Tom94 wrote:
Would you mind telling us how the system detects easier maps if not through contention / map heuristics?
peppy wrote:
Think stuff like sparsity of scores relative to highest SS, ratio of SS to S to poorer ranks, pass ratios.
Also consider that it isn't all about finding the most insanely hard maps (though this would be optimal); higher played maps should inherently be considered more because they are more prominent in the community. It doesn't matter so much if they are "only" hard, because if you are a good player you will also be able to do well in these maps.
Sparsity of scores cannot be used to evaluate difficulty or anything
at all because of the simple reason that not everybody plays every map. It was already mentioned multiple times and yet it seems like you choose to disregard this.
Lybydose wrote:
the "Hard" difficulties are more worthwhile to play for a number of reasons:
1. Large number of scores from "average" people playing that difficulty because they can't pass Insane maps. Thus, high "contention".
2. Very few "pro" players bother to even play the Hard, so it's very easy to rank top 10 or better with only HD or HD/HR or DT at < OD7.
3. It's very difficult to get top 10 on Insane maps AND get high accuracy, because this usually requires playing HR or DT on something that's OD8 or higher.
You said you've taken Lybydose's post into account, but when saying things like "because if you are a good player you will also be able to do well in these maps" it seems like you don't fully comprehend what was being said. There are maps that almost nobody can "do well on", a few examples:
http://osu.ppy.sh/s/29691 - 323117 plays, 12 FCs
http://osu.ppy.sh/s/41686 - 240405 plays, 7 FCs
http://osu.ppy.sh/s/34097 - 190070 plays, 15 FCs
http://osu.ppy.sh/s/32661 - 62063 plays, 10 FCs
http://osu.ppy.sh/s/48979 - 44550 plays, 9 FCs
http://osu.ppy.sh/s/37292 - 49767 plays, 12 FCs
http://osu.ppy.sh/s/45528 - 37345 plays, 6 FCs
http://osu.ppy.sh/s/34348 - 122467 plays, only about 30 plays with over 90% accuracy
http://osu.ppy.sh/b/142772&m=0 - 43564 plays, very few people with over 95% accuracy
http://osu.ppy.sh/s/27712 - 155553 plays, 11 FCs
http://osu.ppy.sh/b/138008&m=0 - 44496 plays, 23 FCs
http://osu.ppy.sh/b/152199&m=0 - 56798 plays, 21 FCs
http://osu.ppy.sh/b/127047&m=0 - 109244 plays, 19 FCs
http://osu.ppy.sh/b/119021&m=0 - 87665 plays, 24 FCs
All of those are popular maps and yet nobody can seem to FC them (except for a few top players, who had to work very hard for it) or get high accuracy. With a system that gives extra point for FC and SS or even S you're basically telling people "don't bother playing maps that are actually hard, you won't get rewarded for those at all. Instead go ahead and play many easy maps with a popular song where a lot of new players had bad plays".
sparsity of scores relative to highest SS... using sparsity of scores makes no sense at all if you ignore who those scores are coming from and the amount of retries. Easy maps (and by this I mean, all of the [Hard]s and all of the low-bpm and low-od [Insane]s) are going to have way more plays on them, simply because
more people are going to try playing them at all (compared to actually-hard maps), and yet those more-plays will usually be lower quality plays for the same reason - that they're being played by newer/inexperienced players. The end result is that those easy maps reward a ton of pp because pp thinks that if a lot of people tried a map and did poorly on it it must mean that map is hard; but no, it usually just means that map is easy and thus more accessible for a bigger playerbase.
Also, since the system doesn't take into account amount of retries taken to get a score, low OD maps are waaaaaaay easier to get pp on than high OD ones, because if you just grind a low-od map enough times you will be able to get good accuracy on it for free, meanwhile on a high OD map you won't get good accuracy no matter how much you retry until you actually get good at being accurate.
