Riven Trading & Toolbuilders: Phase 1

[(PSN)Spider_Enigma]

what would solve this problem is if the riven had there average sell price (ignoring sales of 0-10 plat in the equation) listed on the riven itself, based on the stats on the riven and weapon in question as well 

[Zyga21]

is there any way to cull out "friend" trades? I have a friend that I play with all the time and I've given him rivens I didn't need for basically nothing, a redirection, or on another occasion I needed 20p and I gave him a riven for it, like it might be harder to cull out trades like that but for the sake of the integrity of the data it should be done. Just worried that these minimum trades or maximum trades that are very high/low are skewing the data a bit.

Then there's the whole issue of the stats on these mods.... you can't sell bad stats for 1000p. Goodluck, I'd rather see some sort of in game trading market.

Edited March 31, 2019 by Zyga21
added extra tidbit

[AVAQ]

21 minutes ago, Zyga21 said:

is there any way to cull out "friend" trades? I have a friend that I play with all the time and I've given him rivens I didn't need for basically nothing, a redirection, or on another occasion I needed 20p and I gave him a riven for it, like it might be harder to cull out trades like that but for the sake of the integrity of the data it should be done. Just worried that these minimum trades or maximum trades that are very high/low are skewing the data a bit.

Then there's the whole issue of the stats on these mods.... you can't sell bad stats for 1000p. Goodluck, I'd rather see some sort of in game trading market.

I was just thinking about this. This could also result in market manipulation, because a group of people could decide to sell a single riven in a circle for very low prices, which would lead to the overall price graphs falling.

[Soulless_Spartan]

12 hours ago, (XB1)Kuva Queen said:

It's because you cannot get those rivens easily. They were out for less than a week then taken out of the game so the chances of finding one are so so low. 

...I so very much want to say something snarky and smart-arsed about how I was already aware of that... seriously, what made you think I didn't know about that?

[---.CCC.---TradeFrame]

I'll try to be positive and ignore the fact that the "riven mafia" is going to be smug because we are back in the "stone age" when it comes to trading rivens.

DE currently is giving us data that COMPLETELY ignores riven stats for pricing.
It's merging trash/unrolled/ so called "godly rolls" numbers to give average / std. deviation prices. What's the point??
It's a lot of numbers but it's utter garbage.

The only thing this data was semi-useful was for checking data for veiled rivens.. and even that is only partially useful thanks to the extreme range of min-max, with no idea of actual quantity of trades.
Heck, give us median data as well, not just mean data.
Give us data, ACTUAL trade data so we can number crunch it ourselves PLUS how many were sold, to reach other conclusions instead of limiting us with semi-digested data.

DE, till you don't add attribute details for an item that's priced ONLY on what attributes it has, this phase is trash.
The only good thing about Phase 1 is, you've done a stress test of your delivery mechanism and the webpage to provide "some" statistics to your players.

Hopefully something more meaningful next time.

Till then, Riven Mafia, pewdiepie "congratulations".

Edited April 1, 2019 by IronHair

[Xolgys]

vor 11 Minuten schrieb IronHair:

I'll try to be positive and ignore the fact that the "riven mafia" is going to be smug because we are back in the "stone age" when it comes to trading rivens.

DE currently is giving us data that COMPLETELY ignores riven stats for pricing.
It's merging trash/unrolled/ so called "godly rolls" numbers to give average / std. deviation prices. What's the point??
It's a lot of numbers but it's utter garbage.

The only thing this data was semi-useful was for checking data for veiled rivens.. and even that is only partially useful thanks to the extreme range of min-max, with no idea of actual quantity of trades.
Heck, give us median data as well, not just mean data.
Give us data, ACTUAL trade data so we can number crunch it ourselves PLUS how many were sold, to reach other conclusions instead of limiting us with semi-digested data.

DE, till you don't add attribute details for an item that's priced ONLY on what attributes it has, this phase is trash.
The only good thing about Phase 1 is, you've done a stress test of your delivery mechanism and the webpage to provide "some" statistics to your players.

Hopefully something more meaningful next time.

