… on the Surface


By Lisa Pinskaya, an analyst with HASH CIB
l.pinskaya@hashcib.com

We have recently published a Markets report covering much of the recent action in the cryptoasset space. There, among other things, we presented a case study of XRP, which caught some attention during the November 2018 sell-off. Its recent listing on Coinbase, with just announced retail wallet app support, sparked anew the faded interest in XRP’s price action. So we decided to publicise some of our understanding.

Unfortunately, our reports are unavailable to general public. If you are a financial/digital asset markets professional, you can fill out the form on our website with actual (and verifiable) information to request free-of-charge access. Below we do not give financial advice, but only present part of our findings for informational purposes only. We do not present it as any fact or evidence, but rather a perspective on certain data points we found curious. Any substantiated critique or commentary is welcomed. Thank you.


This Medium post is mostly aimed at people familiar with XRP (the digital asset) and Ripple (the company). If you’re new to the subject or would like to refresh/enhance your knowledge — take a look at the following materials we find most valuable and informative:


Summary:
  • The market is used to seeing XRP price spikes and subsequent deflations on positive news triggers. We have summarised all 20 of such periods and 3 additional, seemingly unfounded, over 2017–2019 YTD.
  • We’ve charted XRP’s price and turnover volatility as well as rolling correlation to the market, to better visualise its historical performance.
  • In search of particular causes to XRP’s price action irregularities, we decided to look further at some on-chain data.
  • We have found a potential discrepancy between the latest Ripple Markets report (Q4 2018) and our findings.
  • Also, we have found a potential divirgence in the use of funds that are released from special set of on-chain escrows and Ripple’s declared intention to “help support the XRP ecosystem” with unlocked funds.
  • Finally, we have found additional XRP balances, possibly controlled by Ripple, that increase the company’s share of the free float of XRP to 36%.

Again, we do not allege that any of our findings is accurate or true. We’ve used and referenced only publicly available resources, so third-party analysts could validate our assumptions.

XRP market action

As crypto traders know by now, XRP tends to “pump” quite sharply around the positive news feed (often times related to Ripple, the company, rather than directly to XRP, the asset) — with subsequent cooling off. Recent listing of XRP on Coinbase turned out to be a non-event, due to the relative insignificance of the price spike (5% over 5 days). Nevertheless, it reminded us of prior, much more noticeable “pump” events. To understand the XRP price action patterns, we have analyzed all of its price spikes connected to positive news throughout 2017–2019 YTD and compared XRP returns to those of the market (an equally weighted portfolio of the three most liquid assets, ex-USDT).

Source: CMC, HASH CIB; “Market” means a portfolio consisting of BTC, ETH and LTC (equally weighted); “EOP” means end of period.

Additionally, we have spotted three alpha-generating (i.e. better-than-market performing) price spikes in XRP, that have no seeming relation to the news flow surrounding Ripple or XRP.

Source: CMC, HASH CIB; “Market” means a portfolio consisting of BTC, ETH and LTC (equally weighted); “EOP” means end of period.

Below we have calculated the averages for duration, total excessive returns and daily excessive returns for the XRP price spike periods for: the whole period of 2017–2019 YTD (including and excluding one extreme outlier event in 2017), as well as 2017 and 2018 separately — as we would argue those two had different market conditions throughout.

Source: CMC, HASH CIB

If we visualize the above presented findings, XRP looks very much like an asset, that trades with some irregularities. The chart below shows a rolling 15-day price volatility for XRP compared to the market through 2017–2018.

Source: CMC, HASH CIB; “Market” means a portfolio consisting of BTC, ETH and LTC (equally weighted)

Looking at turnover volatility (see below), we can see the same pattern as on the price volatility chart, but the divergence with the market is even more apparent. XRP turnover volatility is always greater than that of the market, while the asset has been seemingly liquid, usually among top 5–7 traded (based on daily volume), for years.

Source: CMC, HASH CIB; “Market” means a portfolio consisting of BTC, ETH and LTC (equally weighted)

XRP correlation to the rest of the market dips fairly regularly, especially during the dates related to news flow. The chart below shows a rolling 30-day correlation of daily returns coefficient between the XRP/USD and the market (a portfolio consisting of BTC/USD, ETH/USD and LTC/USD pairs, equally-weighted). Coefficients with magnitude of less than 20% have little if any linear correlation. For example, between ETH and LTC the portion of such days in a year is 14% on average while between XRP and the market — 24%. The periods of abnormally low correlation take 10 days on average.

