That doesn't sound stupid - makes perfect sense in fact. That is a lot like run or control charts used in production. The idea is that as machines wear or other variables change slightly over time the average value for a given product parameter will change over time. So the question is do you target the average value of the last batch, which can result in an average that varies over time, or do you stick to another standard such as a total historical average or the value of the original master standard established at the outset of production.
From memory I think the method you refer to - the updated average - results over the long run in a wider variation in the total population produced historically tan if you try and control the process to stick the the original master standard. But a market is not a product which you can control to a original master standard or which needs to be controlled to a customer specification (which the master standard would be set to) - it is what it is - so using the average of the last "batch" (in this case the last 12 months) is the better approach than sticking to the long term average or original average, because in theory this year's batch should be more similar to last year's batch - the only caveat being this assumes the market conditions change slowly over time compared to the batch duration (if it changed more quickly then you may need to re-calibrate every 6 months or 3 months, etc).
