Analysis of Sticker Affect on Price -- a Look at Over One Million Market Listings in Counter-Strike: Global Offensive

April 30, 2016

Abstract

Over 1,100,000 Steam community market listings with at least one sticker have been recorded since Jan 31, 2016 (up to the time of this analysis April 29, 2016). A baseline minimum price is established for any given item and the difference between the baseline price and the price of an item with stickers is recorded. The presence of stickers is then correlated to a difference in market price from the baseline.

If you just want the raw data summary, check the appendices at the bottom.

Note: I apologize for the unusual names, these are not standardized to what you see on the market, but instead are ripped straight from steamapps\common\Counter-Strike Global Offensive\csgo\scripts\items\items_game.txt. If so inclined, you will be able to match the id from there to the stickers in the data.

Introduction

In late 2015, it became widely known that detailed information could be obtained from items listed on the community market. Since then, a number of services have become available to obtain detailed market listing information (csgozone, glws, etc). An (anonymous) owner of one of these services has graciously allowed me to perform analysis on market data collected, totalling over twenty million CS:GO community market listings since Jan 31, 2016. Of these, over one million listings had detailed item information available and had at least one sticker -- note that not all market listings recorded have detailed item information.

A dataset of this size should allow reasonable conclusions to be drawn, and the goal of this post is to show the relative value any given sticker adds to a market listing.

Methodology

A market listing only contains basic information about an item, most importantly the price. Retrieving detailed information from a listing will show additional details such as the wear value, paint seed, and any stickers the item has. Internally, stickers are stored in sequential order, from the first to the fourth slot. An item with one sticker has a sticker in the (internal) first slot regardless of where the sticker appears on the item; and an item with two stickers will always have a sticker in the (internal) first and second slot, and so on. This allows simple data correlations to be determined based solely on the first sticker. More accurate conclusions could be drawn by performing some type of mulitple linear regression.

As all price information is logged, a baseline price is established for each item by finding the minimum price seen on the market over the last seven days (seven day period ending as of the time of this analysis April 29, 2016). The difference between a listing price and the baseline price is recorded as the price delta. This implicitly normalizes all price information, and the underlying item can be ignored. This has a small (unmeasured) affect of biasing more liquid/higher volume items to a "true" price, while lower volume items will display a higher amount of variance. This may or may not affect the conclusions drawn here.

Limitations

The set of market listings was filtered to only include items with at least one sticker. To state the obvious, only market listings observed at least once where included in the data set. That is, there were a number of unobserved market listings for items listed and sold between market gathering intervals, but these are generally items listed at and immediately sold for market price. There were several technical malfuctions, and at times Steam was unreachable. There was at least one six-hour period when the infrastructure to log market data was down. In this particular case, the ability to gather detailed item information from a market listing has evolved over time and the amount of data becomes more sparse closer to the beginning of the logging period (Jan 31, 2016). These are the various factors affecting the ability to gather a true picture of the Steam market.

This post is not attempting to prescribe an answer to the question "how much value does this sticker add?" In order to do that, the amount that any listing sold for would have to be known, as there are surely a large number of listings "overpriced" on the market due to the stickers included and have not yet been sold. Instead, this report evaluates the price of items on the market and describes how certain stickers affect the price an item is listed for.

Another factor that may influence the conclusions drawn here are recent events. That is, the most recent major will probably have the most stickers in circulation, as older stickers will be on items in "retired" accounts; this is an untested assumption and may or may not have significance.

Analysis

The sticker seen the most frequently (in sticker slot 1) is by far the generic ESL Cologne 2015 sticker (seen 18095 times)[Appendix A]. It appears twice as often as the sixth most frequently seen sticker (ESL Cologne 2015 Team EnVyUs seen 9930 times), and almost three times as often as the 20th most frequent sticker (Headshot Guarantee seen 6379 times). Of the twenty most frequently seen stickers, seventeen (85%) are from tournaments. Four of the top twenty are simply generic major stickers; eight are for teams; five are for individual players, all of which are either current or former EnVyUs members. Of the seventeen tournament stickers, sixteen are for events that occurred in 2015, with the only exception being the 2016 Columbus major at #4 most seen.