Till then, Riven Mafia, pewdiepie "congratulations".

i agree but its also difficult though because stats are still opinion. stats could have a grade or rating tag based on average opinion to keep it flexible, like status chance for the kohm has S if its above 120% and so on. such ratings would also give newer players a hint to which stats are good, even if its something more specific like for some shotguns.

im also curious how many of these silly 10k+ rubico and whatnot get actually sold although thats more my personal curiosity than anything else.

[Malaki_7]

@IronHair it is not specific rivens but the type of riven.

There is the rolled or unrolled version.

[Girunugal]

I mean, selling expensive riven is cool. But, selling it even faster without waiting for days or week AND still get decent prices, is way cooler... 

[Neohazukibr2]

SUPRA for 25k PL?

[Alselein]

A link to this thread or the actual links for each platform should be displayed in-game for everyone to be aware of this, maybe put it between the information that you get when clicking on trade chat, or have another trade chat entirely dedicated to rivens. 

Also maximum prices I think should be limited by the developer, so that people dont even have the chance to ask for astronomical amounts of plat. Maybe the amount of plat you can pay for a riven should be capped at 1-2k.

[(XBOX)Kaancer]

On 2019-03-29 at 5:43 PM, [DE]Rebecca said:

Our first batch of live data is here! Read on, Riven masters!

You may recall:

Today I want to update you all on this part: 'When the data is set up we will post links and explain the update cadence / data, then toolmakers can either start or continue to provide Riven trade analysis for players.'

Right to the goods:

DATA LINKS:

PC: Updated weekly based on the prior week's trades: http://n9e5v4d8.ssl.hwcdn.net/repos/weeklyRivensPC.json
PS4: Updated weekly based on the prior week's trades: http://n9e5v4d8.ssl.hwcdn.net/repos/weeklyRivensPS4.json
XB1: Updated weekly based on the prior week's trades: http://n9e5v4d8.ssl.hwcdn.net/repos/weeklyRivensXB1.json
SWITCH: Updated weekly based on the prior week's trades: http://n9e5v4d8.ssl.hwcdn.net/repos/weeklyRivensSWI.json

SIMPLE EXPLANATION:
There are 8 fields to start things off in Phase 1:

itemType:  What type of Riven is it?
compatibility: Which Weapon is the Riven for?
rerolled: Has it been rolled or unrolled (true/false)?
avg: Average value of trades with this Riven Type.
stddev: The average variation in the prices that the Riven trades for.
min: Lowest amount of Platinum per trade of this Riven Type.
max: Highest amount of Platinum per trade of this Riven Type.
pop: Popularity of this Riven Type being traded within all traded Rivens.

WHAT YOU NEED TO KNOW:

This is actual trade data, not trade-chat data, released with the purposes to help toolbuilders and players learn more about market value of their Riven Mods. It is planned to auto-repopulate at 00:00 UTC Monday (the first week might require manual push as we get set up here). Of course, value is always open to interpretation. We are experimenting with this whole idea in Phase 1 - we don't really know what to expect, as is often the case. While it doesn't matter what we expect, what matters is why we are doing this: to remove any foul play in the Market. After a couple of weeks of feedback and results, we plan on bringing a bit more data, so please be constructive in replies.

All of this data is made up of Riven + Platinum trades. We do not include '1 Platinum' trades in our averages, etc, because we consider these to be between friends and not reflective of Market reality (although, friends, why not choose Ammo Drum)?

Riven Trading is indeed an optional part of the Warframe experience, we hope that those who participate in it find this helpful! Thank you to all who helped push this idea along, you know who you are <3!
 

 

Neat. The Corpus role-playing got a bit too real.

[Im_a_fat_boy2]

alternatively, DE can overhaul the entire riven system and make it rng independent

this should send the market crashing to affordable levels for all players 🙂

 

Edited April 2, 2019 by Im_a_fat_boy2

[NerfFrame]

its crazy that there is still no in game market place with this kind of info but this is a very good first step by DE 

[Obi-Juan.Kenobi]

The problem has been pretty widely stated at this point, being the massive disparity between most trade prices and the wildly bloated ones.