There's also the fact that high end players would usually not want to play the easier maps at all because they're boring to them, which means that getting a
high-rank on those is again waaaaay easier because even if the amount of plays is really high the true contention you're dealing with on those maps is very low - by this I mean, there's less players you actually have to pass for that high rank because there's less players who actually play good enough to get that top rank who care to actually play that map and get that top rank on it. If you told WW, Niko, Silvia, etc, to play a bunch of [Hard] maps and try to aim for a top rank on those they'd be able to get it very easily and that would knock out a ton of people who previously had top ranks on that [Hard], but since the top players are not going to waste their time on playing maps that don't challenge them for a rank it means that other mid-range players can get a top rank on those maps for free.
peppy wrote:
The cases you mention where Hard is worth "more" than insane (I checked the provided ones) isn't actually a huge difference
Just don't know how you can say this. There
is a huge difference and the examples people have given shows it clearly. All of those combined:
1. PP not taking into account amount of retries
2. Top players not bothering to play every single low-difficulty
3. Most players choosing to play low-difficulties over the higher ones, even if they do poorly them, because they'd do even poorer on the higher ones
Means that if you want to get high PP, you need to play low difficulties where the superficial contention is going to be high due to amount of plays from players who can't handle even that difficulty, and you can just grind those difficulties again and again until you get a good score on them, because even if a better player tries that difficulty he might only try it once, possibly get a random 1x100 or even a miss due to not caring, and not really care to fix that because that map isn't fun for him, and then somebody else who grinds that map for 50-100 times would get a higher score/rank but that doesn't mean he can play better.
Meanwhile on maps where top players are actually going to care to retry a lot, getting good acc or anything close to a top rank will be near impossible, and that means those maps are essentially pointless to play if you want to be ranked unless you're already a top-30 player who can contend with all the other top-30 players who have many retries on those maps.
If you actually want
more examples for this, you don't even have to search hard because it's
everywhere, just open any top-50 or hell any player at all profile and look at their top plays, there will usually be [Hard] maps there that were probably easy as hell for that player to get a top rank on, and the only reason he bothered to do it is usually because he liked that specific song or whatever; meanwhile a bunch of A/B ranks from actual hard plays that player is capable of doing on maps like what I listed in the earlier part of this post give that player nothing, even if he had to work much harder for those and they take a lot more skill.
;;
To sum this all up: the premise of pp was that it would let you just play whatever you want normally and then rank you correctly based on that.
In reality, pp only ranks correctly (and even that not very) the highest top players who all contend on the same few hardest maps. For everybody else, to get PP you basically need to find maps that are not hard enough to interest the top players, and yet have a popular song that attracts a lot of the general mass to play them, and then grind those maps until you get a better score on them than everybody else who cared enough to grind them.
To put it bluntly, if you want PP you need to farm it. This is just a farmable stat that requires you to farm a few select maps instead of farming many different ones as was with scores. That's it.
---
PP should completely ignore the total amount of plays on a map. This stat is more related to how popular a song is and has nothing to do with contention. If you really want to know how contended a top rank on a map is, you need to check how many top contenders had a lot of retries on it. One map can have 100,000 plays and only 50 of those come from players with the ability to S that map, while another map can have 30,000 plays but 20000 of them would come from much better players than the previous case; the latter should reward you way more, but in practice it's the opposite.
A map like
http://osu.ppy.sh/s/48979 is very popular amongst top players, but even getting 80% accuracy or passing it at all is hard for the large majority of the players, and most inexperienced players wouldn't dare to play this at all - in reality it's very contended, and yet PP is just going to ignore most of people's plays on it (due to low accuracy and falling much lower in the ranks under all of the top players' scores on this). Meanwhile there's a ton of old maps with 100k+ plays on them but very few plays from top players (and even if there's a score from a top player, it's probably just an FC he did while setting a minor goal for himself, and not his actual best potential for a play on that map if he cared to practice to rank on it) that give a ton of PP despite realistically much lower contention.