If tournament stickers are ignored, then the most frequently seen stickers are "Drug War Veteran (seen 9061 times)", "Thug Life (seen 7175 times)," and "Headshot Guarantee (seen 6379 times)." Non-tournament stickers are much more common than any individual tournament sticker, including stickers expected to be relatively rare such as Howling Dawn, which comes in at position 313 (seen 1026 times). The three most infrequently seen non-tournament stickers are "Bombsquad (foil)" (seen 57 times), "Trick or Treat" (seen 57 times), and "Witch" (seen 61) times.

Unsurprisingly, the sticker that has the most affect on the price of an item is the iBUYPOWER Katowice 2014 (holo), which increases the price of an item on average by $204.73 [Appendix B]. In fact, all of the largest price increases come from Katowice 2014 stickers with three exceptions: 1) at position number 3, the iBUYPOWER DreamHack 2014 (gold) sticker with an average price increase of $96.96, 2) at position number 4, the LGB Katowice 2015 (foil) with an average price increase of $92.98, and 3) at position number 7, the jasonr (gold) signature foil; this sticker was only seen seven times, five of which where between $30 and $170 above the current market price and occurred around the time of the Columbus major, while the remaining two times occurred within the past week and were $0.03 and $0.20 above market price. Tournament stickers make up the bulk of known stickers, and of the 160 stickers which cause the average price of an item to increase by $10 or more, 153 of those are tournament stickers.

If non-tournament stickers are considered, then the sticker which affects price the most is the Howling Dawn sticker, with an average premium of $44.62, followed by the "Crown (foil)" at $35.15. This falls off sharply, such that the seventh sticker, the "Fire starter (holo)" is the last sticker to raise the price of an item by more than $10 above market price.

An aggregate of the average price delta was calculated per collection, and shown in the table below, in USD cents.

+--------------------+-------+
| Collection         | Delta |
+--------------------+-------+
| tournament_assets  |  2961 |
| emskatowice2014    |  2914 |
| dhw2014            |   521 |
| community01        |   490 |
| columbus2016       |   480 |
| dreamhack          |   468 |
| stickers2          |   366 |
| eslkatowice2015    |   362 |
| cologne2014        |   289 |
| community02        |   277 |
| pinups_capsule     |   248 |
| enfu_capsule       |   245 |
| standard           |   211 |
| slid3_capsule      |   205 |
| cologne2015        |   203 |
| team_roles_capsule |   202 |
| cluj2015           |   180 |
+--------------------+-------+

note: The "tournament_assets" collection seems to only contain the "allstars a" and "allstars b" holo from the recent Columbus 2016 major, and only seems to be found on souvenir items. example: 20160430csgo-image-1.jpg.

Conclusion

The average of any price change, due to having at least one sticker (tournament or not), is $5.71. There is strong evidence that it is generally believed stickers add value to an item. Of all stickers, only four (in the first sticker slot) have reduced average prices over the time period measured. Of the 1285 stickers available, 1141 are on items listed for a premium of more than $1 above the minimum price; 711 are on items listed for a premium of more than $2.50 above the minimum price; 355 are on items listed for a premium of more than $5 above the minimum price; and 160 are on items listed for a premium of more than $10 above the minimum price. Whether or not these items listed on the market sell for a premium due to stickers can not be decided from this data, but the available data supports the belief that stickers add value.

Appendix A -- CS:GO Sticker Count https://docs.google.com/spreadsheets/d/1VofYrhXUgcom3CEBKxvXImxCj91Q94aSb1PMC9kfRKo/edit?usp=sharing

Appendix B -- CS:GO Sticker Price Delta https://docs.google.com/spreadsheets/d/1ns_y_NP5Bno-PUBMgpIUWvAgxBWcp1DmOtSnXcZJadY/edit?usp=sharing