EthCC 2 : Ankit Chiplunkar

hello everyone thanks for your patience my name is Anne ketchup linker and I work for token analyst we are a team of researchers who mainly do data analysis of on chain data and today I would like to present you one of our research topics that we also collaborated and published with bit max research and some parts of it were also published in circle how IC o---- is basically invested the money that you give them in 2018 so this is what is that this is what I'm going to talk about so before proceeding forward just let us have a look at what the structure is going to be so first I will introduce you what we are what is the team and what was the I assume any of 2017-2018 what was the mindset of people that who were actually investing at that time period then there are three main technologies that we had to develop in-house to actually be able to do something like this and so I will talk about the three main infrastructure pieces that we have built in-house and then finally like a brief macro overview of how I see us actually invested that money so start this is our team token analyst this is a shaker and Jehu our co-founders Sid did his MS in machine learning in UCL this is Jade bug that seed invest initially then gender akeno's almost any computer language possible this guys me actually says I did this PhD in semantic web and Seamon did his PhD in machine learning so we are a bunch of people who have deep expertise in machine learning and data analysis and before this we have also published several research pieces this one was in collaboration with bit max we have done research in collaboration with dyer circle and also write our own blog post so on and so forth these pieces so before moving forward let us just have a look at how the ICO market let us just have like an overview of what the ICO market look like in 2017 2018 before April 2017 before because he theorem is a lack of highly flexible smart contract language you can have a huge amount of flexibility in how you basically structure your investment or how you basically structure their ICO and before April 2017 there were quite a few teams who were raising money using IC O's but in my opinion they did not have idea of what Hugh scale it would encompass over a period of over the coming ones in it and so like what was initially before April was happening was I SEOs were being so teams would have a cap or an e theorem cap that they want to reach and this cap was being reached a very quicker and quicker so gnosis basically came up with an idea that they would have a reverse dutch auction where the price of them they had a fixed amount of money that they wanted to raise and so their valuation the conversion between ether and that money would change or drop over time and so they basically wanted like if people wait long enough they can buy canosa's tokens for a cheaper price but that did not actually happen canosa's raised 12.5 million in 12 minutes and the team had the team held almost 95% of the tokens this is so this was the mindset that people were going on going at that point of time because like if you don't submit money as quickly as possible then you basically move or you cannot basically participate in that particular ratio then aragón sale happened there is 25 million dollars and I'm talking about million dollars in 26 minutes which is almost equivalent to 1 million dollar a minute and if you have an average and if you have 4 blocks in a minute then it is 250 K average then the bank or sale happened and they were supposed to raise 100 million dollars but they released 153 in three hours so this was like the mindset that people are actually going into while investing in these companies with the white paper basically and then basically OS came along and just said oh we we want each and every one to participate and we are we don't have a cap on we don't have a time limit and they they raised for the whole year so in huge number they're 450 raised in 2018 there is 12.6 billions in USD in 2018 and gos itself raised almost four billion dollars so that was what we were going into so there are three main methods so like our objective now was to somehow be able to identify how tokens are being moved around from these icy wallets and also he theorem that people gave them are moved around from these icy or Treasury wallets and so there were three main pieces of infrastructure that we had to build the first one was somehow being able to query historical token balances which so there are two main methods on how to do it basically one way is if there is function available in the smart contract you can just query the balance or the token balance of that particular address in that smart contract but that is not any a/c 20 standard so many teams do not many teams do not enforce or provide that function provide that function but then they have a huge storage overhead on the on the blockchain basically another method is because you can assume aetherium as a state-transition machine where you have a state n at a particular block number and correspondingly the smart contract store their values in storage so each storage as a key value pair if we can somehow and plan and transactions are basically changes that happen or changes that operate on that state so if we can somehow extract the friends between state and storage