gm from the #2 on the @Lighter_xyz leaderboard. On Saturday I traded more than $90M (total exchange volume was around $300M that day) by manually PvPing several bots. As a hook for the reader: This all was done clicking buttons manually (no API/algo trading at all, so this is a bit off topic from our regular posts about algo trading) from a couch while watching TV. Since this caught quite some interest (and the alpha is gone now), allow me to make some corrections to the quoted tweet and tell the story of how I went from 10 points to rank 2 (400+ points) within 24 hours. Late Friday evening, one of our @PlanemoTrading order book alerts started firing every few minutes, most often for QQQ and SPY. Basically our system identified an operator that was blindly crossing the spread on those tickers every time liquidity was posted within 1 tick, so for instance a buy order at 775.98 and a sell order at 775.99 would be instantly taken by this account, therefore instantly realizing a trading loss (more on that later). Quick background on how we found this in the first place: We run (cross-venue) statistical arbitrage, most visibly across the HIP-3 deployments on Hyperliquid: same ticker, different venues, and we take the spread when it exists. When the @RobinhoodCrypto instance launched on Lighter we were integrated within hours and traded it for a few days, but the lack of liquidity didn't make it worthwhile and we stopped. However, the connection/websocket streams stayed live for the past month and ran through our regular system which includes something like an "anomaly layer". Put simply, it watches order book behavior on everything we touch and flags anything structurally odd for manual review, so we can decide if we have to adapt our logic. We get these flags constantly (we are now already past 1,000 individual orderbook streams) and most of these flags are noise, depending on how sensitive the filters are set. This one on Friday night however repeated every couple of minutes, in the same two books, and for hours at a time. So Friday night I tested it by hand. Posted two sided liquidity one tick apart, and it got taken almost immediately for a tiny profit. Saturday morning the alerts had run all night, faster and in bigger size. So I sat down on the couch and started posting liquidity manually. To be clear about the level of "sophistication" involved: This part was one guy (me) typing orders on the Lighter frontend UI while watching TV. The clever part had already happened when the system flagged the pattern. (To be fair, you could have easily spotted this behavior manually too by just looking at the orderbook for a few minutes fwiw.) Early on, orders sat in the book for minutes and got nibbled in bites of 3 to 5 units, a few thousand dollars at a time. By midday the counterparty woke up properly. Fifty unit clips, about $35K a side, gone within a second of posting. Then 100s. Then 200s. At some point I was resting 500 to 700 units per side (so up to $500,000) and watching it clear in seconds, a few dollars of spread each cycle, over and over. Two or three hours in, company finally arrived. Top of book size (in units) that had been mostly mine became 1,000, then 1,500, sometimes north of 2,000 (roughly $1.5M deep top level) as other makers piled into the same trade. For a few hours it was pretty fun when there were multiple people constantly cancelling and requoting to make sure they are in the first slot top of book. New takers showed up too. The original one always swept both sides in the same second. Others went sequentially, one leg, then the other, 2 to 3 seconds apart. I didn't check the chain in detail, but I think there were at least three distinct takers absorbing everything posted based on the different manners of execution and clip sizes. So, why would anyone blindly cross spreads all day? We can only guess at the motive, but the math points somewhere: at 0.1bps spread and 0 taker fees, the cost is about $10 per $1M of volume, and that Saturday $1M of taker volume (free account) earned around 6 points. So <$2 per point, even without knowing details about the current campaign, is probably a good deal. Of course its their trade, their economics. Ours was simply earning the spread on huge size, almost (!) risk-free (more on that soon). A few hours in is when I realized that I was suddenly number 4 on the leaderboard (lol). By end of day: roughly $90M traded, $2K net profit, around 400 points. Rank zero to rank two in under a day. Late Saturday I did another 10M or so of volume that, it turns out, earned exactly zero points. The program had changed at some point in the afternoon and I hadn't noticed. I of course kept quoting anyway since the PnL trade was the goal. Now, about "wash trading" from the quoted tweet. Wash trading means you control both sides of a trade (resting maker orders and being the taker), so that beneficial ownership never actually moves. That's zero risk, fake volume, and of course should always be filtered out/banned. What I've described is the opposite, and the roughly 5% of the time it went wrong (resting orders not immediately being filled) clearly shows it. Most of the quotes got taken on both sides immediately. The rest of the time, usually against the sequential taker bot when the price ticked more than usual, only one leg filled. In that moment you are naked long or short a single leg of $300K to $500K on a $20K account. It happened a couple of times an hour, and at least five times I sat properly holding the bag for minutes before I could get it flat. On a weekend. Anyone who has traded on high leverage knows exactly what that feels like. One geopolitical headline and a 5% move and that account is zeroed out in a second. This was deliberately degen with a small test account and I was fully aware of the risks of an unhedged position of that size. I have traded perps for the better part of a decade and this trading behavior is like the famous turkey happiness graph. This will go wrong eventually. I do not recommend anyone doing this, you will get rekt eventually. Wash traders do not get left holding an unhedged half million dollar leg on a trade they control both sides of. There's a structural/technical point to make too. On a busy book like Lighter you can't even guarantee your own maker gets matched against your own taker when trading on the UI. Anyone who truly wanted to wash trade there would have to build it algorithmically, precisely timed, purpose built to self match. Since Sunday the books on QQQ/SPY have changed. Spreads on those markets are far wider, and the unit economics moved with them: roughly 2 points per $1M of maker volume now (only premium accounts), against about $110 of maker fees on that same volume. Assuming break even trading, that is real money per point ($55 per point), for rewards nobody knows the value of. Whatever one thinks of the program changes, and I don't have a settled view yet, the window that made Saturday work is closed. So that was my weekend. PVP happens on every exchange with every single trade (by definition), this was just a very visible one.

