Thursday, 15 August 2013

Are you better than a Coin Toss?

This evening I went along to SkillsMatter to watch Richard Warbuton and John Oliver give their talk "Are you better than a Coin Toss?"


  • Having a backup for the presentation platform (i.e. second laptop/iPad) is tantamount!
  • Test your presentation beforehand! Fail fast!
  • Keep water to hand, fizzy drinks don't help.
  • Always good to adapt (i.e. basic introduction to buy boot up time)
  • Er's & Um's distracting
    • Use a pause before speaking to build confidence and removes need to think mid-flow.
    • This goes away with practice
  • Great to use examples away from the core field (like Moneyball)
  • Generally great to illustrate points with concrete examples
  • Beware of standing in front of the slides and blocking them!
  • Try to speak slowly, looking for cues that your audience has understood you.
  • Great to have humour intermixed into the presentation!
  • Good to reaffirm each others points.
  • P-value diagram didn't really help to illustrate the point to those who don't know about it, try something more universal? Dartboard with % probabilities of hit area?
  • Beware naming things part 1, 2, 3… because it can lead to "we are only on part 2?!" syndrome
  • Avoid reading off slides directly
    • You say what you need to, the slide should only illustrate your point
  • Throughout it felt like a few different slides or an animation could have better illustrated the point
    • On the clustering slide, show actual clusters and why they are good/bad
    • On the elbow slide, show a few elbows and highlight your point
    • Good example is the over/underfitting, the pictures made the point perfectly!


  • Producing useful data analytics is really hard.
  • Using hard data analysis techniques is better than using subjective expert opinions.
    • Science FTW!!!
  • Without concrete and unbiased measurements, algorithms may be not better than randomness.
  • Holy cow how much work does jClarity do with their mad polygot skills!?

No comments:

Post a Comment