Source: CMC, HASH CIB; “Market” means a portfolio consisting of BTC, ETH and LTC (equally weighted)
Ripple’s On-chain

In search of what could have caused some of the irregularities on the price charts, we looked into on-chain data. The full XRP ledger takes up ~9 TB in storage — containing history from January 1, 2013. While the functionality of the protocol doesn’t support smart contracts, the reason it takes up so much space is that each block of transactions is recorded with much meta-data, we would consider useless. Even with a good bandwidth and hardware setup one would need around 6–7 months to download the full history. However, a recently Ripple-sponsored third-party developer Wietse Wind has made the full Ripple ledger available via Google’s BigQuery public data analytics tool. While we have since found a bug in the dataset, that relates to the storing of transaction memos (acknowledged by Wietse), other data seems to be accurate. We have double checked over a 100+ random transactions from Wietse’s dataset with the most-popular third-party block explorer for Ripple — Bithomp.

Wietse and the Bithomp team each host a full node that is part of the 26-member Unique Node List of RippleNet’s validators, recommended by Ripple (out of 153 total, at the time of writing). We used data from both as our main sources — hoping, if anything, that it proves the lack of bias on our behalf. Both Wietse’s and the Bithomp nodes were added to the “canonical” UNL as part of the strategy to make XRP more decentralized. The criteria for chosing such validators is past “performance, reliability, and security”, per Ripple. As a result of initiatives to further decentralize the system, Ripple currently maintains to have control over 7 of all 26 validator nodes (i.e. 27%) in the recommended UNL. We hypothesyze it could also have more informal connections to the majority of the remaining UNL nodes, bringing its potential affiliations up to 19 of 26, or 73% of all the recommended nodes.

While the aforementioned hypothesis doesn’t mean UNLs could coordinate their transaction proposal/validation activity, it could imply less decentralization in terms of power being spread out between totally unrelated, independent parties.

Also, though not directly relevant to the subject of this post, we have found that throughout all of RippleNet’s available history 500 most active accounts are responsible for 93% of total transactions, with the top 100 having executed 78% of those. In the table below, we have compared the stats with another top-3 network by market cap, Ethereum (comparison with Bitcoin is complicated by the UTXO count).

In respose to accusations of controlling the XRP supply, on December 7, 2017, Ripple moved 55bn XRP to 11 programmed escrow accounts that automatically unlock 1bn each month. We have analyzed the amounts of publicly declared un-escrowed Ripple-held reserves, but came across a number of inconsistencies.

The total supply of XRP was 100bn on its launch. Ripple initially controlled XRP 80bn. As of March 5, 2019, total supply of XRP ammounts to 99.992 bn. RippleNet does not have an inflation algorithm to incentivize validators. They provide their services for free, with tiny XRP transaction fees just burnt instead of being redistributed to validators. So, XRP has a slowly contracting supply which was reduced by 8.3mn over at least 6 years.

When announcing the escrow plan earlier in May 2017, Ripple claimed to have control over 61.68bn XRP, which left them with 6.68bn (ex-amount to be escrowed). Since then their XRP balances have been changing, per company’s own reporting. Both the Q2 and Q3 Markets Reports, put out by Ripple in 2018, stressed that the company held 13% of total supply — roughly 6.05bn and 6.1bn XRP, respectively (per our calculation). Yet Ripple’s own assessment as of 21 January and September 23, 2018, presented on its website — claimed XRP 6.98bn in January and 6.96bn in September to be in the company’s holding. This suggests that some 860mn XRP or $490mn (at XRP price of $0,57, on September 23) were added to Ripple’s holdings in just 23 days since the end of Q3, 2018. Further, on 2 December, 2018, the company’s Market Performance web page claimed even more — 7.36bn of XRP, to be held by Ripple. The webpage also states that: “Since 2012, Ripple has methodically sold XRP and used it to incentivize market maker activity”. Ripple could potentially have used open market buys or access to funds released from escrow to increase its XRP reserves. The funds released from escrow are to be used “in a variety of ways to help invest in the XRP ecosystem”/“support the ecosystem”, according to Ripple Markets Reports.

Source: Ripple.com, HASH CIB

According to our estimates, around $1.449bn worth of XRP has been used out of the amounts released from escrow since the beginning of 2018. The remaining funds of the released total went back to escrow — further extending the tally of XPR to be released over time with 1bn/month speed. We have put together a table with all of escrow releases as of now —see below. The Ripple-controlled account referenced in the table is a multi-signature wallet, containing 8 signatories, that has been labeled as Ripple-controlled in the Bithomp ledger explorer. We assume this to be the distribution account for the funds released from escrow. The multi-sig functionality was introduced to Ripple ledger along with the escrow lockups in December 2017. All of the accounts created since then and labeled as Ripple-controlled in the Bithomp XRP ledger explorer are using this multi-sig scheme with the same 8 addresses as signees.

