Where and How Crypto Actually Trades: Straight From the Source | Kevin March - Jpeg Trading

That was such an interesting interesting chat. Um, we're doing a bit of a topic change over to Kevin March with who is a founder of JPEG trading. We're going to learn more about how crypto trades actually work. Round of applause. Thank you.

Thank you so much. It's hard to follow Caitlyn Long, right? She's a legend. Uh to get up here and talk about DeFi, which is where all of us want to play. Uh she she really deserves a shout out because she's fighting on the uh the inside of the system that we're all trying to uh convert.

So yeah, shout out Caitlyn Long. So we set about answering this question of where and how crypto actually trades. Everyone thinks that crypto is so transparent, that blockchain is so transparent. But the problem is there's so much data that we're working with that it's really uh it makes it very difficult to actually get insights out of it. Over the last week or so, we set out with about 300 questions that we drafted uh with data that we've aggregated across 30 different chains uh in 32 agentic research cycles, which I threw in just for fun.

Um, and this presentation is mostly drafted with the research that they put together. And the idea is that if we can if we can gro the entirety of the blockchain data, uh, what can we actually find in it? And there was actually some really interesting findings. So let's dive in. There are a few uh few notes on the methodology here before I uh before I take you actually into the insights.

Um, first and foremost, as I as I just described, uh, there's heavy use of of AI in the research that we did for the data itself. These slides I made by hand because, you know, I'm I'm old school. Um, but I don't want to take away from the fact that uh this data might not be perfect. We will clean this up and we'll post it publicly and if anyone wants to uh to engage with us and help take this research forward, uh you'll be able to find all the data posted to our Twitter. Uh, secondly, for data sources, we pulled a lot of this from Dune Analytics.

Um, part of it came from our own internal data as well. Uh, also Coin Market Cap, DeFi Llama, and a lot of the project documentation for a bunch of the uh, various projects that we're going to talk about. Uh, lastly, the data that we pulled for this is from approximately September 1st, 2025 to uh, today. Um, as of like 8 a.m.

this morning. Uh, so this data is very recent. Uh, now with that being said, uh, we're going to bounce around between a bunch of different insights. Our goal was not to cohesively answer the entire story of what actually drives the crypto markets because if anyone tries to answer that for you, they're probably wrong. Um, but our goal is to set people up where they can take things forward from what we've done so far.

Uh, so we're going to bounce around a little bit. It might feel a little bit disjointed. So, hold on to your britches cuz we are jumping in. Now, first and foremost, bot empires rule DeFi. We all probably intuitively know that.

But the question is how much and why. We set out actually to study really broad user behavior. Um but what we found is that there are few clusters of addresses that engage with so much DeFi and they touch so many assets across so many chains that it's almost guaranteed that if anyone in this room has ever used Devi, you've interacted directly with some of these addresses. So let's start with those first. Uh there are two addresses here.

the first of which is 0x632. So this address on on on the left um this is their their Kim live activity over the the time period that we measured. Um they're running a yield farming strategy. They're touching pretty much any decks in existence across uh I think 15 different chains. They are uh farming anything that has incentives that are pushed out by by any token issuer.

They're constantly rebalancing those pools. they're maximizing their ability to extract those incentives and it's turned them into single-handly one of the largest uh addresses that's engaged with DeFi over the last 6 months. They're also related to a second address 0x278 um which is running an HFT strategy that is rebalancing across uh a bunch of different chains for a multi-chain arbitrage strategy. In these two chains, a single entity are single-handedly 7% of DeFi over the last 6 months. There's a third address that we found though that's unrelated to those uh called 0x51C.

We'll post this on our Twitter. I'm not going to read it to you every time. U but they are similarly large. And these three addresses uh in in total pretty much dominate DeFi. any chain that you've interacted with, any decks, any LP pool, pretty much any asset.

Uh, as long as it's on EVM, it's been touched by these folks. This one is a little bit interesting though because they're relatively dominant on Uni Chain and Inkchain, uh, which I imagine probably not the entire crowd is they trade on those addresses, but these three combined actually represent more than 10% of the entire market. that last address that I was just talking about 0x51C. Uh this is a representation of their trading activity over time and the the same sort of macro events that drive the volatility that kind of all of us shy away from that causes uh a lot of people to stress out and and lose sleep over their their day trading activities. That's where this thing finds most of its profit.

