The first 74 days, and what we still cannot claim

Dear Partners,

This is the first letter of a new series, and it is numbered one deliberately. We have several years of research behind us, but the account we now report on is a single account, running a single strategy. Everything else is what we tested and discarded along the way, and half of this letter is about that.

We begin there, because the result on its own reads as nothing more than two and a half good months.

We short ETH against our own ETH. The stack sits in ETH, the short is opened through an inverse (coin-margined) contract, and both the collateral and the profit are denominated in the coin itself. The account therefore compounds in ETH: what it earns does not need converting back, it is already in the asset you hold.

The bet is unglamorously simple: price overshoots to the upside, and it tends to come back. We capture a modest retracement and exit. This is not a secret formula, it is the most thoroughly documented class of signals in the industry, and we say so plainly: the edge is not in the idea, it is in our willingness to execute it without improvisation.

More important than what we do is the list of what we stopped doing.

The averaging-down pyramid. For a year we scaled in: price moves against us, we add lower and wait for the reversal. It sounds reasonable and it feels brave. Measured over nine years, at identical exposure, the single-entry structure returns 11.45× against 1.96× for the pyramid. The reason is unpleasant and obvious in hindsight: a pyramid adds size precisely when the position is wrong. Averaging down no longer exists in the strategy. It is not disabled by a flag, it has been removed as code.

The entry pattern. We used to read candles, hunt exhaustion formations, and take pride in it. We tested it: entry had stopped being a lever. The strategy now enters on a schedule and does not look at candle colour at all. It turns out the market is indifferent to how elegantly we justify our timing.

The second asset. We ran a symmetric leg on bitcoin. We put it through the same audit as ETH: 405 configurations, and the overfitting test returned p = 0.83, against 0.028 for ETH. The improvement did not hold; the attractive optimum proved to be an artefact of the dollar lens. The leg was closed on 1 August. We no longer run a bitcoin account, and that is a decision rather than a pause.

A fund that adds an instrument every quarter grows wider. We chose to grow deeper, and we concede this is the less photogenic option.

The most useful table we produced all summer. We decomposed the nine-year result into stages: how much comes from the shape of the trade itself, how much from the forced-exit rule, how much from the regime filter. All in coin terms, all on ETH.

Read the second row carefully. Our signal, on its own, is worth almost nothing. The entire value of the structure lives in the circuit breaker (×5.29) and the regime filter (×1.48).

We could have left this unpublished. We publish it because it dictates the central rule of our work: if nearly all the value sits in one component, all of the risk sits there too. That component once failed silently on us. It is now the first thing checked, every time, with tests written around it that fail if it dies.

⚠️ The breaker's own threshold has since been revised on a fresh grid with real funding costs, the figures in this table, the SPA/DSR tests, and the budget optimum still await a re-run on it. What changes is the ×5.29 itself; the direction of the finding, that essentially all the value and all the risk sit in the breaker, does not.

Mechanically, the circuit breaker is our equivalent of expiry. A sold option has a term and closes itself; a perpetual short has no term, so something must close it. For us that something is a rule: a position stuck in a confirmed uptrend for roughly a month is cut, and re-entry into that same trend is blocked. What this looks like in practice is in Part II: it has already fired once, and as of this letter we are watching a second position test the same rule again.

A strategy that shows good numbers on its own data is either an edge or a well-groomed coincidence, and the difference is only visible from the outside. So we looked from the outside.

We ran not our preferred configuration but all 315 that can be assembled from our parameters, and tested them with a method that penalises exactly that kind of search (Hansen's SPA): p = 0.028, clearing the 5% bar. A second, independent instrument (the Deflated Sharpe Ratio) returned 0.971. Two methods built on different mathematics agreed to three decimals.

Two things matter more than the number itself. Of the 315 configurations, 270 are positive and not one has a negative mean, that is a plateau, not a needle. And our live configuration ranks 59th, not first: we did not select the maximum. Had we done so, this letter would read better and the account would perform worse.

Separately, we measured rather than interpolated our position size: 62% of the stack short at full deployment. The cost is stated honestly, drawdown in coin terms rises from 28.6% to 34.8%. We capped ourselves at 79%, and that ceiling is set not by fear but by the width of the confidence interval around the optimum: we stop where we stop being confident that we understand what we are doing.

Our nine years are crypto, and they are our own data. So we took the Cboe BXMD index: a covered call on US equities going back to 1988, a different asset, a different era, but structurally the same payoff shape.

Agreement to three decimals on data we could not possibly have touched. Which means what we do is a property of the payoff class, not a quirk of our period. There is a flip side, and it is at the end of this letter.

All of the above converged on one configuration. We call it flat: a flat 3.5% target, one entry level, no averaging down, no pattern. Here is how it performed.

Starting capital: 50 ETH, funded on 1 June. There have been no additions, everything below was earned by the account and reinvested into itself.

The line that matters is the second, not the first. ETH fell 6% over these 74 days, and the stack of ETH in the account grew by 5%. For a holder of the coin, that is the entire point of the structure: it works in the currency in which you count your wealth, though the dollar figure still dipped slightly this time (−1.28%), because the coin gain did not fully offset the coin's own price decline.

Act one, June: the market fell, and we lived on it. ETH went from 2,004 to 1,519, the strategy entered on schedule and exited at target, eight consecutive trades, seven of them closed in under a day and a half. The stack reached 59.67 ETH. Weeks like this make the work look dull and pleasant.

