A new puzzle about daily STEEM production (yes, really.)

What does his lucid explanation amount to but this, that in theory there is no difference between theory and practice, while in practice there is? Benjamin Brewster, The Yale Literary Magazine, 1882

OK, I realize that I am like a dog with a bone here. Sorry. ;-) I guess the appropriate response is probably, "Just let it go!" But, I'll throw it out here, anyway...

This week, I added a new number to one of my weekly visualizations so that I could compare actual/observed new STEEM per day vs. the predicted value. Here's what the full slide looks like after the update:

And here's a blow-up of the visualization that I'm interested in - from the middle row in the slide:

The value that's new is the gray one, labeled, "Avg. Observed New STEEM per day (left)". In particular, I'm interested in the roughly 1,650 STEEM per day difference between that value and the blue one labeled, "Predicted New STEEM per day (left)".

The formulas for the predicted value came from this link by @kouba01, and I think I've verified those formulas in Steemit's github site. Once I have the new STEEM per block, I simply multiply it by 28,800 (20 * 60 * 24) in order to come up with new STEEM per day.

The observed value comes from this page on steemdb.io. It's a simple average that's calculated by subtracting the virtual supply at the beginning of the time window from the virtual supply at the end of the time window and then dividing that difference by the lapsed time (in minutes) and then scaling it up to a daily value.

The numbers are pretty close, and I don't expect them to match exactly, because some STEEM is being burned. In principle, I was expecting the observed value plus the number of burned STEEM to add up to the predicted value. And, conveniently - I'm tracking the burned value, too. On average, during the relevant time frame, we've burned about 955 STEEM per day ((9402+6838)/17), as we can see, here (the numbers came straight from the blockchain API):

With that factored in, as far as I'm aware the numbers should reconcile. But they don't. If we subtract the burned amount from the predicted amount, we're still off from the observed value by about 700 STEEM per day.

Given that we're talking about a blockchain, where accounting is fundamental, it would be nice to be able to reconcile the predicted value and the observed value.

At this point, my one and only guess is that it's due to the changing value of STEEM/MVESTS as time passes, since the burned SP is estimated by converting from MVESTS at a single point in time. Over the course of 17 days, though, I have a hard time believing that there's enough of a change to account for the size of a discrepancy. Also, I would expect to see the gap closing over time - but we don't see that. Any thoughts on how to test that without too much effort? I do see that the STEEM/MVEST value is included in the same steemdb.io link.

Any ideas about what I'm missing?

---

Visit the /promoted page and #burnsteem25 to support the inflation-fighters who are helping to enable decentralized regulation of Steem token supply growth.