I created a visualization of the evolution of the distribution of ring member age over time.

Monthly:

Weekly:

"Week" defined as here:

Thanks to neptune for database management and gingeropolous for computing resources.

Remarks:

1) in late 2017 into early 2018 a clear bimodal distribution forms. It almost looks bimodal-triangular. I know at one point Monero had a triangular distribution. It is strange to me that it looks bimodal at this point. Towards the end of this phenomenon in the weekly visualization, you can see a triangular "wave" move backwards from the first triangle to the second. It may have something to do with transaction volume.

2) In March 2019, the currently-used gamma distribution decoy selection algorithm becomes dominant. According to my initial investigations, the code for the gamma distribution was released in July 2019, so my initial investigations are apparently incorrect. I will investigate in -dev.

3) In 2019 and later, occasionally you can see large isolated spikes in the weekly visualization. I think that may relate to the cached decoys issue that isthmus discovered recently.

4) You can faintly see a small spike in young ring members during the July-August 2021 transaction volume anomaly, consistent with the findings of the Fingerprinting a Flood paper.

This is really cool!

Do you know what the consistent spike for old ring members is from? It used to be more pronounced before the gamma distribution kicked in, but was still there aftwards

jeffro256: It could be a "third-party" wallet implementation or exchange still using an older decoy selection algorithm.

In fact the question that sparked this visualization was: How quickly is a new decoy selection algorithm adopted once it is released in the official GUI wallet?

I've looked more into when the gamma distribution was released. It seems it was first introduced in v0.13.0.2

in October 2018. In July 2019 with v0.14.1.0 some adjustments were made.

Maybe something that could be interesting is a visualization which maps the % change from the perfect gamma distribution to the actual distribtuion for each time period (week/month) like you did here

So ideally the graph would be a flat line, and deviations from the flat line would show how the the gamma distribution doesn't show real spending behavior / older wallet selection algos which haven't adopted the new gamma algorithm

cool stuff. definitely a tricky dataset to visualize. nice approach. would be cool to have a timeline going with the gif with the various development activities on there

for instance, up until.... whenever, there was no min ringsize, and you could make a custom ringsize.

wait so the x axis is the age of the input?

ok, so a day old is about 11

Yes, the x axis is the natural logarithm of the number of seconds that had elapsed since the referenced output was confirmed in a block.

By the way, for anyone wondering what a kernel density estimate is, it's basically a more sophisticated histogram.

Rucknium: posted on MO: monero.observer/rucknium-shares-vis…monero-ring-member-age-distribution