Well that was quite the episode! From Episode 5, Figure 2 and Table 3:

• Phil won his first episode in the series! He finished in 1st place with 18 points.
  • Unsurprisingly, this was Phil’s best episode to date and after highlighting this observation in last weeks post, Phil finally displayed moments of brilliance which contributed to this win.
• Reece finished in 2nd place with 14 points.
  • This was considered a bad (below average) performance by Reece as it in the lower half of episode performances to date.
• Sanjeev accumulated 12 points and finished in 3rd place.
  • Like Reece, this was a bad, below average performance by Sanjeev.
• Ania placed 4th with 11 points in total.
  • Unsurprisingly, this was Ania’s worst episode performance date. To date, she has never scored episode points as low as this, or placed as low as this in an episode.
• Maisie rounded out the leaderboard, accumulating 7 points and finishing in 5th place.
  • This is Maisie’s worst episode to date, and is also the lowest scoring episode so far in the series.
  • Her anger is no doubt justified, but also very entertaining.
• The outcome of the live studio task proved to be extremely influential in the episode and current series standings.
  • Phil and Reece being awarded for their deceitful behaviours, and negative penalty points for Ania, Maisie and Sanjeev ultimately influenced the outcome of this episode.

From the Task Perfromance Table 5:

• We see our first instance in the series of negative points being awarded.
  • [Ania, Maisie, Sanjeev] were penalised -3 points from the live lamé studio task.
• Phil finally broke his rut of not performing spectacularly in prior tasks; he was awarded the full 5 points in 2 tasks.
  • Phil’s win in this episode can be attributed to being awarded no less than 2 points.
  • Phil’s overall task performance will continue to be centered around the low to mid points range, but he now has likelihood of a stellar performance and being awarded 5 points..
• Apart from the live studio task, Ania had a average episode, with points being awarded from either end (1 to 5).
  • Ania has still yet to be awarded 4 points for a task.
  • Ania’s task performance distribution is unlikely to change substantially.
• Maisie’s poor performance in this episode is due to 3 of her tasks scoring 1 point and below.
  • Maisie only had two tasks in which she performed well and was awarded 4 and 5 points,
  • Maisie has previously been disqualified (awarded 0 points), but this disqualification was more targeted (Maisie and Reece suffered), whereas the prior disqualification affected the entire cast.
  • Maisie’s -3 penalty also no doubt added to this terrible episode performance.
  • Maisie’s task performance distribution will become flatter and will be less pronounced
• Ignoring Reece’s task in which he was penalised and awarded 0 points, Reece had a pretty good episode; he was awarded no less than 3 points and ranked 3rd and above for these tasks.
  • The bimodal, extreme nature of Reece’s task performance distribution will less evident following the tasks from this episode.
• Whilst Sanjeev was graded as having a “bad” episode, this can be mainly attributed to the effect of the -3 points awarded from the live task.
  • If we omit this task, Sanjeev didn’t perform as badly, scoring between 2 and 5 points.
  • Sanjeev’s task performance distribution is not anticipated to change greatly; it will retain the same basic shape (skewed to the low to mid tanks) but will be flatter.

And following the surprise outcome of Episode 5, we also see a shake up in the series leaderboard (Figure 6):

• The updated series rankings following Episode 5 are:
  • Ania maintains her 1st place position with 78 points.
  • ot on the heels on Ania is Reece; he places 2nd with 77 points.
  • Also closely behind Reece, are both Maisie and Phil in joint 3rd place with 73 points.
  • Sanjeev rounds out the pack ending with 67 points in 5th place.
• As is evident, the Lamé Live Task has had a substantial impact.
  • Phil has now transitioned from the lower ranks (the 4th and 5th placements shared with Sanjeev) and joined the higher ranks (1st, 2nd and 3rd positions shared amongst Ania, Maisie and Reece).
    • Overall, the two sub-groups we have seen from Episode 3 and 4, is less evident following Episode 5.
  • The -3 points awarded by the Lamé Live task had a detrimental impact on Ania and Maisie.
    • Ania lost her strong lead over Reece, such that it is now a close race between them.
    • Maisie lost her position as a likely, comfortable forerunner of the series. She could regain her position, but she will require lots of good performances going forward.

Figure 10 compares the latest estimated ranking distribution to the empirical distribution based on observed episode rankings to date. If our estimation method is reasonable, the estimated distribution should not be to dissimilar to the empirical distribution.

We observe:

• Reece’s flat, uniform estimated distribution mimics the observed empirical distribution; Reece has finished in all positions to date.
• For the remaining contestants, the most likely outcome under the estimated and observed distribution don’t quite line up, and we don’t see the same level of flatness in the estimated distributions.
  • For Phil for example, he has placed 4th for 60% of episodes so far; however, our estimated distributed shows that Phil’s most likely episode ranking is 3rd place with 23.55%.
  • Whilst this is unfortunate, we stress that empirical distribution is based on 5 episodes and observations only. This may not be a sufficient sample size to form an accurate distribution we can make inferences from.
• We see many instances of contestants not placing in particular positions, but we still have estimated probabilities.
  • For example, Ania has not placed 5th in an episode to date, but we estimate that she will place in this position with 11.02%.
  • Whilst this may seem odd (having a probability estimate for an event which has yet to occur), it is a benefit of our simulation based estimation method and sampling a such granular level (at a contestant-task level).
• Our method also provides a smoother probability distribution that are less disjointed compared to an empirical distribution.
  • Maisie for example, is a woman of extremes in reality with regards to episode ranking; she has placed 1st and 5th each with 40% occurrence to date. The remaining 20% is for placing 3rd and Maisie has never placed 2nd or 4th.
  • Maisie’s estimated distribution has a smoother trajectory, placing the most probability on placing 1st and decaying steadily from there.