Reduced market shares

Comparing market shares between this week and last week, it seems there are only 8 people whose market shares increased this week, everyone else had reductions, with many of the reductions being quite significant. What would cause this?

Each dot on the graph below represents a different person’s entry, with the ranking for week 9 along the x axis and the change in market share between week 8 and week 9 along the y axis. The points in red are new entrants who made their first submission in week 9, who are assumed to have had 0% market share in week 8.


I assume market shares are taken from the realistic market where each person’s model competes with 9 others from the top 10% of models. There are a number of new entrants this week who have significant market shares, so I guess the people in the top 10% would see significant reductions in market share because of this.

I’m less sure why people outside the top 10% would see significant reductions in market share because, in the realistic markets, they don’t directly compete against the new entrants with high market shares. They can only be losing market share to people in the top 10%. My guess is that there are a couple of models in the top 10% taking high market shares when competing against most other models (giving lower market shares to people outside the top 10%), but these models have much lower market shares when competing against some of the new entrants, so the average market share of these couple of models is also lower than their shares last week. Could that be what’s happening? Any other ideas of what could be going on?


Hi @Calico, is it possible to know where you found this graph ?


@Baracuda I copied the market share data shown on the week 8 and week 9 leaderboards and made a graph from it. I was trying to get some insight into how changes in market share and profitability are linked, but the extreme movements in market share make that difficult.

hmm, yes indeed, it is difficult to draw a conclusion from others competitors info unless the model submitted is fix from one week to another :slight_smile: