For discrete question forecasting input, the Y axis is labeled and the out of bounds values are graphed. In uncommon, but not rare, situations, the out of bounds value can far exceed the largest inbound pmf value. This leads to a large bar on the side of the graph, and the shape of the distribution within bounds getting shrunk. The proposed solution would be that if the first or last value being graphed far exceeds the largest inbound value (let's say by 2 times but let's keep that a parameter so it's easy to tweak), then the y-axis should get a break in it such that the out of bounds pillar is no more than 2x the largest inbound value.
Related slack message:
https://metaculus.slack.com/archives/C01Q9AQBVHB/p1784649668288529?thread_ts=1784400978.119999&cid=C01Q9AQBVHB
For discrete question forecasting input, the Y axis is labeled and the out of bounds values are graphed. In uncommon, but not rare, situations, the out of bounds value can far exceed the largest inbound pmf value. This leads to a large bar on the side of the graph, and the shape of the distribution within bounds getting shrunk. The proposed solution would be that if the first or last value being graphed far exceeds the largest inbound value (let's say by 2 times but let's keep that a parameter so it's easy to tweak), then the y-axis should get a break in it such that the out of bounds pillar is no more than 2x the largest inbound value.
Related slack message:
https://metaculus.slack.com/archives/C01Q9AQBVHB/p1784649668288529?thread_ts=1784400978.119999&cid=C01Q9AQBVHB