HyperSwing

Momentum is an appealing starting point for systematic research because the underlying proposition is clear: a market’s recent direction may contain information about what follows. HyperSwing is my attempt to study that proposition in crypto while taking the conditions of a real strategy seriously.

The academic starting point includes Moskowitz, Ooi, and Pedersen’s work on time-series momentum, which studies persistence in an instrument’s own returns across futures markets. That gives me a reason to investigate. Applying the idea to crypto introduces a new set of questions about the assets, the holding period, and the way a position can be expressed.

One question is how much of an asset’s move belongs to the wider market. A collection of rising coins can look like many successful independent decisions while sharing a large exposure to Bitcoin and the same liquidity environment. HyperSwing therefore considers Bitcoin’s regime alongside the trend in individual assets. I want to understand when a signal adds information and when it mostly repeats a market-wide move.

Another question is which assets were available to trade at each point in history. Starting with today’s surviving assets creates a convenient view of the past. The names that disappeared, lost liquidity, or were not yet listed matter to the result. A historical study should reconstruct the opportunity set that existed then, including its less successful members.

The timing of a decision matters just as much. A signal formed from a completed period cannot receive a fill earlier in that period. Fees and funding change the economics of holding it, and turnover can consume a small advantage. I want those costs present while judging the idea, because they influence whether the idea is useful in the first place.

Position sizing is part of the thesis. The same directional view can produce very different outcomes when expressed through a quiet market or a volatile one. HyperSwing studies exposure in relation to volatility, with limits on individual positions. This makes it possible to examine the forecast and the risk taken to express it together.

The evaluation moves through successive historical windows, with research trials kept in the record. A strong aggregate result is only a starting point. I also want to know which market regimes produced it, how concentrated the gains were, and how much broad market exposure remained. An idea that worked during one favourable stretch deserves a different level of confidence from one that behaved consistently across changing conditions.

The project currently gives me a research and paper-trading framework for asking these questions. My aim is to understand the circumstances in which a simple trend idea remains useful. That means retaining the awkward parts of the evidence, from an expensive rebalance to a period when the market stops rewarding persistence.