Source: Bithomp, HASH CIB

As the funds which are forever released from escrow all end up on the same Ripple address, we decided to trace how those funds were spent. The wallet in question (assumed escrow funds distribution account) is seemingly used for multiple purposes. So, it is impossible to accurately identify which of the incoming funds were used for “supporting the ecosystem” and which weren’t. Instead we calculated the net end balance for each period — see “Available Funds” in the table below (“Previous Quarter End Balance” + “Total Recieved” - “Released from Escrow”).

Source: Bithomp, HASH CIB

The next step would be to calculate the amount transferred to exchanges from this account throughout the period. If that amount exceeds the amount of “Available funds”, this means that at least part of the funds from escrow were transferred to exchanges. In Q4 2018 we discovered that at least XRP 170mn (around $80.5mn) was transferred to exchanges through a Ripple-labeled (per Bithomp ledger explorer) transition address. After the escrow distribution account transfers the funds (generally the majority of the recieved amounts) to the “transition” address, they almost immediately get relocated to a number of other wallets, including those of exchanges. We do not derrive any conclusions here, intending to just highlight these findings.

We have also checked the data relating to Ripple’s XRP-selling activity, unrelated to release of escrowed funds. Starting from Q4 2016, Ripple publishes so-called “XRP Markets Reports” with key news, market commentary and reports on its quarterly sales of XRP. Ripple executes direct sales to institutional investors via XRP II, LLC, a Ripple subsidiary licensed as a money service business (MSB), other sales are preformed “programmatically”. Ripple’s CTO in a forum post explained that programmatic sales are done by an algorithm, to ensure no insider trading.

According to Ripple’s Q4 2018 Markets Report $88.88mn worth of XRP was sold “programmatically”. We looked at transactions of all Ripple-labeled accounts (per Bithomp) through the last quarter of 2018 and found an account most probably used for this purpose in Q4, 2018. It also uses the same 8 signatory multi-sig scheme, discussed earlier — with Ripple-labeled signee accounts, by Bithomp. There is no way to determine how exactly the account is accessed by an algorithm to execute the programmatic sales. The amount of XRP sold to exchanges, per Q4 Report, matches this account’s gross outflow in the quarter.

An account labeled as controlled by XRP II, LLC on Bithomp ledger explorer, moved another $80mn worth of XRP to exchanges in Q4, 2018. That conflicts the Q4 2018 Markets Report. Ripple reported selling $40.15mn directly to institutional investors. To be fair, no actual explanations of the purpose or mechanics of such sales is given in any of XRP’s Market Reports or beyond, so we had to rely again on Ripple CTO’s earlier cited forum comments. Supposedly, there should be additional lock-ups and other special conditions imposed on such deals.

The total for the Bithomp explorer XRP II-labeled account’s Q4 potential “sales” (i.e. transactions sent out from the account) accrues to $148.7mn or $108.6mn more than reported by Ripple. The XRP II-labeled account mostly sends funds first to unidentified wallets — yet these funds eventually end up on exchanges. The way XRP is being moved there indicates the use of “shell” addresses, i.e. wallets that receive funds and immediately distribute almost the same amount and are never used again afterwards.

Finally, we hypothesyzed that all of the wallets that were involved in fund transfers to the newly-activated escrow accounts in December 2017, could be controlled by Ripple. We used the Neo4j database (prepared by Thomas Silkjær), that allows to graph relationships on the XRP ledger, in order to trace the history and connections of the addresses in question. Then we assessed those wallets’ aggregate balances at each point in time of RippleNet’s history from the first available block of transactions (1 Jan. 2013). As of 5 March, 2019, there were $17,65bn XRP held on those accounts. We also checked a third-party source to verify this separately. According to Richlist stats, put together by Wietse Wind, mentioned earlier, out of 100 richest accounts 14 are labeled as Ripple-affiliated with XRP 17.5bn on their balances in total (as of 5 Mar, 2019). With free float of XRP currently at 48.892bn, per our estimates, our own calculation of aggregate balances on accounts assumed to be controlled by Ripple, comes to a staggering 36% of current free float of XRP. Again, this is just hypotheses.


Summing up the main point of this case study, we want to see the cryptocurrency markets — the real “make or brake” point for the industry — to mature fast enough. We expect the potential “smoke and mirrors” game, seemingly causing distorting effects and price/volatility ripples across the market to gradually end as the bear cycle extends.