← Shareholder letters
Shareholder letter 06 June 30, 2026 · 2026年6月30日

What We Refused to Trade

第 6 封:我們拒絕交易的東西

Floating-rate transmission and the cost of refused exposure.
Floating-rate transmission and the cost of refused exposure.
浮動利率傳導模型與拒絕曝險的成本。
浮動利率傳導模型與拒絕曝險的成本。

Dear Shareholders,

致股東:

Investment letters usually discuss owned positions. This one discusses rejected trades. The discipline to say no is part of portfolio construction because every rejected exposure preserves capital, attention, and operational capacity for a better opportunity.

投資信通常討論已持有的部位。本封信討論被拒絕的交易。說不的紀律本身就是投資組合建構的一部分,因為每一項被拒絕的曝險,都保留了資本、注意力與操作能力,以等待更佳機會。

Our framework requires three permissions before risk can be added. The market must be observable enough to identify the active driver. The instrument must be executable under realistic liquidity. The loss must remain inside a predefined budget if our view is wrong or merely early. Failure on any one dimension is sufficient to reject the trade.

我們的框架要求新增風險之前取得三項許可。第一,市場必須具有足夠可觀察性,使我們能辨識當期主導因素。第二,工具必須在現實流動性下可執行。第三,如果觀點錯誤或只是過早,損失仍必須留在預先設定的預算內。任何一項不符合,就足以拒絕交易。

Our market thinking identified several seductive but incomplete propositions. We refused to buy an asset solely because it had fallen sharply. We refused to short a crowded theme solely because valuation looked high. We refused to treat AI capital expenditure as a straight line from spending to shareholder return. We refused to assume that long government bonds would hedge every equity shock. We also refused to equate state affiliation, investment-grade ratings, or benchmark inclusion with the absence of credit risk.

本期市場思考辨識出數個誘人但不完整的命題。我們拒絕只因某項資產大幅下跌就買入,也拒絕只因估值偏高就放空擁擠主題。我們不把 AI 資本支出視為從支出到股東回報的直線,也不假設長天期政府債券能避險每一種股票衝擊。我們同樣拒絕把國家背景、投資級評等或指數納入,等同於不存在信用風險。

Our trading action was selective inactivity. We monitored opportunities across short-duration credit, rate curves, energy-linked equities, infrastructure, and volatility, but we did not force a position when catalyst, valuation, and execution were misaligned. In several cases, the expected return looked attractive until we included bid-ask cost, adverse gap risk, financing, and the possibility that correlation would change in stress. The adjusted opportunity was no longer compelling.

交易行動是選擇性的不作為。我們持續監察短久期信用、利率曲線、能源相關股票、基礎設施與波動率機會,但當催化因素、估值與執行條件不一致時,我們不會強迫建立部位。在數個案例中,預期回報最初看似吸引,但加入買賣價差、跳空風險、融資成本與壓力期間相關性改變後,調整後的機會已不再突出。

The position for the year remains +0.00005 unit profit. Part of that outcome reflects what we did not own. Non-events cannot be marked as profits, and we will not claim avoided losses as realized alpha. They belong in process review because restraint is only valuable if it is based on a repeatable rule rather than fear.

本年度部位仍為 +0.00005 單位正收益,其中一部分與我們沒有持有什麼有關。沒有發生的事件不能被計價成利潤,我們也不會把避免的損失宣稱為已實現超額收益。但它們仍必須進入流程檢討,因為克制只有在源自可重複規則,而非恐懼時才有價值。

The lesson learned is that professional skepticism must apply equally to bullish and bearish narratives. A richly valued asset can keep rising if earnings and flows improve. A cheap asset can become cheaper if refinancing or governance deteriorates. The direction of our emotional reaction to price is not evidence.

本期教訓是,專業懷疑必須同時用於多頭與空頭敘事。高估值資產如果獲利與資金流改善,仍可能繼續上升。便宜資產如果再融資或治理惡化,也可能變得更便宜。我們對價格的情緒反應方向,不是證據。

A rejected trade has two possible costs. The first is the avoided loss if the feared risk occurs. The second is the missed gain if the market advances without us. We review both, but we do not evaluate a rejection solely from the subsequent price direction. A disciplined rejection may still be correct if the trade violated liquidity or loss constraints, just as a profitable gamble can remain a poor fiduciary decision.

被拒絕的交易存在兩種可能成本。第一種是如果擔心的風險發生,所避免的損失;第二種是如果市場在沒有我們參與下上漲,所錯過的收益。我們會檢討兩者,但不會只根據後續價格方向判斷拒絕是否正確。即使價格後來上漲,只要交易違反流動性或損失限制,有紀律的拒絕仍可能是正確決策;同樣地,一次獲利的賭博仍可能是不良受託決策。

The rejection log will distinguish insufficient edge, inadequate liquidity, excessive downside, unclear catalyst, crowded positioning, operational limitation, compliance restriction, and portfolio duplication. This classification helps reveal whether we repeatedly avoid one category of risk because it is genuinely unattractive or because our research and execution capabilities remain underdeveloped.

拒絕紀錄會區分優勢不足、流動性不夠、下行過大、催化因素不清楚、倉位擁擠、操作限制、合規限制與投資組合重複。這種分類能揭露我們是否因某類風險真的缺乏吸引力而反覆避開,或其實是因研究與執行能力尚未成熟。

Patience is valuable only when the research process continues to work. Waiting without updating evidence is inactivity; waiting while improving the probability estimate, execution plan, and downside map is disciplined preparation. During periods of low exposure, we redirect capacity toward issuer work, liquidity mapping, scenario design, and the identification of instruments that can express the same thesis with a better loss profile.

只有在研究流程持續運作時,耐心才有價值。不更新證據的等待只是停滯;在等待期間改善機率估計、執行計畫與下行地圖,才是有紀律的準備。在低曝險期間,我們會把能力重新投入發行人研究、流動性地圖、情境設計,以及尋找能以更佳損失結構表達相同論點的工具。

A missed trade can reveal a capability gap. If an opportunity was rejected because reliable data, legal analysis, counterparty access, or execution infrastructure was unavailable, the lesson may be to build the missing capability rather than permanently avoid the market. However, capability investment is justified only when the opportunity set is recurring and the expected benefit exceeds the operational and governance burden.

錯過的交易可能揭露能力缺口。如果機會因缺乏可靠資料、法律分析、交易對手管道或執行基礎設施而被拒絕,教訓可能是建立缺少的能力,而不是永久避開該市場。但只有當機會集合會重複出現,而且預期利益高於操作與治理負擔時,能力投資才合理。

We will also track thesis half-life. Some ideas decay quickly after a catalyst, while others remain valid for several quarters. A rejected trade with a short half-life may never return at an attractive price. A structural thesis may offer repeated entries. Recognizing the difference helps allocate research urgency without converting urgency into undisciplined execution.

我們也會追蹤論點半衰期。有些想法在催化事件後快速衰減,有些則可維持數季。具有短半衰期的被拒絕交易,可能永遠不會以吸引價格再次出現;結構性論點則可能提供多次進場機會。辨識兩者差異,有助安排研究優先順序,同時避免把急迫性轉化為缺乏紀律的執行。

Our next action is to formalize a rejection log and review it at fixed intervals without hindsight editing. For each declined trade, we will record the missing condition, the price at rejection, the trigger that would justify reconsideration, and the subsequent path. Original assumptions will remain frozen, while later outcomes will be appended separately. This will preserve the integrity of the decision record, help us determine whether restraint protected capital or merely disguised hesitation, and allow us to distinguish good process from fortunate narrative reconstruction. We aim to improve both the quality of our trades and the quality of our refusals.

下一步,我們將正式建立拒絕紀錄,並定期檢討被拒絕交易,同時禁止使用事後資訊改寫原始紀錄。每一筆被拒絕的交易都會記錄缺少的條件、拒絕時價格、值得重新考慮的觸發點,以及後續路徑。原始假設會保持凍結,後續結果則另外附加。這能維持決策紀錄完整性,協助我們判斷克制究竟保護了資本,還是只掩飾猶豫,並區分良好流程與幸運的事後敘事重建。我們希望同時提高交易品質與拒絕品質。