Dialogue with Trader Jackson: How to truly capture Alpha through AI + data?

This episode’s guest: Jackson, co-founder of Scope Protocol, Twitter @0xOar

TL;DR

1. For different types of assets and different holding strategies, what data do you need to pay attention to?

Jackson divides his assets into three types. Different types of assets have corresponding holding strategies and data to pay attention to. What is interesting is that he divides asset types according to the holding period:

Long-term holdings (such as mainstream currencies such as ETH SOLANA). Data to watch: The current liquidity situation of the entire industry, there are three indicators:

1. U.S. stock market or U.S. dollar interest rate hike/cut

2. Mint/Burn of Stablecoins

3. Large VCs’ large token holdings

In addition, the non-data level will focus on narrative. Whether a token is recognized by Top VCs, or is an asset that Vitalik and Binance prefer, determines whether it will be held for a long time.

• For assets held for about a week (market value 200-700), the expected return is 10%-20% per week. Pay attention to the data: some abnormal data that can reflect the habits of traders, and then make transactions. For example, after the DEX overbought is triggered 3 to 4 times in a row, it basically represents a very clear upward trend, and you can chase the rise. According to actual tests, most of the tokens ranked 200-800 on CMC and can be listed on mainstream exchanges have a weekly winning rate of 80%-90%.

• For assets held for 4-12 hours, the drawdown can be more than 50%, and the leverage will be 5-10 times. The data to focus on are the indicators that can reflect the change of sentiment in the short term, and there are 3 of them:

• Open Interest, the open interest of contracts in the market. Basically, from the backtest data, if Open Interest is 6% greater than the previous hour, and the price drops by 4%, it is a positive indicator. It is generally believed that it has reached a stage low, or short positions need to be covered. At this time, the winning rate of 4 hours is higher. But the specific situation requires specific analysis. You can check the backtest data displayed on scopechat.

• Funding Rate

• CEX Deposit, if this data has a particularly abnormal change, for example, the current deposit amount is more than 10 times that of the previous week or month, then it is almost certain to fall. The reason is that some large and compliant market makers will not store market making tokens in CEX for a long time. Generally, when there is an action, they will recharge the tokens from the chain.

2. What are the uses of different data products?

Divided into 4 categories:

The first category is macro data products, such as Glassnode, which analyzes the holdings of large investors and changes in liquidity, and is aimed at people who trade around BTC or those who want to see changes in fundamentals.

The second category is products that use data to perform public opinion/sentiment analysis (text analysis), such as Kaito, which analyzes what people are discussing in the entire market, what products some so-called smart investors are tracking, and helps researchers, such as investment managers who have to write reports in VCs, to find what kind of narrative or what products may be popular based on this narrative.

The third category is products that analyze on-chain data. For example, Nansen, ScopeScan, DEX Screener, and DeBank. In fact, on-chain data has changed in this cycle. In the last cycle, we would judge which products would be popular next by the consumption of Gas. In this cycle, on-chain data revolves around short-term Alpha, such as Solana and MEME on Base. The data richness that these tokens can generate is much weaker, so products with fast update frequency, listing speed, and parsing speed will have more advantages, such as DEX Screener. In addition, if you want to track the data of a certain address on multiple chains and conduct in-depth analysis, DeBank is the best choice.

The fourth category is products that meet investigative needs (eating melons), such as Arkham, which analyzes who is behind an address, the specific situation of the positions, and under what circumstances the highest winning rate occurs. This product has a lot of labels, but the accuracy is questionable.

3. What should I do when I don’t know what to buy or whether to buy it?

Jackson believes that it is difficult for ordinary users to use data to assist in trading. On the one hand, there is a lot of data noise, and on the other hand, the threshold for data analysis and use is slightly high. Therefore, his product hopes to provide relatively fool-proof services to ordinary traders who do not have much time to study and do not have special trading talents. There are two typical scenarios:

First, what to buy? Different segmented needs have different designs:

• Copy Trade Demand

1. If you want to trade some more mainstream tokens, the most effective way is to Copy Trade VC-related address. You can directly ask in Scope Chat about the holding structure of the target token, and you will be given a table of VC and whale holdings.

2. If you want to copy trade Alpha assets similar to MEME, you can find out what smart money is selling in the rank.

• The demand for investing in Beta assets. This type of people believes more in narratives and attention. You can refer to two rankings:

1. Narrative Rank allows you to see what segmented narratives people are most concerned about. After narrowing the research scope based on the narrative, you can then look at whether the token should be traded separately.

2. Attention Rank, that is, what 30,000 KOLs are discussing every day, and what tokens are also ranked, essentially capturing the markets attention. If you dont look at these things, you can refer to the score, which represents the AIs judgment through signals that a certain token is likely to rise or fall in the next 24 hours.

Second, now that you know about a coin, do you want to buy it?

There are two possible actions in this scenario:

1. Just ask the AI, “Can I buy XXXX today?” The AI will give you a comprehensive analysis of both bullish and bearish factors, and you can further analyze it to decide whether to trade.