New week, new milestone unlocked. Let's see what we can achieve for the rest of the month. Get started here: ondo.planemotrading.xyz/

👀 Q4

With our upcoming new UI release, we are going to open-source parts of the underlying tech we use for our algos. There are now many funding rate arb tools on the market, some also include "price spreads" across exchanges, but none considers the structural price spread that often persists (think kimchi premium). A price spread is completely irrelevant if its constant and you can't realize it (you close with the same structural difference that you opened with). With our stat arb algo we already automatically only enter when there is a severe dislocation that is NOT normal, and therefore can profit from it. In our new quant tab, you will see (green dots in the screenshot) when the price spread is actually executable in your favor. Still many nuances that we want to implement (maker/taker execution, actual depth of book, etc.), but the end result should be something that is actually valuable information and not a useless "price spread". It's still extremely difficult to actually pull this off manually and for best execution it should be automated (just use our algos), but if you want to manually execute trades, you can use this to at least check if right now is a good/bad/neutral time to do it.

This alpha is still alive right now and can be executed just by clicking buttons on Hyperliquid. NFA and there is no free money in perps and everything carries risk. We've been busy this morning. Across 313 spot pairs and 526 perp listings, there is currently only one basis trade executable on HL. ethereum:0x31eea0edeb85b0cb65c1400bd2f4e487fd61b537's perp listing has been trading up to 2.5% higher than its spot equivalent over the last few hours, and we have been busy building a position here. While surely only temporary, shorting the perp currently pays 2% daily in funding on top. We are happy with the accumulated position for now, so sharing this here for everyone. Biggest risk is a liquidation wick on the perps side, so make sure you keep an eye on things and know what you are doing. We'll have this trade across many exchanges and fully automated soon on @PlanemoTrading.

Fully exited at reversal. ~$1,500 profit in <36 hours Anyone else did this trade?

Buying the cheaper perp and selling the more expensive one doesn’t mean you’ve “locked in a profit.” Today’s big @Ondo news has sent ethereum:0xfaba6f8e4a5e8ab82f62fe7c39859fa577269be3 funding rates all over the place across perp DEXs. A good moment to show one of the less obvious traps in funding arbitrage, and a preview of the terminal coming to our new website in Q4. On paper, the trade is straightforward: long on one exchange, short on another, collect the funding difference. But what counts as a good entry? In the screenshot below, the thin green line shows that Extended has typically traded about 10bps below Hyperliquid over this window. So buying on Extended 10bps cheaper than you sell on Hyperliquid is a pretty ordinary entry. You’re getting the usual discount (aka 0). If that gap is still 10bps when you close, there’s no price spread profit to collect. And you’ve paid fees to get in and out. You can absolutely make money from funding with that entry. But if funding flips or you get a bad fill, there’s less room for things to go wrong. This is why the baseline matters. A persistent gap between two exchanges can look like an opportunity every time you open the chart. The useful question is: how far is the spread from its usual range, and how much of that deviation can you actually capture after fees and slippage? “Actually” is key: The price available for a tiny order may not be available at your size. Our statistical arbitrage strategies only enter on executable price dislocations. We want the opportunity to earn both funding and spread convergence, with more room to manage an early exit if the funding trade stops making sense. The new scanner will let you look for these setups manually. Some dislocations persist for several minutes, so you don’t always need to be the fastest bot to participate (you can just manually click). We’ll also offer historical data, because it helps to know whether a pair is worth watching in the first place. Some tickers trade within such a tight range across two venues that waiting for a meaningful dislocation can be a very long exercise. Funding arbitrage teaches you a lot of these details the first time you put a trade on. One worth learning beforehand: the spread you see isn’t necessarily the spread you get to keep.

