Discussing Potential Leaderboards Rating Adjustments

This thread will be home to a lengthy discussion on potential changes to the Leaderboards ratings for certain levels. I'm not going to promise that I'll change anything discussed here, I'm only opening a discussion for such things. I will commit to changing PQ Hunt and Pack 7 which have both long since needed nerfing. The rest is optional as far as I'm concerned and need strong reasons for change. Much discussion and contemplation must take place before I would commit to changing any of them. 

If you're curious about seeing the rating curves for any levels use this link . You can also adjust the setting below the graph to mess with the curves for yourself (they will not apply to the LBs, it's just for fun). 
 

This forum post was created in response to a fairly in depth discussion in Discord about this very subject. My summary of that discussion is as follows. 

Ideas shared started surrounding Singleplayer Hunt in PlatinumQuest. Gravity Tower was a level that had it's rating for the WR surpass 1 million when Mazik discovered an overpowered strategy on it to blow the old record out of the water and pushing the rating curve further than the developers ever thought possible. I very quickly nerfed that rating curve and now it's much more in line with the rest of PQ's Intermediate levels. However, many other hunt levels were pushed much further than the devs thought possible as well, their curves didn't explode like GT's but they're still granting much more rating than the rest of the levels in their respective difficulties. I made a commitment to nerf those rating curves like I did GT, and I set a deadline for myself at the end of this month, Jan 2026. Other people mentioned that a few other levels (namely: Hydropower, Miscalculation, Manic Bounce) have inflated rating curves as well, and this sparked ideas surrounding which levels deserve 1 million rating. I will return to that idea shortly. 

Of course PlatinumQuest is not the only set of levels with perceived problems with it's rating curves. Many custom levels have wildly inconsistent rating curves, some very difficult levels have relatively low yields, and some other relatively easy ones have much more bloated curves. Instinct being a level that was never thought possible to get a time below 2s now yields over 800k. No Jumping II also being a level that much later became a catch level now gives over 1m rating. Many levels discussed believed to give too little rating as well. Comparisons were made from easy levels giving more rating than other more difficult levels, and general consensus of the vocal participants of this conversation seemed to suggest that Leaderboards Customs needed adjustments here and there. 

On the topic of 1m rating levels. Currently the levels on the LBs where a WR gives over 1m rating include: Morph, Catwalks, The Tale of the Tall Skyscraper, Mastering the Marble, Space Station, Battlecube Finale, Gem Finding Folly, Hydropower, Miscalculation, Manic Bounce, Xployt, No Jumping II, Fun Cage, Moving Platform Master Course, and Stronghold. Much discussion took place about whether or not any level deserves 1m rating, and which levels would be deserving. The joke of a suggestion making DA HOUSE give 1m rating was brought up lots. Otherwise the discussion seemed to think only certain selected extremely difficult or extremely long levels deserve the honor of granting 1m rating for either their WR or in the case of Manic Bounce for merely beating the Awesome Time. 

A Few other noteworthy suggestions were made. Such as custom levels should be organized by difficulty (Beginner, Intermediate, Advanced, Expert) rather than by pack. That or each level should be granted a difficulty in line with the BIAE grading system and their ratings adjusted accordingly. Pack 7 was brought up a lot, this has long been universally agreed to need a nerf, the levels in it consistently give ~30% too much rating, especially since many of the levels have times that much exceeded the devs expectations. And lastly some talk of reworking MBP Expert's ratings were brought up. 

Thus concludes my summary of last night's discussion within Discord. This should give some starting points for the discussion surrounding Leaderboards Rating Adjustments. I want to hear all your ideas whether they're based on anything mentioned here or not. 

Honestly the rating system could be reworked into something like the best time on the level / your time on the level * 100,000 then rounded down to the last digit = your score on the level. Then add up all the scores per level to get the leaderboard ranking like normal. I think all the levels should give the same max score because the harder levels will naturally have a higher skill gap anyways so it makes more sense to focus on harder levels once someone finishes the easier levels (which shouldn't be hard to grind).

This is a suggestion that comes up a lot when discussing the rating system we use, and it's one that I fundamentally disagree with. I hate the idea that you could grind out good times on a level, then through no fault of your own your own rating goes down because other people beat your time, or because someone beat the world record. I much prefer the system where you get a time, you get your score, and that never changes (barring some rebalancing like we're planning here). 10 years from now I'll still have the same ratings I have now even if I never play this game again. That way your hard work isn't erased by other people's better times. The difficulty goes up in either system, but at least in the system that we have your work is never erased either. That's my quick take on that. 

I like the idea of giving custom levels Beginner/Intermediate/Advanced/Expert difficulty ratings. Would that also involve shuffling the levels in each pack around by approximate difficulty, so that they go from easiest to hardest? I like when that is done with the multiplayer levels, so it would be cool to see it for the single player customs as well. Then the ratings could be adjusted accordingly.