2. Search for tokens on the product to view everyones opinions, which summarize all the discussions and reports on this token by KOLs and news media. The winning rate of each KOL is also marked in the product. The higher a persons winning rate is, the more valuable his answer is.

4. How to use tools to improve trading success rate?

Taking 0xScope as an example, Jacksons daily routine is roughly:

The first step is to check the currently popular narratives, and then filter out assets within the selected specific narratives that have relatively not yet experienced a rebound or some other situations.

The second step is to click on the asset to view the Trading Signal page, which clearly states the correlation between different types of Signals and token prices and the winning rates in different time frames. Choose an acceptable winning rate, confirm that the liquidity of the asset itself is OK, and set an early warning.

The third step is that the next time such an abnormal signal occurs, an alarm will be sounded. At this time, fill in a note, for example, PEPE needs to pay attention to the Open Interest indicator, and then start work according to the original established strategy.

5. What do we buy in the second half? Party Game Token

The features of Party Game are:

• Involves the issuance of new assets;

• It is an ecosystem, and members of the party have a strong motivation to make profits through means other than directly selling coins.

If there is normal input of external liquidity, such as an interest rate cut, more attention will be paid to the assets in this new party game, specific examples of which are Monad and AI Infra.

Everyone is welcome to try @ScopeProtocol.

Original interview

FC:

The background for this issue is quite simple. To be honest, I have always been confused about two things. First, what is the use of data? Second, how should I use so many tools? I have seen many tools, such as Nansen, including Footprint, which we invested in. In fact, there are many tools. How to use them and what are their uses? This matter can actually be understood after thinking about it once, so I found you.

And I think from your previous sharing, you are actually using your own signals to do some transactions. This time I want you to explain the tools and data clearly. When I was at 36Kr, we were mainly working on the data of the primary market. At that time, everyone was concerned about two functions. The first one was called trend changes, such as analyzing the trend changes of the intended financing track through various indexes. This is what all VCs want to see. But in fact, the primary market is slower. The advantage of being slow is that others are not much ahead of you, and the impact on transactions is actually not big, so the data at that time could not be charged. The second one I think is useful is called the observation warehouse, which is to see if there is any abnormal movement in your portfolio. As a post-investment, they actually need it. I know you also have such a function. I think this is a business that does not make money. It takes a lot of time and manual data cleaning.

So I think either you can give us a brief introduction of your background and what you are doing now, and then we can start talking about trading.

Jackson:

Yes, this is very FC. I was invited to join this event. I have also listened to the previous episodes. Everyone is a very professional trader. Compared with them, I may not be a trader with a very professional background, but I have been in the industry for almost 7 years. Two years ago, I started the company 0x Scope. Basically, you can understand it as a data company. We cover a lot of data product lines, not only the Scope VC product provided to institutions, but also some products such as Scopescan that focus on on-chain data, and Scope API for developers. In the past six months, I have made an AI product called Scope Chat. This Scope Chat is a very transaction-oriented product. The purpose of this product is to hope that everyone can make the decision with the highest winning rate in the shortest time. So in the past six months, I have spent a lot of time on the direction of trading, and have been looking for how to use data to assist trading, and how to improve the winning rate of your trading strategy in this data.

Before I started this project, I had been in the blockchain industry. I have worked on almost all types of Crypto products you can think of, including Layer 1, Layer 2, and exchanges. I have always been a product manager. After I started working on data, I have always been adhering to the idea of being a product manager. As a product manager, there is a saying that the butt determines the head, but as a product manager you should never have your own butt, your butt should sit with the user. So starting from working on data, no matter what type of data we work on or what kind of customers we serve, basically my goal is to think about problems from the perspective of customers. So in the past six months, when we were doing transactions, we did think a lot, and today I am very honored to have the opportunity to share with you.

FC:

Thank you. Actually, I was particularly impressed by you because a lot of your sharing was related to scenarios. For example, when you were having dinner with your friends, they asked you what you would do if someone recommended a coin to you. We will talk about your product in detail in the next session. Lets talk about your trading strategy first. If you use one sentence to summarize your own trading strategy, what kind of strategy is it? What is your own trading style?

Jackson:

My own trading style is that I have a very low risk appetite, I am very busy, and I am not a very smart person. Anyway, you can understand that my trading strategy is more suitable for people like me who have an IQ of 100 and a relatively low risk appetite. Large positions come from my own research, and small and medium positions come from my own data and some signals fed back to me.

FC:

As I understand it, when people say I Q1 00, they are actually trying to do something called Smart Beta, right?

Jackson:

About the same. I think the winning rate of Alpha is too low. It may happen, but for me, my own trading strategy is still betting on Smart Beta.

FC:

I see. Can you tell me about your expected return and whether there is a stop loss line? Or the trading strategy, I understand, is your capital scale, your expected return? And for example, your risk control? Of course, you don’t have to say how much money, but you can talk about the detailed explanation of the entire trading strategy.