Just beautiful to live observe a trade open (lowest 2 rows) and then close within 10 minutes, with each closing creating net PnL. This one here printed $2 in net PnL on $12,000 volume (or 1.67bps measured against volume). All within 10 minutes and 0 manual input.

Best epoch in terms of points since early February on @extendedapp. If you haven't tried it yet: Check out our Extended algos which include a mixture of maker and taker strategies. Hyperliquid copytrading available, too! You can also unlock access to our delta-neutral strategy with @Lighter_xyz after $3M volume. extended.planemotrading.xyz/

gm from the #2 on the @Lighter_xyz leaderboard. On Saturday I traded more than $90M (total exchange volume was around $300M that day) by manually PvPing several bots. As a hook for the reader: This all was done clicking buttons manually (no API/algo trading at all, so this is a bit off topic from our regular posts about algo trading) from a couch while watching TV. Since this caught quite some interest (and the alpha is gone now), allow me to make some corrections to the quoted tweet and tell the story of how I went from 10 points to rank 2 (400+ points) within 24 hours. Late Friday evening, one of our @PlanemoTrading order book alerts started firing every few minutes, most often for QQQ and SPY. Basically our system identified an operator that was blindly crossing the spread on those tickers every time liquidity was posted within 1 tick, so for instance a buy order at 775.98 and a sell order at 775.99 would be instantly taken by this account, therefore instantly realizing a trading loss (more on that later). Quick background on how we found this in the first place: We run (cross-venue) statistical arbitrage, most visibly across the HIP-3 deployments on Hyperliquid: same ticker, different venues, and we take the spread when it exists. When the @RobinhoodCrypto instance launched on Lighter we were integrated within hours and traded it for a few days, but the lack of liquidity didn't make it worthwhile and we stopped. However, the connection/websocket streams stayed live for the past month and ran through our regular system which includes something like an "anomaly layer". Put simply, it watches order book behavior on everything we touch and flags anything structurally odd for manual review, so we can decide if we have to adapt our logic. We get these flags constantly (we are now already past 1,000 individual orderbook streams) and most of these flags are noise, depending on how sensitive the filters are set. This one on Friday night however repeated every couple of minutes, in the same two books, and for hours at a time. So Friday night I tested it by hand. Posted two sided liquidity one tick apart, and it got taken almost immediately for a tiny profit. Saturday morning the alerts had run all night, faster and in bigger size. So I sat down on the couch and started posting liquidity manually. To be clear about the level of "sophistication" involved: This part was one guy (me) typing orders on the Lighter frontend UI while watching TV. The clever part had already happened when the system flagged the pattern. (To be fair, you could have easily spotted this behavior manually too by just looking at the orderbook for a few minutes fwiw.) Early on, orders sat in the book for minutes and got nibbled in bites of 3 to 5 units, a few thousand dollars at a time. By midday the counterparty woke up properly. Fifty unit clips, about $35K a side, gone within a second of posting. Then 100s. Then 200s. At some point I was resting 500 to 700 units per side (so up to $500,000) and watching it clear in seconds, a few dollars of spread each cycle, over and over. Two or three hours in, company finally arrived. Top of book size (in units) that had been mostly mine became 1,000, then 1,500, sometimes north of 2,000 (roughly $1.5M deep top level) as other makers piled into the same trade. For a few hours it was pretty fun when there were multiple people constantly cancelling and requoting to make sure they are in the first slot top of book. New takers showed up too. The original one always swept both sides in the same second. Others went sequentially, one leg, then the other, 2 to 3 seconds apart. I didn't check the chain in detail, but I think there were at least three distinct takers absorbing everything posted based on the different manners of execution and clip sizes. So, why would anyone blindly cross spreads all day? We can only guess at the motive, but the math points somewhere: at 0.1bps spread and 0 taker fees, the cost is about $10 per $1M of volume, and that Saturday $1M of taker volume (free account) earned around 6 points. So <$2 per point, even without knowing details about the current campaign, is probably a good deal. Of course its their trade, their economics. Ours was simply earning the spread on huge size, almost (!) risk-free (more on that soon). A few hours in is when I realized that I was suddenly number 4 on the leaderboard (lol). By end of day: roughly $90M traded, $2K net profit, around 400 points. Rank zero to rank two in under a day. Late Saturday I did another 10M or so of volume that, it turns out, earned exactly zero points. The program had changed at some point in the afternoon and I hadn't noticed. I of course kept quoting anyway since the PnL trade was the goal. Now, about "wash trading" from the quoted tweet. Wash trading means you control both sides of a trade (resting maker orders and being the taker), so that beneficial ownership never actually moves. That's zero risk, fake volume, and of course should always be filtered out/banned. What I've described is the opposite, and the roughly 5% of the time it went wrong (resting orders not immediately being filled) clearly shows it. Most of the quotes got taken on both sides immediately. The rest of the time, usually against the sequential taker bot when the price ticked more than usual, only one leg filled. In that moment you are naked long or short a single leg of $300K to $500K on a $20K account. It happened a couple of times an hour, and at least five times I sat properly holding the bag for minutes before I could get it flat. On a weekend. Anyone who has traded on high leverage knows exactly what that feels like. One geopolitical headline and a 5% move and that account is zeroed out in a second. This was deliberately degen with a small test account and I was fully aware of the risks of an unhedged position of that size. I have traded perps for the better part of a decade and this trading behavior is like the famous turkey happiness graph. This will go wrong eventually. I do not recommend anyone doing this, you will get rekt eventually. Wash traders do not get left holding an unhedged half million dollar leg on a trade they control both sides of. There's a structural/technical point to make too. On a busy book like Lighter you can't even guarantee your own maker gets matched against your own taker when trading on the UI. Anyone who truly wanted to wash trade there would have to build it algorithmically, precisely timed, purpose built to self match. Since Sunday the books on QQQ/SPY have changed. Spreads on those markets are far wider, and the unit economics moved with them: roughly 2 points per $1M of maker volume now (only premium accounts), against about $110 of maker fees on that same volume. Assuming break even trading, that is real money per point ($55 per point), for rewards nobody knows the value of. Whatever one thinks of the program changes, and I don't have a settled view yet, the window that made Saturday work is closed. So that was my weekend. PVP happens on every exchange with every single trade (by definition), this was just a very visible one.

