From the blockchain transparency dot org website with links removed because this forum won’t let me post links
AUGUST 2018 - INITIAL EXCHANGE RANKINGS REPORT
Current metrics used to determine exchange rankings by CoinMarketCap even with the new ‘Adjusted Volume’ parameters have still left most in the industry questioning their methodology. A new system was developed by BTI to provide better transparency for the crypto sphere on true exchange volumes. In the following report we have discovered at least 7 of the top 10 exchanges engaging in excessive wash trading from 12x to over 100x their true volume. We have also highligted more trustworthy exchanges which move up 40-60 places in the true rankings list.
To determine a more accurate exchange rankings list, data for this new system was compiled using available order book liquidity and unique daily visitor counts for the various exchanges. This data looks to provide a more accurate estimation of true daily volume in the crypto field. The formulas used will be in a constant state of review and evolution as new metrics to estimate volume become available. As a reminder, these are estimates and as we receive more data from exchanges our September rankings will really hone in on an even more accurate count. Any exchange is welcome to contact us and provide data for a more accurate ranking in September. Transparency is key.
Research was conducted on numerous web traffic data websites and SimilarWeb was settled upon as the most accurate source of the technology available. Out of 1,700 websites tested against google analytics, sites with 100k-500k monthly visitors had an average accuracy difference of 23% (3,333-16,667 daily unique visitors), and with over 1 million monthly visitors around 11% (33,333+ daily unique visitors). Sites with under 100k monthly visitors could be overestimated by an average of 94% (Less than 3,333 unique visitors/day).
The formula used to determine the rankings begins with taking their unique visitors per day figure and multiplying it by the percentage of unique access sessions which were entered directly in the address bar by the user and not accessed via a referral site or social media link. This will better hone in on the retail traders accessing the site directly and not count users stumbling on to the sites via searches, online marketing, or blogs. The total volume reported by the exchange is then divided by this number to determine a per visitor volume figure. This amount will be lower than actual for each exchange since it is also including visitors who are checking prices and not actually trading.
Using [slippage research conducted by Sylvain Ribes] on exchanges faking volume by evidence of analyzing order book liquidity, a base set of exchanges that appear to be doing accurate volume reporting was established. This research accurately correlated with the largest number of unique visitors reported by SimilarWeb.
The exchanges with the most unique visitors per month likely have the most customers which would mean more daily traders and higher volumes. Exchanges with a 1.0 adjusted volume differential appear to be reporting accurately. This does not mean there is no wash trading going on, just that no evidence of wash trading could be seen using this model. However, more data sets will be used in future release rankings and these numbers could change.
A few differences can be seen where a few exchanges such as Bittrex, Kucoin, Cryptopia, and Bithumb have much higher unique visitor counts than their volume would expect to show. This gives further evidence of their accuracy as they could report much higher volume. Instead they appear to be accurately reporting that their users spend less per session. These exchanges trade many low cap coins and with most of the big volume spenders on high market cap coins these numbers fall in line with the accuracy of their reporting. Cryptopia, well known for having low liquidity on many of its coins naturally has one of the lowest dollar amount traded per visitor per day inside the top 50.
With a base set of 34 accurate reporting websites (determined by Ribes research and per user volume/unique visitor ratios) data was abstracted and applied to exchanges with suspect volume. Typical per user volume numbers for an exchange with x number of unique daily visitors was applied to all suspect exchanges to determine a more accurate per user volume figure per 24h.
The accurate exchanges outside of the big money exchanges typically have a volume/user to unique visitor ratio of around between 2% and 5% (3.5% average). The suspect exchanges ratio ranged wildly from 10% up to over 655,000%. For example, LBank and ZB exchanges which both claim to be in the top 10 of all exchanges, are also claiming to have volume/unique visitor numbers over $214,000 and $74,000 per day, respectively. This is outlandish considering known high liquid markets Bitfinex, Binance, and Coinbase fall between $5,000 and $8,500 per visitor per day.
Suspect exchanges with extremely high volume/visitor numbers were re-calculated based on the average vol/visitor ratio of known accurate reporting exchanges which averaged out to around 3.5%. This number could then be multiplied by the total unique visitor figure to determine a new estimated volume per day number. It was also noted that the suspect exchanges with a new volume number much lower than their reported figures all have very low social media activity as well as multiple reports of wash trading posted online by community traders.
The largest offenders in the top 10 include ZB exchange which appears to be wash trading their volume to over 390x as well as Lbank to over a whopping 4400x.
Other suspect exchanges in the top 10 include Huobi and Okex (Ribes data estimated them to be faking 80-90% of their total volume bringing their actual totals close to $100,000,000/24h) which also falls in line with this formula as they appear to be over reporting their volume by 12-19x coming in at $93M and $87M respectively. A ratio of 21% was used to determine their volume per day figure by averaging this number from the top 15 confirmed accurate exchanges with similar unique visitor counts.
Another top 10 exchange Bibox appears to be overstating their true volume by 85x. All of these top 10 suspect exchange numbers may sway lower or higher by an average of 23% (according to data measuring the accuracy of SimilarWeb’s traffic estimation tool by a third party).
Binance with by far the largest unique visitor per day count easily tops the new exchange rankings list. Bitfinex comes in at number 2 with a high volume per user figure as this exchange is known to be a landing spot for high dollar accounts with its $10,000 sign up minimum.
The top exchange in the USA is Coinbase which is no surprise, however Bithumb overtakes Upbit as the largest exchange in South Korea. Upbit was under investigation for falsifying book trading and money laundering, however a large accounting firm in S. Korea recently cleared them during an audit of their books. Upbit has contacted us since publication about their mobile user app count and we will be looking further into this in the coming weeks. Their ranking will also be adjusted after analyzing any new data we can confirm. We will be doing the same with any other exchanges that are open to working us and can provide verifiable data by our team.
Kucoin and Cryptopia are the biggest movers up the charts landing in at #19 and #26 respectively. They both report volumes with under a 1% vol/visitor ratio.
The biggest movers to the bottom of the charts won’t suprise many, as anyone who has used these exchanges or seen data on CMC sees the suspect volume happening daily. Most of these exchanges can be found using the same copycat UI.
Toward the very bottom end of the exchange list it can be seen that the very low volume exchanges have mostly visitors who are checking prices as their per visitor volume numbers approach $0/day. None of these exchanges would be appealing to real day or short swing traders and this shows here.
Tallying up the volume numbers of the top 130 exchanges, it is estimated that over $6 billion dollars in daily trade volume is being faked with over 67% of daily volume being wash traded. Over 70% of the CMC top 100 is likely engaging in wash trading by at least 3x their stated volume.