Jackson:

Basically, about 60% of my positions are in mainstream assets, such as ETH and SOLANA. I have bought some new top-tier projects before, such as ARB, but I sold them all later. The logic is mainly Long Crypto. Then about 30% of the assets will be used to buy assets that are suitable for holding for about a week. Why do I say this time? It is mainly related to the product we are making. The data indicators I may pay attention to for assets that are about a week long may be different. I may pay attention to some trend-oriented indicators. There are also about 10% of the assets. I will play with some assets that are about 4 hours to 12 hours. For these assets, I may pay attention to Open Interest, Funding Rate, or timely news, some security events, and some abnormal data of large withdrawals. For this part of the assets, I may even do some contract-related operations.

I may have different stop-loss rhythms for different types of assets. For example, for large positions, I basically only focus on analyzing the liquidity of the entire market. For assets of about 30%, my income expectation may be about 10% to 20% per week. For 10% of the assets within 4 to 12 hours, my drawdown can be more than 50%, or even higher. However, I must do better than expected income, and generally the leverage will be about 5 to 10 times.

FC:

I want to know what your decision path is for these holding periods. For example, I think you might look at data at the hourly level, but what would you do in the long term? I want to hear about your trading decision-making process.

Jackson:

In fact, different types of assets may have different types of holding strategies, and the data they look at are not the same. For example, for the long-term positions of 60% of mainstream coins, I basically only focus on the current liquidity of the entire industry, whether there is overall liquidity input or liquidity output. This data is more common. For example, whether the entire US stock market or the US dollar has a trend of interest rate hikes or cuts, I actually don’t understand these myself, so I basically look at some other people’s analysis to see what everyone’s overall opinion is. Or we will look at the mint and burn of some stablecoins, or look at the changes in the holdings of some large VCs’ large tokens themselves, which are more fundamental.

For assets that are held for about a week, I will focus on what we understand as a counter-trend strategy. What does that mean? The positions I usually invest in are tokens with a market value of about 200 to 700. After a certain stage, you will find that there may be obvious traders behind these tokens. So for this kind of assets that are held for about a week, my goal is to try to capture some abnormal data that can reflect the habits of traders through data, and then do some counter-trend trading. For example, the data we have recently tested are quite useful. Sometimes you may find the data on the chain useless, and sometimes you may find that the trading volume difference between DEX and spot CEX is still a bit large, but the advantage of the chain is that there are many rat warehouses on the chain. Many rat warehouses are generally reflected in some abnormal data bought by DEX. You will find that if you compare the data of the previous week or the previous month, there are suddenly abnormal overboughts on DEX at certain time points this week. Then these abnormal overboughts are very likely to be the traders own rat warehouses. These overboughts will eventually be recharged to the exchange, but the recharge of these tokens will generally cause the spot price to drop. However, if after a period of time, it is found that the price of the token has not dropped, and even the DEX overbought has been triggered 3 to 4 times in a row, we are likely to think that it may be an indicator that can reflect the abnormal data of the trader. If it can be triggered 3 to 4 times in a row, I have now measured that most of the tokens around 200 to 800, which can be listed on some relatively mainstream exchanges, have a weekly winning rate of 80% to 90%. All the data I am talking about now can be seen on our Scope Chat product. If you don’t believe it, you can check it out and calculate it yourself. These are the signals for about a week. Then, 4 hours to 12 hours, this is a very short-term indicator that can reflect sentiment changes, and there may be only a few that are useful. The first is Open Interest, which is the position of the contract in the market. The second is Funding Rate. The third is CEX Deposit. If this data has a particularly abnormal change, for example, its current recharge amount is more than 10 times that of the previous week or the previous month, then it is almost certain to fall. I can give you an example. Last month, I made more money from a position. That position was a social token. It was highly discussed on social media because of the participation of a big shot. We caught the rise in social media discussion in advance, but did not buy it that day. By the next day, the token price had risen a lot. I suddenly found that it had a CEX Deposit of more than 20%, which came from their market maker. We have also studied this market maker before. It has a very rigorous internal operation process, so it will not store the market-making tokens in CEX for a long time. Generally, when there is an action, it will recharge the token on the chain. So after I looked at the order, I opened a short order, and it really fell a lot.

This is what kind of data I will look at for tokens with different holding periods.

FC:

Lets be more specific, is it equivalent to depositing or borrowing coins in an exchange?

Jackson:

Deposit is CEX Deposit. I was actually confused before, because in theory, the market maker can keep the client assets he holds in the exchange and not move them, so that no one can actually see any of his operations.

FC:

You mean he gave it to the trustee, right?