New venue, new integration. We are live on @OndoPerps. Our full algo suite, open from day 1: - orderbook imbalance (our signature strategy) - momentum edge - Hyperliquid copytrading - statistical arbitrage Week 1 users get: - 0 builder fees - 20% off trading fees via our referral - Ondo USDC rewards + Ondo points, paid weekly. More volume and longer holds = bigger share. Start here: ondo.planemotrading.xyz/ Every user runs on their own dedicated server instance, so first trades usually land within hours of signup. Strategy breakdowns below.👇

We don't need 50x leverage and margin calls in the middle of the night to feel something. What's exciting to us is a slow and steady, "up only" PnL graph, with consistency over months. Our stat arb upgrade seems to do its job, all users in the green today @CoinMarketMan. Coming to more exchanges very soon!

Last week we quietly shipped a major update to our flagship HIP-3 statistical arbitrage strategy. As a reminder: it arbitrages fully automated the same tickers across @tradexyz @Markets_xyz @entropyIO and @tradeparagon. Since activity on many non-xyz tickers is rather low, capacity has been very constrained since the @Dreamcash and @felixprotocol sunsets. Meaning that for the last few months, cannibalization between users was rather high. There is only so much mispriced liquidity on a book, and not all users can take it at the same time at the same price. The choice was to either spread out trading hours across the entire user base, i.e. every user was set live only for x% of the day, or have everyone active at the same time, and let it normalize over time (everyone gets some bad trades from time to time and it evens out long-term). We tried both approaches and weren't happy with the performance, so we made some major changes. The engine now also takes, with high priority, funding rate spreads into account (used to consider it before too, now much more though). Which means that on average positions will now be open much longer if you can earn good money on them via funding (instead of rotating in and out more often). The result will be a much more stable PNL graph, increased profitability at the cost of slightly less volume. Attached you see the performance of one of our test accounts over the last week since the change went live. In screenshot 2 you can see the strategy still exits in its trademark fashion of net profitability per close. And screenshot 3 shows that pairs now also accumulate significant PNL via funding. So, TLDR: Better PNL graph and more capacity (can onboard more users and support higher deposits).

At any given time, we have 3-5 "moonshot" algos running on various perp dexes (internal-only). It's usually some weird idea that seems to be worth exploring more and we just spin it up into a formalized strategy and let it run for a month or so with a small account and then forget all about it until checking in on it. Most fail, but sometimes you strike gold too. Time to deposit more now and increase order sizing.

Slow & steady, like clockwork. With no trading fee discount, profitability hovered around 1bps (measured against volume) this week, which might seem low to some, but it's not. If you check the biggest algorithmic accounts on Hyperliquid (billions in monthly volume), the range that most fall in is between 0.2bps and 2bps, and that already includes a sizeable fee discount from volume tiers and staked HYPE. We aren't at the billions per month level yet, but since we are only trading 4 markets with this strategy, very happy with the consistency and profitability.