The easiest way to deal with this would be to eliminate outliers (maybe by the 1.5*IQR method?) and recalculating statistics with those adjusted ones, once outliers are eliminated. Because, let's be honest, people aren't regularly going to pay 4k platinum for an unveiled pistol riven.

[(PSN)ForNoPurpose]

Th is is kinda cool.

[Kontrollo]

5 hours ago, Obi-Juan.Kenobi said:

The problem has been pretty widely stated at this point, being the massive disparity between most trade prices and the wildly bloated ones.

The easiest way to deal with this would be to eliminate outliers (maybe by the 1.5*IQR method?) and recalculating statistics with those adjusted ones, once outliers are eliminated. Because, let's be honest, people aren't regularly going to pay 4k platinum for an unveiled pistol riven.

It's pretty clear there are some funky things going on there:

With a mean (average; avg) & standard deviation (stdddev) given, the assumption is we're dealing with a normal distribution here (bell curve). But with a hard cap of 1 plat at the lower end, you can be sure the standard deviation should not be significantly higher than about 1/3 to 1/2 of the mean, because 2x the standard deviation from the mean includes 95%, and 3x 99.7% of all data:

Spoiler

The symbol there is the Greek letter sigma (σ) and stands for the standard deviation -- in each direction from the mean, denoted with 0 here.

Ideally we'd see values where the '1 platinum' falls even below 3 sigma, because zero or negative trades aren't even possible (and we know the min value anyway). Assuming 1/3 to 1/2 (-3σ to -2σ in the graph) is reasonable, however, because you can't expect a perfect bell shape with this kind of data.

=> It's a simple indicator for how well the idealised normal distribution fits the actual data.

But then we have values like this (first week):

• Supra (rerolled):
  • avg: 567.06
  • stddev: 2700.21
• Veiled Pistol Riven:
  • avg: 31.76
  • stddev: 167.96

With values like that it's obvious we're not dealing with normal distributions on these Riven classes. Here, we should at least know the median, because the mean is much more susceptible to outliers.

 

And as you say, 4k platinum for a Veiled Pistol Riven (in the first week data above it's 3k) -- no one in their right mind is paying that, and I hope the devs are investigating at least some of the more "interesting" trades. (Week two data for Veiled Pistol Rivens: avg: 23.06 / stddev: 118.66 / min: 2 / max: 4000)

Edited April 2, 2019 by Kontrollo
corrections -- it's late & more explanations

[Obi-Juan.Kenobi]

I imagine the main reason we didn't get a median at this point was due to complexity of the algorithm needed to compute it. As it is, they're at O(n) for the complexity to find everything currently listed, but finding median(s) would require the data to be sorted, which would make this a minimum of O(nlogn) for complexity. With the presumably massive data sets involved here, that logarithmic increase makes this take substantially longer, which I imagine they wouldn't want to do.

[C486]

Never underestimate the power of greed.

 

 

Spoiler

Sipping some tea during this post, so meh.

 

[Kontrollo]

4 hours ago, Obi-Juan.Kenobi said:

I imagine the main reason we didn't get a median at this point was due to complexity of the algorithm needed to compute it. As it is, they're at O(n) for the complexity to find everything currently listed, but finding median(s) would require the data to be sorted, which would make this a minimum of O(nlogn) for complexity. With the presumably massive data sets involved here, that logarithmic increase makes this take substantially longer, which I imagine they wouldn't want to do.

Hm, I doubt they're that massive.

Let's say PC has about 120k individual players per day (i.e. 840k per week). Even if all players would trade 1 Riven on average per day, computing the median of 1 million trades is hardly resource intensive (takes far less than 1s). And the data is split into 546 categories, to boot.

 

Edit: also, a quick search told me it's possible to do it in O(n) anyway.

Edited April 3, 2019 by Kontrollo

[RocketSniper]

I would like to see this *in-game* somewhere.  New players are still going to get duped because they mostly don't go to the forums to read excessively about the game, so without some sort of prompting in-game, I wouldn't expect this to change much.  Maybe on the rivens page, when a riven is selected, a 'view trade data' button could pop up, that's populated with the data from the json file?