Jackson:

Yes, it should be custodial. But I found that several large and compliant market makers, when they have such actions, when they want to crash the market, they will charge money from the on-chain. After this set of processes, they will take it out and put it on the on-chain. Later, I asked several market makers, and one of them told me that they have this requirement internally, so sometimes I think this is also very strange, but in the end it is reflected in the data. You can now see the backtest data of our product, and what is its winning rate after similar situations. We found that the more compliant the market makers, the higher the winning rate of similar behaviors. On the contrary, some market makers whose on-chain actions are not frequent enough may be more black box.

FC:

I understand. I heard you mentioned that you held ARB before but sold it later. Why did you sell it? Is it related to your data role? For your long-term position.

Jackson:

For long-term positions, I basically use data to look at some fundamentals, such as liquidity, which I dont think has much to do with data. If you put aside liquidity, some things that can be held for a long period of time are basically related to narratives, and the narratives that these people or institutions who can lead attention want to promote. So I basically rely on research on the product itself to determine what kind of tokens I should hold and how to trade. I will judge that a certain token may be recognized by a top VC or institution, or an asset that Vitalik or Binance likes, and I may hold it for a longer period of time. If none of these are available, then the risk may be too great for me. I used to think that Layer 2 was a very good scenario that could solve the scalability problem of ETH, but later I gradually discovered that people didnt seem to be that loyal to ETH, and were even a little scattered, so I sold it.

FC:

I understand. Assuming we are now focusing on Layer 1, I don’t know if this is within the scope of what you are doing. There may be some dimensions to focus on Layer 1, such as before Sui’s pull-up, its TVL increased by about 2 to 3 times. For example, data with a span of months or tracks, are you doing this? Or what dimensions of these data do you think will affect transactions?

Jackson:

I think that the cost of counterfeiting TVL and DAU is too low. Because in the first year of 0x Scope, we developed the product Scope Scan. One of its biggest features is that we have an algorithm that can calculate which addresses may belong to the same person. Through this function, we found that in fact, the water content in some data related to the fundamentals of the project that everyone can see in the market is too large. I can give you a few examples. There are several chains and several head projects on the chains that were relatively well-known at the time. With a very basic algorithm of deep learning, it can exclude almost 95% of fake traffic. That is, 950,000 of the 1 million users may be fake traffic. So after analyzing the data of that year and seeing a lot of very deep data, I think the difficulty of counterfeiting in this kind of thing is actually very low, and there are too many counterfeiting cases. So if you let me trade by myself, I will never be sure. I don’t know whether the TVL comes from a few people gathering together. In fact, you can also see that there are many chains whose DAU has always been very high, but there are no hot projects on these chains. So I haven’t looked at these data for a long time.

FC:

I have another question that I am more concerned about. For example, you must have done some competitive research, or you must have an understanding of this market now, because you should have had Nansen before you did this, right?

Jackson:

Yes, the opportunity for me to make this product is actually quite interesting. I was actually tagged as Smart Money by Nansen before. Before LUNA exploded, I had several arbitrage strategies, one of which was arbitrage long, and another was short in CEX, which might be an arbitrage strategy, but Nansen marked my address with a relatively high winning rate as Smart Money for a while. Later, after one of our colleagues told me about this, I actually found a point that the on-chain address does not have KYC, so if you use single as the basic unit for data analysis, the noise inside will be very large, so I had an idea, hoping to solve this problem through knowledge graphs, and change the smallest unit of data analysis from single to an entity. In theory, we have achieved part of it, and later we found that everyone in the entire industry may be like this, so we started to turn to some other directions.

FC:

I see. What I want to ask is, in your opinion, how many categories of data products are there? What kind of people can use them for different trading strategies? Or who are the products designed for?

Jackson:

The first and biggest one is the macro data product. Glassnode is the most commonly used one. I see Ni Da often uses it. Because Glassnode does a good job of analyzing the data of BTC and BTC ecology, it can analyze the positions of some big investors and the changes in liquidity. So people who may trade around BTC can use Glassnode, or you can use Glassnode to see some fundamental changes. It is a pretty good product.

The second type is that you need to use this data to do some public opinion analysis and sentiment analysis. For example, Kaito is essentially a large language model plus a search engine architecture. In AI, we call it RAG. RAG is actually very helpful for text data processing. In fact, our Scope Chat can also find some text data and do some Twitter analysis, but Kaito may do better in this direction. These types of data are more suitable for researchers, such as investment managers who have to write reports in VC, because in most cases, the focus of VCs work is to judge trends and help my organization find what kind of narratives or what products may be popular based on this narrative. People may use this kind of product that analyzes text data to judge what people may be discussing in the entire market, what some so-called smart investors are discussing, or what products some so-called smart investors follow. I think if you have this need, you may use a sentiment analysis product like Kaito, which I think is good.

Some people will use on-chain products like Scope Scan or Nansen. Although I have also made a product similar to Nansen, you can actually see that Nansens traffic has been declining. To be honest, our Scope Scans traffic has also been declining. I think a very important reason for this is that the difference between this round and the previous cycle is that the data richness on the on-chain in this cycle is much weaker. The on-chain data in this cycle may not have a so-called intermediate state. It revolves around the alpha that is particularly short, flat and fast, such as Solana and MEME Coin on Base. These MEMEs may only be able to generate a few types of data, resulting in a very low overall threshold. The final users demand for this type of data will be transferred to those products with a very fast update frequency, a very fast coin listing speed, and a very fast parsing speed, such as some products like DEX Screener.