Something else to keep in mind is achievement rating. For example, there's an achievement for getting the Gold Time on Tightrope which gives the player 500,000 rating. So Tightrope effectively becomes a 1 million rating level if we take that into account. There are also achievements like "The Dragon's Destiny" for hitting 60 million rating, which used to be a real challenge in 2014 but has since become an easy achievement due to the huge amount of rating available online. I know there's not really a way to take away achievements, but I just thought I'd bring it up.

Ultimately I don't play enough to be very familiar with the ratings that certain levels give, but I disagree with the idea of certain levels "deserving" 1 million rating. I think a better approach might be to have a big community vote on the difficulty of each level. This could be done with a Google Form, or if there was an in-game rating system like for multiplayer where each user could select one of four difficulties to give each level, we might get more accurate results. But anyway, 1 million is an arbitrary number, and I think that it's better to assign the rating curves, and then if the WR happens to get 1 million, so be it, and if it doesn't, then that's fine too.

Maybe instead of looking at WR on each level, we look at the average score compared to the high score, or the range where the most scores are concentrated, and use that to determine a rating curve instead.

The PQ levels: Gems Ahoy!, Hydropower, and Gunnery Practice, have their rating curves set to barely below the standing WRs, which are all at the very least possible to beat by enough to mess with the rating curve, with Gems Ahoy! just breaking it again. I'll see if I can't fix these along with the rest. 

In January I began working to adjust the ratings of the hunt levels in PlatinumQuest. The scale of this project, combined with heightened real-life responsibilities, quickly overwhelmed me and I set this aside. Months have gone by with no work done, and no updates from me. I'm returning to this now with a better understanding of the scope of the task I face, and a clearer sense of how to approach it.

I understand this is an undertaking that should not be taken lightly, and that this is not going to happen without confusion, disagreement, and strong reactions. My goal with this post is to reduce the confusion and build trust around these proposed rating changes by explaining why I believe that these adjustments are necessary, what principles I plan to use, and where I would like input from the community. This should hopefully lead to us coming up with a plan on how to adjust the ratings in a way that is fair to the game and the skilled players who have taken these levels farther than anyone originally imagined possible. The goal is not to rebalance the entire rating system of PlatinumQuest, but to adjust a handful of levels that give far more rating than they were originally calibrated to give.

To explain what I mean by originally calibrated: most of the rating curves were set by the original developers. Each level was designed to grant a broadly comparable rating to the other levels in their respective difficulty, with higher difficulties granting more rating. But as it currently stands many hunt levels give noticeably more rating than the other levels in their respective difficulties. For example, the world record for Hunting Around, a Tutorial level, gives enough rating that it would not be out of place, rating-wise, in the Expert levels. The original developers did their best to set curves that would grant fair ratings. But they had no idea just how skilled players would become and the heights to which these levels would one day be taken.

I want to carefully balance my approach between two extremes. On the one hand, I want to avoid over-nerfing hunt levels, making these levels not worth the effort to play and devaluing the hard work many players put into pushing their limits. And on the other hand, I want to avoid doing nothing, and allowing these ratings to continue drifting further from the normal range of their difficulties. The question I asked when starting this project is, how do I best approach adjusting these levels without becoming too extreme in the changes I make?

This is how I approached this task back in January. I needed to find a way to judge the rating that a level gave, without relying too heavily on intuition, while minimizing personal frustrations and biases. As we marblers are known to do, I made a spreadsheet, and ran some calculations to see what came of it. I will give a more detailed explanation of my data analysis at a later date, but the TL;DR version is this: The spreadsheet does not give us all the answers, but what I found supports the claim that many hunt levels give far more rating than their respective difficulties normally suggest. It does not tell us what the new curves should be, but it does show which levels are the most inflated and how severe the problem is.

For now I want to lay out the principles I seek to use in adjusting these rating curves. Since the spreadsheet can only diagnose the problem and not solve it, I need principles for adjusting them fairly. This list is generally what I'm looking for:

• Hunt levels should still be worth playing.
• Top scores should be meaningfully rewarded.
• Innovation matters; the curves should not be flattened so much that improvement feels pointless.
• Rating should not far exceed the normal range of a level's difficulty.
• New rating curves should account for modern scores, not just old expectations.
• Nerfs should be proportional to how inflated each level is.
• There should not be a flat nerf applied to every hunt level.
• Community feedback will be most helpful when it engages with the data, the methods, the principles, and the practical feel of the proposed curves.

The goal is a limited adjustment, not a complete rework of the entire rating system. I want to bring certain hunt levels back into the expected range of their difficulties while still valuing the high level of skill and dedication put into achieving their scores. At this stage I would like feedback on these principles and this framework for approaching this project before I share my data and we start working on proposed adjustments for specific rating curves. If you feel anything is missing or needs adjusting, please respond to this post. I don't expect to do this all on my own as I once planned.