of these smart contracts historically then we'll be also able to look at the historical smart contract or token balance basically for each and every token interest that we are interested in so yeah each block has several transactions each transaction has several state dips where what is the change in a particular account and each account has several storages so if for example you are transacting in honor decks and you are changing token a to token B then we also extract how much token was extracted from some medicine token E and how much token was a deposit in some token B and we did this for all the historical data on chain and at one point of time we were running like 95 full sing parity archive notes which was 3k dollars a day but that was for a very brief amount of time so because we have like all data possible to be extracted we were basically able to create the state of each and every smart we are c20 smart contract at any snapshot in time so I will just give you a brief demo of how it works the video will not work yes this is our in-house tool where we have connected all the data possible we extract all the data possible from aetherium and put it into a MapReduce kind of architecture so here what we are doing is we are just querying for historical balance of one of the holders of B&B token so this is like a very flexible way to generate queries and to analysis on the data that we have this is a 20 token Yesi balance diff table is basically generated from all the storage differences that we extract in the box chain I will this demo yesterday so yeah I'm getting the smart contract address of B Envy and also like the address of the holder that I want to extract basically and then once you have these once you have you once you have made that very flexible query you can execute that query and you get results in seconds and then you can plot it out and so this is how the balance of that smart or this is how that is how that address was holding the BnB token on number basis and so what we can also do is we can combine these flexible queries to make dashboards so this is some dashboard that we made for fiber and unis hope this is another dashboard that I was working on for died yesterday so that is one piece of infrastructure that we have built to basically very quickly get access to all the token holders and how how it is distributed bienvenido very severe very severe nearly description daily monster lick homie but that little is a poor control inordinate oh yeah so that was one piece of infrastructure the other piece of infrastructure is like we had to build really low latency alerts for some of our clients so what we did was we run huge amount of nodes so we have a get node we have a parity node in you parity node in u.s. and then there is this open software open source software that we have made open source which basically queries a node for as soon as a block or a transaction comes in and pushes it to a Kafka pipeline and on top of that we built a confirma which is like finality filter so it combines all the data of all the duplicate blocks and transactions and imposes a finality of six block confirmations and then you get a stream of confirmed blocks transactions events whatever and using this we have also built very low-level API in which you can just query the blockchain using our JavaScript SDK and in three lines you can get connected to the ethereum blockchain and get yourself connected to the matrix basically so here is the stream of all events that are happening on the blockchain in real time and what you can do basically do with this is you can just grab this data from the terminal and create really simple BOTS and we have as a demo we have made some telegram and slack BOTS as well using this API and this API is open source and anyone can use it right now you can you just need to go to this website and you can plug yourself into the etherium matrix the third piece of infrastructure that we had to build was we needed to somehow intelligently identify address labels that correspond to a particular ICO or a particular exchange so if this is a chart of all the addresses possible on atiim and this and the ones in colored are what we were able to extract from ether scan if we apply very general machine learning algorithms we were able to basically extract a much more wider range of address labels when we compare it from like hand-picked labels from either scale and one use case of this is for example this is our graph database that we built for Bitcoin basically but you can extrapolate it to a theorem as well so there is this dark market address which we have labeled as dark market and then there is this sum exchange and we can basically identify in how many hops money is being transferred from this address to this address basically so these are the three main pieces of infrastructure that we had to build to actually be able to do some sort of analysis like this and like if you guys are interested to use our services you can just hit me up on Twitter my email address whatever now coming to the main point why most of the people are over here is how actually the ICU spent their money so this is like a macro analysis I am NOT going into how individual teams have spent money right now but this is a macro analysis of what happened to the eath that was raised so AOS raised almost 7.