Another type of demand is investigation-oriented. For example, if you want to know who is behind this person, or the specific situation of this persons holdings, what is the most accurate one? Many people will use products like Arkham. Because it does have a lot of labels, but we doubt whether it is accurate. If you have an investigation need and want to eat melons, I think Arkham is very useful.

If you want to do some in-depth analysis of an address you follow, to see its holdings, and to see its multi-chain data, DeBank is actually the best in this regard. Although they are all competitors, I can honestly say that DeBank is indeed the best in this regard.

Therefore, in the on-chain data, you will find that it rarely has the same intermediate states as before. In the last cycle, we would also pay attention to which guest consumption is higher, and we can see which gas contracts represent which projects. We can analyze and judge the projects to find out which products may be popular next. We will study the products in advance and participate in its DeFi Summer or other similar actions. However, there are few such things in this cycle. This is the situation of on-chain data.

So what are we doing? After working with data for such a long time, we gradually discovered that data products actually have a big problem, that is, there are too many noises or influencing factors. In addition to eating melons, or doing some very deep analysis yourself, in fact, for ordinary users, I think it is the kind of level of watching the excitement. I have seen a lot of media and they also forwarded some of the things we posted, but is this kind of thing useful for your transaction itself? I think most people don’t really feed back the data to their own trading system. The main reason is that the data has a lot of noise. Different tokens have corresponding data indicators. You may need to analyze them according to local conditions and analyze them case by case. This actually leads to the fact that if you want to use the data well, its cost is very high. So you will gradually find that only some Degens and some VC investors usually have a high demand for data.

Degen might study the data to see who the rat-warehouses of the local dogs he wants to attack are. There is no safety risk. After reading it, this earthquake with a relatively high gambling nature may rush directly. After rushing, it depends on fate. PVP. Researchers look at sentiment analysis and smart followers. Several American institutions follow several projects with less than 2,000 followers. I quickly follow and DM to see if I can get some quota. In the end, you will find that everyones demand for data and actual usage needs will become these two categories. We will do a test later. We have launched an AI product and let everyone ask it casually. In the end, we found that ordinary retail investors have two questions. One is what to buy OKEN? The other is whether this Token can be bought? I think this is actually very interesting. I have been thinking about how we can lower the threshold for analyzing data so that more people can use data to assist transactions. After the big language model comes out, I think this should be feasible. So what are we doing now? I don’t directly tell you how much CEX has been recharged, what tokens a certain smart moeny address has bought, or you need to open etherscan to check who bought first, whether this person has sold, and other similar data. I filter out all the data, whether it is on-chain or off-chain, that may affect the price, through deep learning algorithms, and then report the abnormal values of these data to users. The most important point reported to users on our product is what we call smart signal. If this signal is reported to you and you think it is abnormal, I will give you these signals and some previous measured data. If you believe us and think that this matter has a high chance of winning, you can follow the order.

What you are actually following is not a certain smart money. It is possible that this smart money is not smart. What you are following are these abnormal data signals. If you are very interested in this abnormal data signal, our product also provides you with the ability to do in-depth analysis. In this way, we can actually solve the problems of some ordinary retail investors. These people may be like me, with little time, and no extraordinary talents or super high wisdom. We can remind you through a smart signal. In fact, I think your winning rate will be significantly improved, because we have tested it ourselves and found that the winning rate will be significantly improved after using this product.

FC:

How do you usually use it? For example, how can your product be used to improve the winning rate?

Jackson:

It’s actually very simple. You can look at some of the sharing we did before. I usually do this: every morning I check what the hottest narratives are. When you open the product, there is a column for hot narratives. This column will look at which good narratives are available based on the discussion and the increase in the narrative itself. After selecting the narrative, one of my common strategies is to find out which assets have not yet made up for the increase or in some other situations. I click into the assets and I can see a page called trading signal. It explains in detail what the correlation between different types of signals and token prices is. What is its winning rate? What is its 4-hour winning rate? What is its 12-hour winning rate? What is its 7-day winning rate? Once I find which winning rate is acceptable to me, for example, 70% is acceptable to me, and I think the liquidity of this asset is also OK, and it does not have any abnormal situations, then I will set an alert. After setting the alert, it will alarm the next time there is such an abnormal signal. After the alarm is completed, I can fill in a note in the alert, and I will write it clearly. For example, pepe is a token, and I often pay attention to its open interest indicator. When I find that its open interest has an abnormal value, it will alarm me. After the alarm, I will start working according to my original established strategy.

