RetroPoints 2.1 Proposal

[kiwitaco]

First let me start by saying that I'm a big fan of introducing diminishing returns on Retro Points, and that I welcome bringing the point where such a dampener starts applying much lower than what previously proposed. Unconstrained linear growth was absolutely ridiculous and set the expectations of the community in a weird place. For that reason, I think this is a step in the right direction, where Retro Points more accurately reflect rarity. I believe, however, that it went too far in nerfing the ratio of rare achievements, in a way that introduced problems listed below.

The PlayerCountAdjustment seems to have been well received, so this proposition does not touch that adjustment in any way.

Problems

• The formula fails to capture a proper gradient between rare achievements. For reference, an achievement with an order of magnitude more rarity (e.g 1 in 40,000 vs 1 in 4,000) receives 40% more points than the other. Previously, they would have received 400 times more points (i.e 40,000%) which clearly was absolutely ridiculous. But 40% fails to capture rarity in the case where they arguably matter the most. Without getting into the most egregiously rare achievements on-site, you get to ~50-70% more points for something ~10 times more rare, which again would have 40,000% more points before the change.
• While the dampener starting at a certain percentage of the player base is not the worst concept, I don't see why that is the case. When using a much higher threshold, like in the original proposition, it made more sense, because it was applying to extremely low unlock rates, and that notion depends on how many players there are on-site. Now that the team settled on starting diminishing returns at around 1 in 250, this threshold has no reason to vary depending on the total player count. If the site only had 5,000 players, that ratio would still make sense, and we wouldn't want it to start with a 1 in 10 players.
• The "soft cap" is used twice in the computation for two entirely different purposes. One is to determine when it starts applying, which was addressed in the previous point, and the other is to completely change the curve. Dividing by it before applying the log means that the curve is allowed to grow more when we have more players. This effectively duplicate the effect of new players, allowing high unlock ratio by its nature AND increasing the retropoints of every achievement on-site with an unlock rate over the threshold. This seems clearly unintended. If we get 50k new hardcore players, it should not just buff the retropoints of every rare achievement across the site. This is the current behavior. If 50k new hardcore players are logged in tomorrow, every achievement with an unlock rate below 1 in 250 would be buffed by a good margin (talking 2 or even 3 digit percentage).

Proposed Solution

• Change the formula to use a fixed threshold of 1 in 250.
• Change the formula to use base 2 instead of base 10.

This allows for dampening to start as early as it is currently, but with a slope that isn't as sharp, allowing for a greater gradient. This helps distinguishing achievements that are rarer with more granularity while still having a strong dampening effect that make it impossible to get a multiplier above 1000 with the current player base.

For reference, here's a dynamic graph that allows you to test various parameters. At the moment, a= is roughly 180 (as it uses the number of players with at least 1250 retropoints for the base total, which might be a bug as the code comments clearly state the intent was roughly 250). And b=10. The proposal is to set a=250 and b=2. You can easily test the effect the change https://www.desmos.com/calculator/yyh34zir58

On that sheet, you can easily test the effect of various parameters and compare it with the old and current implementation of retropoints.

Sample:

0.4% unlock vs 0.04% unlock rate vs 0.01% unlock rate vs 0.005% unlock rate

Current: 83x | 155x | 198x | 220x
Proposed: 101x | 433x | 633x | 733x
Old: 101x | 1000x | 4000x | 8000x

[kiwitaco]

While changing the base to 2 is a big change, I believe everything else should be something to consider as soon as possible. The fact that the multiplier starts to be heavily reduced at a variable point depending how many total hardcore players there are kind of makes sense, but the reduction itself being tied to it makes this more unstable than it should be.

For reference, over the last 2 weeks, the number of hardcore player grew enough to make very rare achievements gain more than 5% retropoints. That's not the number of players playing a certain game that affected the rarity of achievements of that game, that's just the total number of hardcore players growing. Every very rare achievements growing from 15k to 25k retropoints each over the next couple months seems unintended.