So when the rest of us are are recoiling because crazy mo uh movements are happening in the market from the you know the US election cycle to Trump announcing tariffs that's when this address jumps in and from that these folks have have uh pretty much single-handedly cornered the the DeFi market. This the table somewhere is in but we can skip it. So next up, instead of looking at individual addresses, uh what we're looking at are there are there insights that we can pull out about individual protocols or patterns in in the uh usage activity around these protocols uh that can drive interesting discoveries that hopefully we can use to design better protocols on the future. First up is fluid. Uh fluid as a dex is probably largely unappreciated.

Fluid is is kind of had its moment over the last uh over the last six months or so. But fluid is predominantly preferred by the institutional users in crypto. Uh what that means is that their average transaction size is much much much larger than the average transaction size you see across the rest of crypto. Um you know pretty much like any chain you look at any DEX you look at fluid dominantly is the one that is most preferred by institutional users. What they have that drives this is their unique um debt liquidity model which I don't have time in this presentation to dive into exactly how fluid works but uh the usage data around them indicates that they're loved by institutions in a way that nearly no other DEX has has achieved not a drum not unis swap not meteora prediction no one the next step that we found is that activity on each chain is uh pretty consistent.

So in in the in the top left that you see here uh it's the whale versus retail uh volume share and what we're trying to indicate is over time how much activity is made up of certain users consistently. Uh the top left chart Ethereum pretty consistently represents 90% or the whale activity pretty much uh represents 90% of activity on Ethereum. The next up is Binance Smart Chain, which is a little bit more volatile. And then uh and then Arbitrage and Bass. Bass is actually really interesting.

So across each of these three charts, BAS is pretty much the only one uh that consistently has some sort of volatility to it. Um and really what that means is there was something that happened on base on approximately January 20th that fundamentally changed how people interacted on base. And really it was it was driven by an enormous amount of retail activity uh that that dwarfed the amount of of whale activity that happened on base on that day. Uh which unclear what exactly it was but largely each chain has its own personality. I think that's the most important takeaway that if a certain type of user is dominant on a chain, it's very likely that type of user is going to uh pretty much exclusively uh engage in that activity on a chain going forward.

Next up is the Salana token factory. Now, a lot of people have talked about the the memecoin era of on Salana and how it has uh has driven the the uh downchain activity for Salana for a very long time. I think the the surprising thing is that the Salana memecoin era is not over. In fact, it's better than ever. since January of this year, so year to date, there have been 1.

3 million new meme coins launched on on Salana. Um, which means that for every uh for for every memecoin, there are two active users on Salana. It's a 1 to2 ratio. Uh, so yeah. Yeah, thank thank you for that.

So, you know, me and Aquiver just laughed. You know, we can share our memecoin and then, you know, everyone else can uh can share theirs. It's still dominantly what drives activity on on Salana. Uh Pump Swap, so pumped out funds, uh new decks that they that they've launched is still dominantly made up of memecoin activity that that comes out of this. It's still the casino style distributions that that you'd expect.

So this this bar chart on the right is is uh showing the outcome of each meme coin that launches. So obviously 87% end up with a market cap of sub $1,000, 12%, 0.5%, 0.4% it it drops off dramatically. So, it's it's very it's very much uh lottery style distributions.

If you see a memecoin launch on on Salana, you actually have a better chance of winning the lottery in most US states than to pick a winner out of the the the memecoin madness that's happening on Salana right now. Despite that, it still drives the majority of onchain volume. Next up, we had a question which is if a very hyped project has a ton of activity on test net, what does that actually mean for the future usage of uh of of that project? Now, I I I think intuitively a lot of us could guess that that the answer is uh that it wouldn't necessarily lead to long-term usage of the project just because it has a lot of uh test net usage. Um but we we don't want to test it.

