Re-insurance and bigger markets! 🥳

Last week, we asked you, our community, how we can increase engagement in the competition…and we received a resounding response on some key ideas.

I am thrilled to share that we are incorporating more and more of your feedback, and today, we are announcing 2 major updates. (stay tuned as more updates will be coming soon)

  1. Capping of claims at €50K. We have implemented market-wide excess-of-loss re-insurance for claims above €50,000. We are effectively capping any claims above €50K at €50K.
  2. Tripling leaderboard market size. We are increasing the weekly leaderboard size from 6K to 18K rows

These updates are already applied to the week 6 leaderboard :muscle:

Why capping claims at €50K? :thinking:

In the real world reinsurance is optional and companies can negotiate the price and buy it. One among many reasons for this is to protect insurers (that’s you) from extreme claims.

What is an extreme claim? Different firms assess this differently but one common strategy is by looking at your revenue. If your company had a revenue of €1B and one claim cost €50M, then 5% of your revenue went to one claim. That is an extreme claim and this % is called a Combined Ratio for those familiar with the terminology, simplified here for our markets.

What we are doing here is to introduce a re-insurance cover that protects against:

  1. Claims above 2% of the total market revenue of each leaderboard (in reality this number is closer to 1%)
  2. Claims above 1% of the total market revenue for the final evaluation leaderboard.

€50K corresponds to these values in our markets :point_up_2::point_up_2:

To put your mind at ease, this only affects a handful of claims dataset-wide.

Why more data? :weight_lifting_man:

There are two key reasons for doing this:

  1. It further enhances the effect of re-insurance when each model has more revenue by being able to win more policies.
  2. A larger dataset is more representative of the final leaderboard in terms of size.

How does it affect the challenge? :hammer_and_wrench:

  1. Your feedback. The reinsurance combined with larger markets will lead to less market volatility and less of the lottery effect in your feedback so that one claim does not completely wipe your profits away.
  2. The training profit. As with a real insurance company, you would account for capping of claims in your pricing and claim estimation, therefore we are applying the same rule to the training data. That means we will check the training profit rule with the €50K cap in place. You can download the data from the resource tab on the challenge page if you wish.
  3. The RMSE leaderboard. This will be updated and all submissions re-run. Expect your score to go down, but your leaderboard position to remain (roughly) unchanged. This is because these handful of claims are largely unpredictable and likely affect everyone in the same why, by inflating their RMSE scores.

Lastly, as always we have new feedback cooking up for you, so stay tuned this week :headphones:

I hope this will help you improve your strategies and not worry about rolling the dice in the markets.

Well, worry a little less :game_die:

Good luck in the markets!

Ali, The pricing game team

7 Likes

Thanks for this update. This competition stays fun and exciting! :partying_face:

Quick question:
With regards to the second update, the trippling size.

  1. Tripling leaderboard market size. We are increasing the weekly leaderboard size from 6K to 18K rows

Are you referring to this on the main page?

WEEKLY PROFIT LEADERBOARDS

This contains a total of 30K policies with 4 years of history (~115K rows). It is split into 9 weeks such that:

  • Weeks 1 - 5 each use approximately 7K rows of data from 15K policies with 4 years of history
  • Weeks 6 - 9 each use approximately 20K rows of data from 30K policies with 4 years of history

Or is it something else?

1 Like

Hi @michael_bordeleau

I updated the overview page today! Yes it used to be 7K and from today it will be around 20K. :slight_smile:

Is the “reinsurance” going to be in place for the final evaluation too (or is this only a change to the weekly leaderboards)?

It will be in place throughout, including in the final evaluation :slight_smile: