HOW DOES PECCALA’S TRADING ENGINE WORK?
QUESTION — What is the investable universe (ie. what assets are you trading)?
We trade crypto derivatives, in particular perpetual futures. We currently trade the ~140 currency pairs available in Binance Futures that are settled in USDT.
QUESTION — What is the exposure to each instrument through time (top 5 if you are trading a lot of different things)?
Approximately 5% of the funds sit in a pool that we use to redeem tokens, and the remaining balance is actively traded. The situation is highly dynamic so there are no global top 5 instruments over time. However, at any given time, the PECH strategy allocates on average 28% of the capital to the top instrument. On rare occasions (~2% of the time) this allocation exceeds 80% of the capital. The PECM strategy is more conservative, and the top instrument uses on average 10% of the capital.
QUESTION — Where are the funds held?
The assets are held in our Binance Corporate account, with the funds of each token segregated.
ABOUT TRADING BOTS
QUESTION — How often do you place trades?
Each hour, the bots decide if a position is going to be opened, increased, decreased, or closed.
QUESTION — How much of the total AUM is being traded at any one time?
See the answer to question 2.
QUESTION — How does the trading engine make decisions?
Each bot has a unique set of parameters and internal rules, but basically, they all try to answer the following questions: Is there a trend? If so, what’s the probability of further trend amplification?
If the assessment is positive and the trend is upwards, the bot opens a long position. If the assessment is positive and the trend is downwards, the bot opens a short position. Upon uncertainty or a negative assessment, the bot does not use the capital and waits for the next assessment.
QUESTION — How do you decide which of the 1000s of bots are put in charge of the funds?
Once the bots are trained, we use independent data to test them. Instead of ranking them, we impose strict requirements (in terms of profit and risk management) on them before taking them to production. The bots that pass the battery of tests are then deployed to production.
QUESTION — Do you ever change the bots? If so how often?
Yes. Cryptomarkets are open 24/7, so literally, every second new data are created. Regularly we recalibrate the system, performing again the above battery of tests with fresh data.
QUESTION — Is there ever human intervention in the trading?
No. Humans are involved only in creating, evaluating, and implementing the strategies. No discretionary decision-making is involved in the trading. The aim is to guarantee that every single trading decision has been thoroughly backtested. Furthermore, we have a policy of zero emotions in the trading decision-making process.
QUESTION — What happens if bots make different decisions on an asset? E.g. what happens if 10 want to go long, 5 want to go short, and the rest don’t want to do anything in that position?
All bots have equal weight. Bots that disagree cancel out each other, and what we see on an hourly basis is the consensus-based output.
QUESTION — Does each bot always control an equal portion of the pot, or do they change based on performance?
They all have the same funds allowance. Market conditions are constantly changing, and what could have been a remarkable strategy yesterday, is not necessarily going to be even profitable tomorrow.
QUESTION — Will a bot have to stop investing in a high-performing position if it “uses up” its assigned capital?
Yes. There’s no spillover of the capital between bots, as we don’t want the global result to be biased towards a specific bot’s strategy.
QUESTION — If the bots pool their positions, how do you know which bots are performing best?
We log every single action and all the variables that led to the decision-making process itself. With this information, it’s easy to track the profit or loss of each bot.
QUESTION — Why is there always a dip after a peak?
There are different angles to this question. Fundamentally, I’d simply argue that realistically there’s no other way. A peak after a peak with no dip would mean infinite growth with zero risk, which would require some sort of crystal ball. One could ask then, why don’t you stop trading when you reach a peak? That is a very valid question. But first of all, a peak is only evident when we’re far past it, i.e., when we’ve lost some money, unless, again, we have a crystal ball.
We could then ask ourselves what happens if we temporarily stop the trading after we’re happy with the recent growth, not necessarily at a peak. Doing so will certainly work in the short term, but eventually, we’d miss a large swing that would wipe out all the small profits we made from overengineering the process. We’ve thoroughly backtested the latter, and there’s no straightforward way to be better off in the long run than just keeping the trading on. Intuitively it makes sense, as leaving the capital unexposed to risks prevents it also from making profits.
QUESTION — Why are drawdowns so high, especially on high risk?
The drawdowns are not only comparable to other crypto products but actually smaller. Bitcoin, one of the least volatile crypto assets, has a max drawdown of approximately 90 percent, whereas that of PECH and PECM are 65% and 23% respectively. In one of the answers above we mentioned imposing strict conditions on the bots in the testing phase before they are deployed. One of these conditions is that the drawdown should be less than other crypto products. We deliver crypto returns with controlled risk.
QUESTION — Why can’t the bots close positions faster at the top of the market, or as soon as they see the market is turning?
This is a question we get often. If markets were frictionless (no trading fees), we could theoretically turn 100 euros into a million in no time, precisely by adapting super fast and capitalizing from changes at small time scales. Real markets though are far from frictionless, and every decision has an associated cost. We need to see the market at the exact time scale that allows us to be agile, yet efficient in terms of fees.
Many human traders get ruined precisely by wanting to adapt too fast. We often see quick drops in a rather upward trend. Human traders may panic and change direction very quickly, only to realize that soon after they switched, the market continued its original trend.
QUESTION — Are you using leverage? What does that actually mean — are you borrowing funds to trade with?
Although leveraged positions in traditional assets might lead to a user losing more than the initial capital, crypto markets are very different in this regard. In crypto perpetual futures, positions get automatically closed (liquidated) when the balance approaches zero, and by construction, there is no way one can lose more than the original capital. Furthermore, thanks to our automated hourly rebalancing, we’ve never even been close to being liquidated in any of the hundreds of thousands of positions we’ve opened.
To further clarify, we do not borrow funds for our leveraged positions. Futures in the crypto world do not involve borrowed money, positions get liquidated automatically instead.
QUESTION — Can you end up owing more than you invested?
No, neither we as a company nor our individual users. The reason is outlined in the answer above.