Improved SorareData Lineup Projections
The SorareData lineup projections are a constant work-in-progress, as we are determined to make them as valuable as possible for our members. While we are satisfied by how many people participate, and on the general accuracy of the projections, we’ve received various feedback that we wanted to address to keep improving this feature.
The projections are based on the community, as each and every SorareData member can participate (and win prizes), as described in our previous post about the competition: https://soraredata.medium.com/projected-lineups-competition-updates-9c1e43a076fb
Because it’s community-based, it allows us to cover every match in every competition in every game week. However, it comes with a price: not everyone in the community is good at projections. In order to avoid some issues, we designed a weight system to ensure the best projectors were weighted more heavily than others.
And, as we’ve determined, that wasn’t good enough. For example, if a player has started the last 50 game weeks but turns out to be ill for the upcoming game week, most users would have had an equal weight in our confidence meter for this player because most of those projectors would have been correct over the last 50 game weeks.
Another thing to consider is that the members of the community who have been the most the most accurate with their projections usually wait until a few hours before the game week deadline to submit their picks. This shouldn’t be surprising, as that is when most information is available.
We are therefore evolving from our weight system to a more selective approach. As we get closer to each game week deadline, we will reduce the subset of managers included in the voting system based on their rank in the lineup projection seasonal leaderboards. This will also be reflected by the reliability indicator:
At the Low level, all projections are used; at Moderate, we have enough projections to only take into account the top-25 managers in the seasonal leaderboard; at High, we only use the top 10.
We have tested these modifications over the last few game weeks, and we hope this change will help managers make even better SO5 lineup decisions. It should also be noted that we explored other methods, including the use of an external data provider, but feel our evolved methodology produced better results.
There are numerous ways to participate in our Lineup Projections Competition, and we encourage all SorareData members to give it a try: https://www.soraredata.com/SO5Results/sections/community-lineups