I redacted this because I don't want to publicly FUD the these projects quite yet until we uh until we validate this data further. Um but we pulled two projects. We pulled all test net activity for both of them and then we record those transactions over time. Um the project on the left uh what we found was that it built up to a crescendo obviously right before it launched to to mainet and the second it launched to mainet it was a uh I think 86% drop down um from test net usage to the production usage that that metriculated onto mainet hype test net 2 for a very similar project followed a very similar pattern except at least it had a a slower drop down uh from the peak into uh in into the the lows that that happened post mainet. Uh but still it was a 95% uh draw draw down from the peak test net activity.

In both of these cases there were no clear incentives that were being thrown out for test net usage. It was more so uh people speculating that test test net usage leads to to further incentives. Um but the but the answer is clear. If you see KPIs and metrics being ted as part of test net usage for a new project that's coming up that raised a bunch of venture capital, they've done all sorts of really impressive things, it doesn't necessarily metriculate into real onchain usage. We still have to look at the fundamentals that that come next.

The next is something that that we're coining ourselves. I think um and keep me honest here. If anyone's discovered this and they've talked about it, uh feel free to yell at me from the crowd and say, "Hey, Kevin, this this isn't necessarily original research." Uh but we're calling this the 11 trade rule. And the question is uh what is it that makes certain users more likely to continue using a D5 protocol um than others?

Retention is one of the biggest problems that most D5 protocols have. And the 11 trade rule is actually something we discovered that really helps with it. So what this chart is showing is the likelihood to continue using a protocol based on the early activity of a user on that protocol. Um, and the more a user uses a protocol in the first seven days of interacting with that protocol, the much the more likely they are to continue using that protocol long term. Um, so much so that after they've used a protocol 10 times in the early usage of that protocol, um, they are somewhere between like 3.

5 to 7x more likely to metriculate into being a long-term user. We saw this pattern actually across a huge number of of dexes and a huge number of D5 protocols. Here we only have two. Um, but this also showed the only protocol that actually broke this trend was uh Pancake Swap on Binance. Um, which I'm open to thoughts on on why they they broke that.

Um, but Pancake Swap has a lot of uh uh wallet rotation that happens for users. So it might mean that uh Pancake Swap has high retention. um or it might be very heavily bought it. Who knows? The last area that we're looking into is how does geography impact the trading trends?

And then the question everyone asks is like you know who's driving the market. Is it America? Is it uh you know is it China? Is it Singapore? Is it Korea?

Is it Europe? We try to break it down uh just broadly based into Asia, Europe, North Kore North America. Um what we found is that the the timing of the day matters tremendously. Um and certain activities in certain time periods are much more likely to be successful than others. It's probably no surprise that Asia does most of its trading from 9 to 2 p.

m. Singapore time. Europe follows a pretty similar pattern that America follows. What we found is that uh by using these patterns you can you can optimize the uh you can opt optimize uh strategies around when to launch products, when to launch tokens, when to launch marketing announcements based on the geography that you want to target. Um so projects that are targeting primarily uh US users from 1 to 5:00 p.

m. Eastern time happen to be the the the most important time that they can find on J. I'm running a little bit out of time here. So, let me wrap up with with my key takeaways. So, first is that the median trade size that that you see across protocols uh is very important.

The larger the median trade size, the more likely it is that the activity on that protocol is organic, the less likely it is that it's botted and the more likely it is that the the activity for the average retail user is going to be profitable. The second is the the 11 trade retention rule. um the more a user uses a protocol in the first seven days of of interacting with it, the more likely it is that they will turn into a permanent user. Third is that when in doubt, use Ethereum. Ethereum across all metrics that that we uh we're tracking uh has proven to be the least spotted and the most common source of organic activity.

And last but not least to end on a high note is that uh discount test net by 95%. is a general rule of thumb when you're trying to imagine like how much of this test net usage actually metriculates into real world usage. On that note, I'll wrap it up. Uh this is where you can find us, JPEG trading. My name is Kevin March.

That QR code uh takes all of your crypto. I'm kidding. It goes to our Twitter account where we will post this presentation and all of our data from it and anyone is more than welcome to come join us and uh comment on it. So, thank you very much.