Live data from Hacker News

95%-ile isn't that good (2020)

danluu.com

141–150 of 191 posts

Re: 95%-ile isn't that good (2020)

#141

Are there any good platforms out there that people would recommend where you can have your code reviewed? Either paid for or for free. As this article points out, feedback is an essential part of learning, and open source projects would be a great opportunity to get feedback on something you have sole agency over as opposed to many workplace projects where it's harder to receive feedback for your individual decisions…

If the code is fairly short, there is https://codereview.stackexchange.com/

Re: 95%-ile isn't that good (2020)

#142

When I was younger, I used to enjoy competition and climbing leaderboards. Now, parenting always comes first. Reaching the top 5% is possible for most people with sustained effort. Many participants aren't dedicated to reaching the top of the leaderboard.

I aim to be a 95th percentile parent.

It's not as hard as you'd think

Re: 95%-ile isn't that good (2020)

#143

Are there any good platforms out there that people would recommend where you can have your code reviewed? Either paid for or for free. As this article points out, feedback is an essential part of learning, and open source projects would be a great opportunity to get feedback on something you have sole agency over as opposed to many workplace projects where it's harder to receive feedback for your individual decisions…

Send PRs to open source projects you’re interested in?

Open source projects are flooded by low-quality PRs, they can't afford to give mentor-quality feedback to them. Better to join the project's chat server and have back-and-forth discussions.

Re: 95%-ile isn't that good (2020)

#144

The argument seems like a cheat, because it rules out people who don't participate, but doesn't limit itself to people who practice regularly. For video games, an analogy could be limiting to people who bought the game on steam, but not limiting to people who ever installed and played it. Or maybe a better analogy is website users, limiting to people who visited, but still including all the people who immediately bou…

You can weight by frequency of activity—say, approximately "players sampled from matches selected uniformly at random"—and it'd largely still be true. The top few percentiles do practice more often, but overwhelmingly so.

Glicko metrics for two-player games basically have this property, so if you want to try it out in the wild, you can pick up a game with a Glicko-rated ladder and see how hard it is to get your mean rating estimate up +2 SDs.

Re: 95%-ile isn't that good (2020)

#145
Great guide to get to top 95-99% of a video game. Completely different guide needed to get from 99% to 99.99%. Would love to read that one, as someone who's ranked top 1% in a video game (rocket league).

Re: 95%-ile isn't that good (2020)

#146
post #47

Earlier quoted context omitted.

> A streamer (...) says that if you are (...) watching streams of the game, that makes you a top 10% player automatically. Sounds like pandering to the audience. Performing better then 90% of people makes you a top 10% player. Self soothing about how what you like to do in your spare time makes you better then others is not making you any favors. It doesn't matter how much effort you put into being effective. In fact…

I got to top 10% of Trackmania players (extremely competitive) just by watching Wirtual’s streams. Couple of advices (and learning to implement them) were worth more than hours of (bad) grinding. And he is „only” in like 99,9th percentile in time trials, not even competing in top tournaments.

Same: when Path of Exile was less popular and had races I was often able to finish top 10% (or even much better) by just copying techniques from streamers who were good.

The concentration of knowledge that was available in a (successful player's) stream was far more useful than what I could gain from personal experience playing alone.

Re: 95%-ile isn't that good (2020)

#147
post #121

Earlier quoted context omitted.

> claims a 352 lb bench is ~95th percentile for a 180 lb male Surely you mean 353lb dead lift.

That's kind of the point: people online will say that's the 95 %ile level[0], and either they're taking the piss or they're self-selecting into a group that actually tries, so it's the 95 %ile of the 95%ile, or the 99.75 %ile. Meanwhile your average gymgoer that sometimes wanders into the free weight room thinks putting the big plates on the bar (i.e. 135) is a lot. There's obviously no way 1 in 20 dudes at the gym h…

And not to distract from the point (which you made well), but weight lifting in general is tricky to compare since you generally want to take body weight into account. Benching 225 lbs is a different prospect for a 125 lbs individual vs. a 250 lbs individual.

Re: 95%-ile isn't that good (2020)

#148

Earlier quoted context omitted.

last 5%? I've been top 1% and slightly below 1% at peak in video games that have milions of gamers and I wouldnt say that top 5% is filled with "truly talented people" Top 5% is just decent people, but they are very, very, very far from top 0.1, 0.5%, etc. At 1% I've been faaaaar from 0.5%, 0.1% people.

I'm no good anymore, but in the VERY early days of SC2, I was good enough to occasionally get matched up by the ranking system with players like WhiteRa and Idra. Shocking how much better they were than me. Technically in the same league, but in terms of skill it wasn't even close. The top .1% really are something special in these types of cases. Pretty neat stuff.

I just looked and the top 4% of the SC2 ladder is Master+, which I consider to be quite an accomplishment. At the same time, there's so much range of skill within the master's leagues as well as even in the Grandmaster (top 200 players in each region). Even the difference between the top 5 pros and next 10 are visible in matches.

This reminds me of the range of skills in Go (board game), from absolute beginner 30-20k (kyu) beginners, 19k-10k casual, 9k-1k amateur (me), then 1-7d advanced amateur (dan), then 1-9p (pro dan). Each few 'stones' of difference can typically squash or run circles around the other. So many orders of magnitude.

Re: 95%-ile isn't that good (2020)

#149
post #66

I don't think I agree with this at all. I can see how you would have this impression with something like a video game or I guess public speaking where maybe a lot of people do it un-seriously and you can be better than most with some intentional effort. It's weird to choose a super mass market game like Overwatch to make that point though, if you were in the 95th percentile of Starcraft players, that would truly be s…

I'm not sure what the 352 is supposed to be. strengthlevel.com claims a 352 lb bench is ~95th percentile for a 180 lb male, but I've never seen anyone in a gym that can bench anywhere near that. IME even a 225 lb squat feels like 95th percentile. I don't think I've ever seen someone else do a 100 lb shoulder press (~20th percentile on SL). Usually I see something like 75 lb (doing fixed barbells), and once I saw 95 l…

This thread is truly bizarre.

A 352 pound bench at 180 pounds is bordering on being an elite lift. I would wager that it's unattainable for most guys, even if they worked at it and did everything right (in terms of training, nutrition, etc). Success in athletics/strength-based activities is, in large part, determined by genetics.

I'm sure there are domains where being in the 95th percentile just takes "some work." If we're using video games to make this point though, then I'm at a loss for words.

Re: 95%-ile isn't that good (2020)

#150

I would have to agree about 95th percentile being not that difficult. A streamer that I occasionally tune into (Asmongold) says that if you are seeking information outside of the game such as watching streams of the game, that makes you a top 10% player automatically. Most people are passive consumers and participants in their activities and that isn't exclusive to the individual. I am a passive consumer & participan…

In StarCraft the reverse it true. There are many who get information outside the game by watching, but don't actually play that much (1v1 in particular). Just by playing it puts you in a high percentile of SC2 players (that used to play or only play co-op and other forms).
Post reply on HN