FC:

Understand. I dont know if you have seen a video where Erbao and his team predicted on an AI that the price of Bitcoin would reach 190,000 US dollars or 1 million US dollars in a certain year. The question I want to ask is, this so-called AI prediction of long-term, I think its business model is particularly interesting, because it uses that video to attract traffic, and after attracting traffic, the most important point of charging is to predict the price, 3 U for each prediction. I think the business model is actually quite good, but it is easy for it to deceive people to come and buy these 3 U, you know. So I want to know, from your perspective, like this kind of long-term prediction, first, do you think it is accurate or not, and second, what do you think is the logic behind it? And how should we view this kind of prediction brought to us by the so-called AI?

Jackson:

I think AIs prediction logic is actually very simple. You feed AI some technical indicators about the past of tokens, such as MACD, EMA, etc. The big language model will give you a more general logical judgment. For example, if a golden cross or a dead cross appears, what will be the trend? I think it is likely that what you are talking about may be such logic.

In fact, we call this thing indicator trading. This indicator trading is to a certain extent judging the future trend of the entire token through some abnormal indicators that are happening now. I can give you an example. Generally, in our products, we find that dex is overbought. If it happens about three times in a row, the winning rate will suddenly rise from more than 60 to about 70, 80 or even 90, right? It will even continue to rise. In fact, this logic is a bit like chasing the rise. You find that the token in the market may have risen by 1% or 2%. If you chase it now, it will most likely rise to 8% or 9%. In a similar situation, it is essentially a case of discovering certain trends through abnormal situations, and you chase the rise or sell it. I agree with this method to a certain extent, but this method must be based on a large number of different types of data that can reflect this trend, and it may be effective. For example, when we first started making this product, we only had some so-called technical indicators, but not all tokens are suitable for these technical indicators. Only some tokens with particularly good liquidity and no wild dealers may be more in line with these technical indicators.

However, a large number of fast tokens may be backed by several wild dealers, and there are many tricks, but these tricks will definitely leave some traces, and these traces may be on the chain or off the chain. We caught these abnormal data through backtesting, which may be some traces left by them. When we follow the traces to make transactions next time, at least most of them are still valid in the three-month cycle. So to sum up, I think it is possible to achieve it to a certain extent. But the logic of realization is actually that we use data to capture the so-called trend of the future rise and fall of Tokens. If you want to make the trend capture more effective, the better way or more practical way is that you must cover more possibilities before it is OK. We really spent a lot of energy in this regard.

FC:

So there need to be enough factors. First, you need to cover as many possibilities as possible, and then pick out the relevant ones from them.

Jackson:

Yes.

FC:

Assuming we use data to trade, should we pay attention to its risks, or what kind of failures might occur?

Jackson:

In fact, I think this is all case by case, but this is also a benefit of doing data, that is, in fact, our requirement for all indicators is to backtest for at least 180 days, and then provide users with valid data, only the most recent 30 days, 60 days, and 90 days or so. Generally speaking, I think that a certain indicator may become invalid in about 30 days or even two months due to changes in market sentiment itself, so the larger and wider the range of your backtest, the higher the possibility that you can cover the same type. I cant pat my chest and say that this thing must work, but at least through our testing, and first of all, our backtesting, but backtesting has a certain degree of overfitting. We have thought of many ways, but now we have tested that most of the indicators are still valid for at least one to three months. If its market environment has changed greatly, then some other indicators may come into play at this time.

FC:

I see. I remember that there is a token trend on Nansen. There is a foreign user whose name I dont remember. What does he think of Nansen? First, look at the changes in the overall transaction volume trend. Second, look at who the transaction volume is with. For example, if there are 6 people he knows, KOLs or Smart Money, and the depth of the pool is OK, he thinks that this thing is likely to be good, so he will enter his initial screening. After the initial screening, he will look at Twitter, the subject matter, and the content. If he thinks it is not bad, he may make a judgment within an hour and jump in. I listened to your sharing before. In fact, there is a very interesting thing when you design products. I think you are very sensitive to scenarios. You will say, for example, we are drinking in a bar now, and a friend says you are going to buy a token. Generally, people will ask you if this is reliable? Or you can give me a reason. Maybe now you tell me a coin, and search in Scope Chat, it will tell you a general logic, whether it is reliable or not. I dont know what other scenarios you have when designing? Or do you have some assumptions that you can use?