I would also like to have the date range included in the json file somewhere, so it's easier to compile / chart week over week.

I really like the transparency here. THANKS DE!

Edited April 3, 2019 by RocketSniper
included date range request

[(XBOX)GrandSageDragon]

Now they just need to buff tatsu's disposition with melee 3.0 and my riven for it might not be so trashy 

[Kontrollo]

I've only had a quick glance at the new data. Some thoughts/observations:

• Median field has been added, thanks for that. This can already give us a much better idea with this kind of data.
• Feels like you were using some JSON-producing library before, now you're doing it manually.
  • Might be because JSON doesn't support NaN values? Luckily my (Python's) library accepts that by default. Others don't, however.
• I have the impression that veiled Riven data hasn't been updated here. Except for the added median field and rounded floats, the data seems to be the same (are the prices this stable?). Example Kitgun, but all look like this:

Spoiler

Week 2:
   {
        "itemType": "Kitgun Riven Mod",
        "compatibility": null,
        "rerolled": false,
        "avg": 31.870082342177,
        "stddev": 28.855534284122,
        "min": 2.0,
        "max": 800.0,
        "pop": 3.008
    },

Week 3:
    {
        "itemType" : "Kitgun Riven Mod",
        "compatibility" : null,
        "rerolled" : false,
        "avg" : 31.87,
        "stddev" : 28.86,
        "min" : 2,
        "max" : 800,
        "pop" : 3.008,
        "median" : 30
    },

• There are some more oddities, but I'm going to look into it at some later point.

 

P.S.: If anyone was brave enough to use the script I posted earlier, this needs to be adjusted for the new field, too. However, the forum is not a good place for these scripts (forum software messes it up somehow), so I'm not going to update/repost here.

[TenGaugeBoson]

Nice .To my own credit I asked for something like this 2 years ago cause I am so awesome.

Seriously though.Thanks for this.Hopefully some one can put it in a spreadsheet or something easy to read.

DE and Warframe. literally the only light in an industry that has fallen from grace.

[Kontrollo]

Ignore my previous post. The first time I pulled the data about 1 hour after midnight UTC, and just did it again manually.

Looks like we now have new data (week 3), and the algorithm which has resulted in NaN values has since been fixed.

 

I still haven't had a close look at everything, but can you clarify what the goals and non-goals here are? Because it's very easy to infer some data that maybe isn't meant to be clear. It's easier with lower total number, so examples from the NSW data:

• There are 134 categories with only 1 trade (min and max identical, min. pop value)
• There are 63 categories with only 2 trades (2x min. pop value)
• There are 38 categories with only 3 trades (3x min. pop value)
• We know the exact value of every trade in these, i.e. the categories in the previous 3 bullet points (using the available data)
• The highest is Rifle Riven Mod with 195 trades (using pop value and because we know the min. pop value)
• Total number of trades is 1261 (again, using min. pop value)

I haven't used any statistical methods on anything yet, these values can simply be inferred. (My statistics knowledge isn't that great, anyway.)

 

Maybe you'll have to increase the timespan on some categories? If you eventually decide to do that, please include timespan data. Single values or aggregates with low numbers aren't that interesting anyway, because we have no information on the Riven stats, so these are most likely just misleading anyway.

 

 

Edit: For those who just want to see some useful data. Median values are definitely interesting, and veiled Rivens don't have any variable stats. I hope I've copied these without any mistakes:

 

Week 3 Median values of Veiled Rivens

	Rifle	Shotgun	Pistol	Kitgun	Melee	Zaw
PC	60	50	15	30	30	25
PS4	20	30	13	15	20	15
XB1	30	40	20	22.5	23	20
Switch	50	58	20	25	30	30

Edited April 8, 2019 by Kontrollo

[(NSW)EdgeWalker]

5 hours ago, Kontrollo said:

Ignore my previous post. The first time I pulled the data about 1 hour after midnight UTC, and just did it again manually.

So the midnight UTC still isn't set, I had something waiting 10 min to pull the data and after looking at this week the data appears the same so was going to push the pull time back. Looks like an hour or more push back is in order for now.