2 million aetherium and the USB equivalent of that is 3.8 billion dollars the way we calculated this was we identified a cluster of Eos the accounts and then we were noting how money was going into those Treasury accounts on a monthly basis and the USB value is calculated by the etherium price of that month so basically if 100 eath is entering that Treasury account and the ECM price of that month in 200 basically is what he was raised for that particular month and as a comparison other projects rate 7.9 million each which is 1.6 which is equivalent 1 six billion dollars there's a disparity because like the price of e theorem itself fluctuated significantly so even if these guys there's much more amount of heat but the USD equivalent was much lower so effectively in before September 2018 ico says fifteen million dollar fifteen million aetherium which is equivalent to five point four billion dollars and we estimate that yes almost all of their aetherium when we sold it has to be in quotes because we track etherium being transferred from their Treasury account to some exchange or some other account and we assume that if they are not holding custody to the etherium then they they have an intention to sell basically so we cannot identify if they have sold when they have put the money in an exchange but we are assuming this they are not holding custody then at least there's an intention to sell and similarly the money that was or the Fiat equivalent Holdings were calculated in the same manner by comparison other projects transferred 4.1 million of Heath and all in all I think on a global scale if we compare then the total USD equivalent of aetherium that was raised actually sold by now there is still a huge amount of aetherium that is remaining in with other teams but they it's still either unrealized profit or unrealized loss now let us have a look at what happened on a time basis like how the token tragedy etherium we read on a time basis so in blue we have the etherium balances of Treasuries for all the all the i ce o--'s and in orange we have the price of aetherium at that point of time so we actually see that when a theorem is very high there are some team that sold there it and then when the second peak occurred and it started dropping down team started significantly liquidating and converting their etherium holdings into fiat Holdings and maybe like it it might have caught another like a secondary effect to raise to create like another price drop so like all in all if we compare the gains us in our estimates in 68 million dollars by just converting from their aetherium Holdings to fiat Holdings other projects gained almost six hundred ninety-two million dollars and lost almost thirty four so total net gains are almost 800 million dollars as of September 2018 the other study that we did was what happened to team tokens basically so if we look at team tokens like the distribution of team tokens is highly centralized we identified again for those 222 icos what is the distribution like how much percentage of tokens are held by the top 20 accounts so almost 90 to 95 percent of I SEOs in almost 90 to 95 percent of I SEOs top 20 accounts hold sorry almost 32 percent of I co-taught to at or 20 accounts hold 95 200 percent of the tokens and so on and so forth and these are some some of the examples what is very unique thing that is happening is the team then total raised $5,000,000 $5,000,000 but they gave themselves 24 billion dollars of tokens out of thin air which is a pretty good business model I think so we assume that out of the out of the eyes here and out of what does contract was holding teams gave themselves 21.

5 billion dollars and after the ICO they gave some in total they gave themselves 24 billion dollars but they lost significant amount of money mostly because the value of the tokens decreased over a period of time so a lot of teams are still holding their tokens but the value in itself has decreased significantly yes tomato nee 1.5 billion dollars of sell-off has happened of these tokens and like an over just to give an overall picture so in in a few numbers there is 5.4 billion dollars of etherium gave themselves 24 billion dollars of tokens any theorem terms there was like an equivalent amount of money was like converted to feared but there's a huge loss in terms of price decrease due to the tokens themselves so that is all my talk if you want to read like a detailed analysis of what happens to each and every token teams just go to this research that we have published and thanks a lot to conclude I think what we have come up with is good infrastructure to basically be able to do real-time analysis and have an identify at this clusters I SEOs sold and overall made it 800 million dollars and on the other hand gave themselves 24 billion of which 80% value was lost and like from a personal perspective I think it is super ironic that we have in this ecosystem we always strive to be decentralized and personalization but we see Nico we see a white paper and in minutes we are sending millions of dollars to that remote team based on advisers and memes and I think from possible pecks perspective investment has a huge history in even in Addie's even in a centralized world and there are certain structures that you can identify for example there should be a cliff of investment and the people should be paid money little by little and these kind of structures can be actually enforced by smart contract regulations and I think we can imbibe those structures into smart contracts and enforce them as regulations basically yes that is my talk do get in touch if you want access to data and do we have time for questions or but yeah [Applause]