Jackson:

In fact, we can summarize it into several scenarios:

First, when users don’t know what token to trade, you have to provide them with a choice and a set of methodologies to teach them how to choose. So we designed the rankings into several types, such as copy trade, which is the demand for copy trade. Generally speaking, if you want to trade some more mainstream tokens, the most effective data we found is to look at the changes in the holdings of some VCs. There is also a tricky thing here, that is, our product has two VC labels, one is the label disclosed by VC, and the other is VC relative label. There are many relatively large assets, especially the top 100 assets. The addresses disclosed by VCs often hold these assets. Some related ones are very likely to sell these assets. If you hear a lot of Layer 1 and Layer 2 now, they are basically in the category of their reduction. I think you can find some so-called trends here. It is best to copy trade some so-called VC related addresses. How do you check this thing? You can directly ask in Scope Chat about the holding structure of the token you want to trade. We will give you a table that includes the holdings of VCs and whales. If you want to copy trade all these alpha assets, such as memes, you can find out what smart money is selling in the rank. There is another type of person who also invests in Beta assets. He may have an investor background and believes more in narratives and attention. We give this type of person two ranks, one is narrative rank. We divide narratives into very fine categories. AI is divided into AI Agent, AI MEME and other ranks. You can see what narratives people are more concerned about through ranks. After narrowing your research scope based on narratives, you can then look at whether these tokens should be traded separately. Another thing is that we have made a rank for what tokens 30,000 KOLs are discussing every day. In essence, it is to capture the attention of the market. For example, today TON and BLAST are ranked higher, but some of them are more positive feedback, and some may be more negative feedback. If you don’t look at these things, we still give you an option. We will directly calculate a score through the signal. This score represents whether our AI thinks it is likely to rise or fall in the next 24 hours. However, this is the Beta stage and you can believe it or not. So these are the major scenarios for selecting coins that we have set ourselves. You can choose tokens in the copy trade scenario, choose tokens based on narrative or attention, or just believe our signals and choose tokens.

Second, you know a coin or you have heard about a coin, what should you do in this scenario? In terms of product design, first you can directly ask AI, can I buy XXXX today? Just ask a question like this, our AI will give you a comprehensive indicator, but this thing is a bit like you can buy and sell, because we need to give you a more comprehensive analysis, it has bullish and bearish, and try to give you some of its good and bad, and then you can analyze whether to make a decision on the transaction. Another situation is that you may see someone mention a token on Twitter, but you are not sure, because everyones evaluation may be consistent, then you can search on our product, and after clicking in, you will find that there is a view of everyone. We directly summarize this view. All KOLs mentioned it, what they are discussing, what they are discussing 24 hours a day, and whether all news media have published some news about it. There are many interesting points here. We just posted a tweet today. Some KOLs have a really high win rate. In the past month, the win rate of shouting orders was about 86%. There are even a few KOLs, perhaps some of the more famous anti-pointers in our Chinese circle, I wont mention their names, but their win rates are quite high. There are some KOLs that everyone may respect, or these KOLs that many people trust, but their win rates are actually very low. The win rate of each KOL is also marked in the product. You can directly see who is discussing this token and what the win rate of this person is. The higher his win rate, the more valuable his answer is. If his win rate is very low, but he posts a long string of words, and many people still like it or something, it is possible that he either brushed it himself or it may not have much (reference value).

FC:

Understand. In fact, I have seen it before, for example, you also heard my previous conversation with the trader. Do you want to know what answers you are looking for in the process?

Jackson:

Actually, I said at the beginning that I am very busy and don’t have much time to conduct in-depth analysis. I am not a so-called Degen. Another point is that I don’t think I am a smart person. So when I found that this smart signal is more useful, I found a lot of professional quantitative people, some professional traders, and I discussed with them how you use the signal itself in the quantitative process, why this thing is so easy to use, or why this thing sometimes has such a high winning rate. So like you hosted these few events, I listened very carefully. I hope to hear some of their combing of the entire trading process. For example, one thing that everyone mentioned is the so-called emotional transmission. Where you are in the emotional transmission link represents what kind of money you may make. In fact, it is the same when you do signals. If you catch this anomaly in advance, you may be in the emotional transmission, maybe not the top level, but definitely the second or third level. If it can be transmitted to level 6 and level 7, then you are likely to be able to capture part of Beta, right? So essentially, I still hope to learn through communicating with different people to prove that this thing is indeed valuable.

FC:

I see. Finally, I am curious about some growth paths. If you were asked to choose, do you have a favorite trader or trading style?

Jackson:

There is no particular trader or trading style. We have now reached this point, and I can honestly tell you that we are trying to develop a quantitative strategy based on data and indicator trading. I mentioned earlier that, especially in some high-frequency scenarios, I don’t trust my own judgment or my luck. I have done so many things, and ultimately I hope to have a so-called trading robot that is data-driven and can capture a lot of data that others may not be able to capture, and then make money based on this data. Now we are working towards this direction. Based on this direction, we have actually been talking intensively with many quantitative funds recently to learn from their experience and ask them how to adjust their models, etc.

FC:

OK. Two more questions: First, in the second half of the bull market, people may pay more attention to AI and MEME. From your perspective, what should we pay more attention to in the second half? Second, if we want to use data to trade better, what kind of content do you recommend we should read?

