Thanks @Calico for sharing your approach! I tried (like many others) to test different ranges between rounds but got discouraged when I saw people who didn’t update their submissions were going up and down the leaderboard ; really cool to use them to get a baseline !
It is also fun to see that, at the end, you targeted a conversion rate very different from mine (5% for you vs 10% for me)… and we might both miss our target given how much the market will vary in the last round 
@kamil_singh the graph with fuel type on the x-axis and the relativities on the y axis was created with Akur8 ( https://akur8-tech.com/ ) actuarial modeling and pricing solution (I work for the company developing it… and it is a great pricing platform, of course ! 
)
@chezmoi : in general I would clearly say that demand models need a good and well identified price-sensitivity component to provide actionable results (and it brings really huge value). But here it was clearly a futile experiment, mainly for fun. While my estimates may be true for the week 10 market, it will not be for the final one, given how much the competition premiums changed. It also provide no clear vision on the individual conversion rate and behavior, as you point out.