A Few Responses Before the Capital Long March Goes Live
It's February 12, 2026, in the afternoon.
Yesterday, we achieved a very promising experimental result. The FMAB signal performed exceptionally well, and we are preparing to launch it live. After sharing this good news yesterday, we received some sincere questions and concerns.
- Returns in the thousands or tens of thousands of percent sound unreasonable. Are the returns overestimated?
- Has tail risk in real markets been overlooked? What specific mechanisms are in place to ensure risk control safety?
- The log articles seem a bit too scattered, mixing in many AI-related posts, making it difficult to follow. Could they be written more focused?
I'd like to address these concerns.
Regarding Excessively High Returns
As the "Impossible Trinity" states, high returns, low risk, and high capacity cannot be achieved simultaneously.
We are not violating this principle. The premise for our high returns is accepting higher risk. Although I have mentioned multiple times in the logs how I manage risk, this does not mean risk is completely eliminated. High returns are fundamentally backed by high risk. We also do not intend to pursue low-capacity strategies; we have already achieved that with our arbitrage project. Arbitrage can offer high returns and low risk, but its capacity is very low, which is not what we are aiming for.
However, we can reconsider what constitutes "high risk." Which is riskier: "steadily losing money" or "an unstable return curve"?
This is a debatable point. My personal view is that steadily losing money represents a higher risk. It is a tangible, realized risk—not something that might or might not happen, but a certainty.
Yet, interestingly, as humans, we eat every day, consume for survival, and have expenses. This "risk" is essentially steady monetary loss, yet people accept it with peace of mind. Why?
The answer is expectation management. People accept the risk of steady loss through the psychological defense mechanism of expectation. They expect that through work, investment, etc., they can earn enough to cover their expenses, enabling survival and growth. When the highest risk is transformed into consumption, people can accept it.
Some might argue: No, risk is volatility. Steady loss doesn't mean high risk because volatility is low; it's stable, so how can it be called high risk? In that case, consider whether the "Impossible Trinity" of high returns, low risk, and high capacity remains so impossible under this definition. It clearly becomes less impossible.
In summary, the premise for our high returns is accepting high risk, but the key distinction is that this risk is an expected risk.
I once used skiing as an analogy. When skiing, you must orient your posture downhill, overcome fear, and avoid "shifting your center of gravity away" from the fall line, or you will fall. Shifting away is like a bodily instinct, akin to the instinct to fear losses. Only by fully accepting the expectation of loss can one achieve higher returns in investing. It's an interesting analogy.
In this framework, the Capital Long March is essentially a high-return, high-risk, high-capacity investment strategy.
The next question is: If it's about steady loss, which is riskier: losing 100 yuan per day or losing 1000 yuan per day? Obviously, losing 1000 yuan per day is riskier. Interestingly, we can adjust the level of risk by adjusting the rate of loss. Even at the highest risk tier, there are still degrees of risk.
Thus, we approach the truth about the returns: The returns in SandTable are actually relative to the invested cash flow, not relative to total assets.
Take SandTable's experiment with the FMAB signal as an example. The baseline signal's return is 100%, while the anti-Martingale betting strategy's return is 4600%. Both were achieved within the same time window (one year). For the former, you invest principal X upfront, wait a year, and end up with total assets of 2X. For the latter, you invest a daily cash flow Y. During operation, it continuously outputs cash flow Z through profit-taking, and upon profit-taking, all previously invested cash flows are recovered, returning the principal. The latter can achieve self-sustaining after a certain investment period. After one year, the latter achieved a 46x return on cash flow. The principal? It was recovered long ago.
The latter scenario more closely resembles real-world investment scenarios in business. Imagine opening a restaurant: you need to invest daily cash flow for ingredients, employee salaries, and other operating costs, while the restaurant also generates daily revenue. Over time, the restaurant's total assets grow, and this growth rate is the return you truly care about.
Regarding Tail Risk
Some are concerned about the inability to control tail risk during execution. I think this is an excellent question.
By definition, tail risk refers to unpredictable extreme events in system modeling. No matter how complex your model, you cannot predict the occurrence of tail risk. You can use stress tests to simulate more extreme market conditions, but you cannot predict when tail risk will occur. Is 3 sigma enough? 5 sigma? 7 sigma? You can never be certain.
The real method to guard against tail risk is through zero-liability assurance. For example, on HyperLiquid, accounts are non-custodial and anonymous. No one knows who owns a particular account, so no one can pursue debt against a specific account. Thus, even if a tail risk event occurs and an account goes to zero, it does not affect the safety of funds in other accounts. This is the true foundation of the risk control line in the Capital Long March.
Preventing tail risk is more of a concern for exchanges. They handle basic liquidation issues through forced liquidation mechanisms. In extreme cases, they use Risk Reserve Funds and Auto De-Leveraging (ADL) mechanisms to protect the overall financial safety of the exchange from tail risk events.
In short, the exchange will help zero out your liabilities; at worst, your account goes to zero, but you won't owe money to the exchange. Of course, some traditional exchanges might have debt recovery mechanisms. For instance, commodity futures exchanges, which are not 24/7, might face gap risk overnight, potentially leading to debt recovery situations. In such cases, additional safety margins are necessary.
Consider historical events like short squeezes in futures markets, the "Crude Oil Treasure" incident, and the famous LTCM collapse. When the safety of larger systemic levels is threatened, policy interventions in the market can lead to extreme events.
That is to say, zero-liability assurance remains, to this day, only a locally valid assumption. But there's no need for excessive pessimism. With the trend of Real World Assets (RWA), more assets are being introduced into the DeFi ecosystem, making the zero-liability assumption increasingly realistic.
Often, it's due to limited liquidation capacity and the absence of legitimate risk reserve funds that debt responsibility shifts to investors, leading to debt recovery mechanisms. The stronger the liquidation capacity, the less necessary debt recovery becomes. Crypto, with its unique characteristic of continuous trading, has developed very robust liquidation capabilities. For example, during the recent major crash on October 11, 2025, over $19 billion in liquidations occurred. The liquidation capacity of Crypto exchanges was well tested; no debt recovery situations occurred—the worst outcome was accounts going to zero—and no exchanges collapsed due to the event.
Regarding Article Focus
Since I am writing logs, many experimental results are iterated and adjusted during the process, leading to some fragmented content.
Naturally, following the logs involves sifting through a lot of noise. Read the fresh updates when you have time; if not, wait for the later summary and overview articles. I will have AI periodically help summarize project-specific overview articles for easier reading.
