity.

Popularity formula

The Popularity Score formula can be summarised as a ratio of the number of unique attendees to the maximum room capacity, scaled by 100. In mathematical terms.

This formula gives a percentage-based score, making it easier to compare talks across rooms of different sizes. It's a simple yet effective way to gauge the popularity of a talk while accounting for the room's capacity.

Most popular talks were :

Popularity score top-10

As you can see there were more connected devices than seats available, which makes sense when attendees could have a connected portable, phone, tablet and phone during a session.

Room 8 packed during opening keynote (with 3 overflow rooms)

The Weighted Score

The Weighted Score formula combines both the Average Rating and Max Attendees of a talk. It assigns 60% weight to the Average Rating and 40% weight to a normalized version of Max Attendees. This balanced approach ensures that both quality (ratings) and popularity (attendance) are factored into the final score, providing a comprehensive measure of a talk's overall impact.

Weighted score top-10


I've also included the top-100 list for all the other formulas, which can be found

???? HERE ????


Thanks again to all the speakers for making the 20th edition of Devoxx Belgium an amazing experience.

Make sure to mark your agenda for our 21st edition which will take place in the same venue from October 7th until the 11th. See you there!

-Stephan

Exponential Weighted Top-50 list




Corneil du Plessis

Staff Software Engineer at VMware by Broadcom on Spring Team

1 年

Am I the only one who tried to right click on the columns to format numbers for 2 decimals?

回复
Gerald Venzl

Strategist | Technologist | Keynote Speaker

1 年

Pretty cool, well done! Next step: loading that data into a #database and compare trends over the years :)

回复
Luc Duponcheel

Retired Mathematician and Functional Programmer

1 年

Keep on doing such an excellent job Stephan!

Holly Cummins

Senior Principal Software Engineer ? Making Developer Joy Happen

1 年

IANADS, but that exponential weighting formula doesn't look right to me. It should be said that the reason I noticed this is because it wasn't very favourable to me, and my talk scored more highly using the other formulas, so *of course* I don't like the exponential weighting. ?? But I think there's a normalisation factor missing. The formula suggested by ChatGPT is # attendees + e^(average rating). The maximum average rating is 5, which means the maximum of e^(average rating) is 148.4. Now let's imagine Devoxx makes a commercially reckless decision, and launch Devoxx Marshall Islands, but only 10 people turn up to the conference. The most popular session, "how to make money with pyramid schemes," has 9 attendees, but the speaker is awful and everyone rates it 1, so its exponential score is 12. On the other hand, 1 person goes to the session on "why you should use punchcards in 2023," but they love the session and give it 5/5, so the exponential score is 153. But only one person went, so the amazingly high score doesn't seem right.

Julien Top?u

Vice-President Technology SHODO Group | Socio-Technical Coach & Architect at SHODO Paris | International Keynote Speaker | Founder of BeyondxScratch, CraftsRecords and le Tremplin

1 年

Hi Stephan Janssen ! What is the difference between the weighted score and the exponential weighted score ?

回复

要查看或添加评论,请登录