Jackson:

The first question is what to watch in the second half. I think there is a consensus with everyone that we have been discussing a theory recently. After I returned to Singapore recently, many founders sat together to discuss the party game. In particular, many Chinese founders felt that they could not join the party game, which was actually difficult for the entire startup. If you ask me what I will pay attention to in the second half from a trading perspective, I will still pay attention to some of the so-called party games that everyone has newly collected, such as Monad, and some AI-related infra that may be available in the future. I may pay more attention to it because I think this involves the issuance of new assets. At the same time, it is an ecosystem. The members of the party behind it should have a great motivation to make profits through other means besides directly selling coins. If there is normal input of external liquidity, such as interest rate cuts, then if the logic of this party game itself still works, I will pay more attention to the assets in this new party game.

The second question is how to look at it from the data point of view. I think it goes back to what I said before. First, determine what kind of trading style you have, or what kind of person you are. If you are an investor, then some text-based public opinion trend data should be the focus. If you are very confident in yourself and feel that you can become a degen and win in PVP, then what you need to focus on is how to find the so-called earliest alpha assets that you are concerned about. The mouse positions, based on the movements of the mouse positions, then determine your entry and exit points. If you are like me, you usually have a job and don’t have much time to pay attention to the entire market, pay attention to various data, and don’t have much time to do very in-depth analysis. Then you can look at some of the learns of the entire data, this information flow, and eventually be able to continuously simplify it through AI, and only pay attention to some abnormal signals, which may help your trading more.

FC:

Just now you mentioned party game. Actually, I have been thinking about how to sort out the difference between the so-called value token and MEME from a communication perspective. So I want to ask, who do you think is in the party game? What are its elements? Who is the organizer? Or who is the inviter? Who are the guests? What is the process like?

Jackson:

Didn’t we compare VC and MEME Coin before? Actually, we analyzed the data later and found the logic of party game behind it. Of course, some degens may have made money by following the party game. For example, according to the report released by GoPlus yesterday, only a dozen or so people made money in the entire meme ecosystem, and most people still lost money. This meme may have used Fair Launch, a method that makes you lose more comfortably, but the logic is essentially the same. So whether it is a party game of a large new ecological project or a party game of Meme Coin, its essential logic may be that there is a group of the earliest rat warehouses. This group of rat warehouses may have strong data analysis capabilities now, and cannot directly sell their assets in some more obvious ways. So what should they do? Suppose I am the leader of a large Layer 1 public chain. First of all, I cannot openly sell my tokens. I may have several ways to make a profit: First, I set up a fund to support some projects in the ecosystem. I can make a profit by selling the tokens of the ecosystem projects. Second, I continue to increase the use scenarios of native tokens. Then the share of the entire secondary market circulation will decrease. Then I will control the price of spot or contract, and there may be more ways. So how to create the use scenarios of native tokens? Either you create a lot of ecosystems, like the continuous nesting of dolls in the Terra ecosystem in the last cycle, and the leverage in the market is also very high. The native tokens are also staked in various ways. Or, like Solana now, maybe a few people will gather a new MEME and a new ecological project, such as an issuance platform. In these issuance platforms and DEX, you will find that the other side of its trading pair is always SOL. I think this logic is actually very obvious. After several projects that can go out of the circle are promoted, more people who want to play party games will come in and continue to gather new games. But no matter what kind of game we play, mainstream tokens may be part of a trading pair, so I can sell the coins in various ways, or I can do OTC. Recently, there is an ecosystem that has sold a lot of OTC.

FC:

I think what is the difference between Europe and the United States and Asia now? In fact, the money in Europe and the United States is relatively long-term, this is the truth. In 2017, the cycle of Asian funds was 6+6, that is, 6 months plus 6 months. The next round is 2+2? 1+1, 2+2, and now it is 2+3+n, but in fact, European and American funds are basically 4+4, right? Basically, they are all funds with an 8-year cycle. This means that everyone’s LP is different, and LP’s expectations and duration of returns are different. For example, a16z may invest tens of millions of dollars at a time, so how to sell it? I also heard an answer, that is, he may have an OTC team through Coinbase, for example, to sell it to family offices that may be longer-term, so I think they are quite mature in this path. But for Asia, in 2017, SOLANA and other projects were all roadshows in Beijing, and we even attended the meeting at that time. But now that there is no Chinese market, everyone has no such advantage. Actually, I think we can find another time to talk about this. Actually, the problem with Asian entrepreneurs now is that they can only keep launching new projects, because the original projects may eventually be listed on BN, and they don’t know what to do afterwards. It’s not necessarily that they don’t want to do it, but they don’t know how to do it better. Yesterday, I saw a tweet saying that the most important thing for Asian entrepreneurs now is to find foreigners, how to play with foreigners, or how to convince foreigners. I think it’s actually quite right, but it’s also quite disappointing. Because I recently used your product, I think it’s pretty good. I think the most important thing is to save time, this is the first. Second, it was actually quite troublesome for us to dig up the overall information of a team before, for example, we had to look at the root data and Twitter. The integration you did is actually very important. Third, when I don’t know what to do, I go to your place to check it out, and at least it makes me feel like I haven’t worked in vain today. Another thing is the MEME thing, you have a lot of trading signals that are helpful. So I think today I also thank you for helping me improve the data.