Dr David de Lorenzo
@daviddelorenzo
Followers
2K
Following
9K
Media
581
Statuses
4K
Researcher, Senior Data Scientist @TheNeotree @UCL @ich_ppp. Passionate about unlocking the secrets of data for better health. https://t.co/3HFQ8baffh
London, UK
Joined July 2010
The Faithful’s only real weapon isn’t deduction — it’s chaos. #TheTraitors #GameTheory #BBCTheTraitors #Mafia #Werewolf
0
0
1
They already coordinate the kill at night. Doing the same in daylight would have secured control faster. The difference? Strategy, not luck. The maths of #TheTraitors is ruthless: - Uncoordinated Traitors win most of the time. - Coordinated Traitors win almost every time.
1
0
1
At the start of the game: - With no coordinated voting between traitors → 69% win chance. - With coordinated voting (all Traitors target one Faithful when voting) & night kills → 95% win chance. By simply voting together, they would have had almost guaranteed win.
1
0
1
They’ve also been lucky: no Traitor has been banished yet. But they could have secured victory even earlier. In #TheTraitors the night kill is perfectly coordinated… …but the day vote often isn’t. From a game-theory perspective, that’s a huge strategic mistake.
1
0
1
Let’s apply that to the current state of the show: 8 players left (5 Faithful, 3 Traitors). Plugging it into the BEM equation, and you get a 96.9% chance of a Traitor win. Even without perfect coordination, the maths says the Faithful are almost out of time.
1
0
1
The maths behind this comes from the Braverman–Etesami–Mossel (BEM) model, which describes how the odds evolve as players are eliminated. Here n = number of Faithful, m = number of Traitors.
1
0
1
Each Faithful vote is effectively random. But the Traitors’ votes are deliberate and aligned. This creates an information asymmetry: the Traitors’ votes are worth more than the Faithful’s scattered votes. That imbalance gives them near-certainty of winning.
1
0
1
Based on game theory, the Traitors in #TheTraitors have already won the game. With 8 players left the maths gives them a 97% chance of victory. Why? Because when a small group shares information and a large group guesses blindly, probability bends in favour of the informed few.👇
1
1
1
Today's date in dd/mm/yy: 2²/3²/5²
0
29
144
Yesterday, 20 of the most ambitious startups from London’s universities took to the stage at London Demo Day - showcasing the incredible entrepreneurial talent emerging from UCL, LSE, King's College London, Imperial, and Queen Mary University of London: https://t.co/xZTwWXyLM3
1
5
11
You don't need a PhD to master bioinformatics. What you really need is: ✅ Consistency ✅ Curiosity ✅ One solid Bash script Master the basics. Everything else follows. #Bioinformatics #CareerTips #Learning
0
1
0
25 students. 5 finalists. $1000 to support further AI studies. What struck me most? Finalist Cyprian (centre), preparing his pitch, helped by Wisdom (right) and Morris (left), who weren’t selected. No rivalry. Just Ubuntu. I rarely see this spirit in Europe. #AI4Africa #Ubuntu
0
1
2
Just coming back to London after participating as a mentor and speaker at the #AIinHealth conference in Harare. Amazing to see the energy, innovation, and local talent driving the future of AI in health in LMICs. Thanks to all organisers and students for the inspiring work!
AI-driven predictive analytics, machine learning, & real-time data integration are transforming healthcare delivery in Zimbabwe and Africa. Day 1 of the AI in Health Conference showcased innovations accelerating diagnostics, treatment,& health system efficiency 🎥 @nrtvzimbabwe
0
1
2
What a great 3 days we have had - just wrapped up with a dragon’s den of early career researchers pitching their apps in TB, sexual health, diabetes, medical records and diagnostics. Blown away by the skills and innovation in 🇿🇼 youth #ZimbabweHasTalent
AI-driven predictive analytics, machine learning, & real-time data integration are transforming healthcare delivery in Zimbabwe and Africa. Day 1 of the AI in Health Conference showcased innovations accelerating diagnostics, treatment,& health system efficiency 🎥 @nrtvzimbabwe
0
2
6
Excited to be part of the AI for Health in Africa event! With my talk, "From Accuracy to Confusion: A Practical Guide to AI Model Evaluation", I'll be sharing lessons from my work with @TheNeotree on why evaluation & context matter, especially in low-resource clinical settings.
Get ready for 𝗔𝗿𝘁𝗶𝗳𝗶𝗰𝗶𝗮𝗹 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗳𝗼𝗿 𝗛𝗲𝗮𝗹𝘁𝗵 𝗶𝗻 𝗭𝗶𝗺𝗯𝗮𝗯𝘄𝗲 in Harare. More here ⤵️ https://t.co/K7uhYXnfKc
@UKinZimbabwe @imperialcollege @NIHRresearch @TheNeotree @ucl @wellcometrust @UniversityOfZim @healthtimeszim
0
1
0
In classical stats, multiple testing corrections (Bonferroni, Holm, FDR) help control false positives across many tests. But ML/AI doesn’t need them—no hypotheses are tested. Check my latest blog post at https://t.co/K7Grfgwt6f
#MachineLearning
#multipletestcorrection
drdelorenzo.com
In classical statistics, multiple testing correction is essential to control false positives when analysing numerous hypotheses. However, machine learning breaks free from this constraint by focusing...
0
0
0
This local gem needs our help to stay open! The McGahern Barracks Museum is crucial for our community and beyond. Let's raise the funds needed to keep it alive. Please share and donate!
gofundme.com
The John McGahern Barracks Museum is operated by a l… McGahern Barracks Museum McGahern Barracks Museum needs your support for Keep the McGahern Barracks Museum alive
4
5
5