Act two, 26 June, 22 July: the market turned, and we paid for it. Entry of 5,709 contracts at 1,543.16, and instead of retracing, ETH ran to 1,915. The position went 24.1% underwater. After 26 days the circuit breaker did precisely what it is written to do: it cut the position for a loss of −7.20 ETH. That single loss consumed roughly three quarters of the profit of the eight preceding trades.

Act three, 22 July, 9 August: about two and a half weeks of silence. And this is the most important thing that happened on the account this period. The strategy placed no trades at all. Not for lack of opportunities: roughly 70 entry slots came and went. Every one was blocked, the rule forbids opening a new fade inside the same confirmed uptrend that has just cost us money.

The rally we declined to fight kept running. The strategy returned to the market on 9 August, once the trend stopped confirming, re-entering with 6,248 contracts at 1,916.17, target 1,849.10. This position is open as of this letter, we do not know whether it closes at target or becomes a second forced exit, and we would have written this letter differently if we did.

Note the eighth row: that position sat underwater for nearly twenty days and went 17% against us before closing in profit. It came within a hair of becoming the second forced exit. We show it because the difference between patience and stubbornness is entirely invisible on a chart, in our work it is decided by a rule, not by how we feel that morning.

The formal win rate for the period is 89%. We are obliged to say this is the least useful metric in our reporting, and here is why: eight wins produced +9.62 ETH; one loss took −7.20. The average win is 1.2 ETH; the single loss is six times larger.

That is how this strategy is built: many small gains and rare large losses. We measured the shape of that distribution over nine years, the skew is negative, and it is a design feature, not a defect. Anyone selling you a "90% win rate" without immediately saying the next sentence is selling you half the information.

We would rather you heard the unpleasant part from us in advance than from the market on time.

We will lag in rising markets. From that same 38-year history of our strategy class: the worst shortfall against simply holding the market is −48%, the longest uninterrupted stretch of lagging is 14.6 years, and at the time of measurement it had not ended. From our own data: in up days we beat the market 27% of the time; in down days, 72%. The formulation we ask you to remember: we reduce your beta to roughly 0.8, and the entire excess is earned on declines. In a sustained bull market we will look foolish. That is the invoice for the structure, not a malfunction.

But our "lag" runs on different physics than the paper version. BXMD is cash-settled: a rising stock price means a real dollar payout on the sold call, unsoftened by anything. For us, a rising price shrinks the coin-denominated exposure of the short while simultaneously making the collateral worth more, the mechanism works FOR us structurally, and our "lag" is foregone upside measured in coin, not a realised cash loss. BXMD also has no "enough, we're out" rule: it is a passive index, selling a call every month regardless of regime. Our own worst stuck duration is 34 days, not 14.6 years, precisely because we have an age/trend circuit breaker, 14.6 years is likely the ceiling for a strategy with no exit, not a forecast for us. The price of our breaker is different: not "sit for years" but "sometimes cut at the wrong moment" (see the −7.202 ETH cut above). And the worst case for the two structures differs in kind, not just in size: the equivalent of our own mistake on paper, selling the covering stock while leaving the short call open, turns the position naked, with unlimited loss. The same class of mistake for us (a breach of short ≤ spot) produces a stepped, not a cliff-edge, loss of stack. Beta and capture match BXMD to three decimals, the tail risk does not, and in our favour.

Drawdowns will be deeper than the one in this letter. The 13.09% of this period is the mild case. Over nine years, at the current position size, we measured 34.8% in coin terms ⚠️ (this figure still awaits a re-run on the current breaker threshold and will most likely come back deeper), and we put the probability of ever losing 30% of the coin stack at roughly 20%. In dollar terms the figures are several times uglier, and that is equally true of simply holding ETH.

There is no stop-loss in the conventional sense. A position closes at target or by the breaker rule. This is an architectural decision rather than an omission, and it is precisely what requires capital to be held with a margin of safety and position size to stay away from the limit.

And 74 days is not a sample. Nine trades, one forced exit, one open position. The period demonstrates that the mechanism behaves as designed and says exactly nothing about the next quarter. We would be grateful if nobody, ourselves included, multiplied +5.25% by five.

Over the first reporting period the account grew from 50 to 52.62 ETH (+5.25%) in a market that fell 6% over the same span. The strategy gave back most of what it had earned in a single forced exit and then refused to enter the market for about two and a half weeks, both times this was the same rule, and both times it was right, and as of this letter it holds a second position open, testing that same rule again.

Our next letter will not present a brilliant new idea. It will most likely present the same one, tested once more, alongside a list of what we tried and discarded during the quarter. If that sounds dull, it is precisely the impression we are working towards.

Thank you for your confidence. We are glad to answer any questions in person, including the awkward ones, for which we are usually best prepared.

Yours faithfully, Isochrons Capital

The circuit breaker cost us 7.20 ETH this period and saved us an unknown amount. The first is measurable and the second is not, which is the whole difficulty of explaining our work.

"Two and a half weeks without a single trade" is either discipline or idleness. This letter is not authorised to make that distinction, and neither is the rule that produced it.

We are better at finding our mistakes than at making them. So far.

Past performance does not guarantee future results. Returns are calculated in the contract currency (ETH); the dollar-denominated result for the period differs and is stated separately. Single-period figures are not suitable for extrapolation.