2. Design assumptions

The engine rests on a small number of empirical premises about how markets behave. This section states each one, gives the evidence behind it, and describes what the engine does when the premise does not hold. That last part matters more than the first two: these are the assumptions that, if wrong, would invalidate the approach; a milder error in the others would only reduce its edge. A reader deciding whether to run the engine against real capital is entitled to know where that line sits.

2.1 Price oscillates at the ladder's operating timescale

The assumption. Returns at the interval the ladder trades tend to mean-revert. The ladder buys declines and sells them back one interval higher, so it needs price to revisit levels it has just left.

Evidence. Mean reversion is documented across equity and commodity markets (Poterba & Summers, 1988), and cyclical boom-and-correction patterns in many markets reinforce its viability as a trading basis (Urquhart, 2016). In digital assets, Baur, Hong, and Lee (2018) found that while short-term movements are highly volatile, longer horizons revert toward fair value, and quantified a daily-return autocorrelation of -0.12, direct evidence of reversal. The same short-horizon negative autocorrelation appears in liquid large-cap equities and heavily traded index products, making those the primary non-digital candidates.

If it fails. The ladder accumulates without realizing, and no amount of sizing discipline repairs that, because the oracle can only choose how much capital to commit, not manufacture reversals. This is the primary failure mode of the system and is covered in section 9.1. It is also why asset selection requires the oscillation to exist at the interval the oracle actually resolves, so the oracle derives spacing from measured history.

2.2 Volatility clusters

The assumption. Periods of high volatility are followed by periods of high volatility, and the same holds at the low end.

Evidence. Volatility clustering is one of the most robust stylized facts in empirical asset-return research (Cont, 2001): large price swings are followed by large swings, which raises both trading frequency and the profit available per cycle. That second effect is the one the oracle depends on: it makes recent measured behavior informative about near-term behavior. Backtests on digital assets report that grid strategies with optimized spacing and capital allocation outperformed trend-following over the majority of sampled periods, at a Sharpe ratio of 1.45 against 0.97 for a simple moving-average crossover, with a 28% reduction in maximum drawdown (Jia, 2023). A separate backtest reaches similar conclusions on Bitcoin (Zhou & Zhang, 2022). These are third-party backtest results; the period definitions behind them differ between studies, and none of them measures this engine.

Why the design depends on it. The oracle's whole premise is that recent measured behavior carries information about near-term behavior. If volatility were independent across time, sizing from recorded history would carry no signal, and the survivability model would be a more elaborate way of guessing. The search is only worthwhile because of clustering, which also lets the aggressiveness factor lean toward tighter spacing in volatile conditions without becoming reckless.

If it fails. Sizing carries less information than intended, and the engine drifts toward its conservative corner. The failure is gradual, which makes it the more manageable of the two.

2.3 Systematic execution beats discretionary judgment

The assumption. Rules applied without exception outperform judgment applied with discretion.

Evidence. This is the oldest and least technical of the four premises, and the least dependent on a specific result. It rests on the general finding that decisions made under uncertainty involve systematic biases (Kahneman & Tversky, 1979), which is the reason discretionary execution carries an error term a rule does not. Dio depends on this premise more strongly than most systems because it removes the operator from the loop entirely. There is no stop-loss, no manual override, and no discretionary exit path.

Where Dio sits among approaches. The engine belongs to an existing family of systematic strategies. Trend-following systems take directional positions from price trends (Carver, 2019); momentum systems buy recent winners and sell recent losers (Jegadeesh & Titman, 1993); pairs-trading systems exploit relative value between correlated instruments (Gatev et al., 2006). Dio sits among the mean-reverting liquidity-provision strategies: it holds fully-backed inventory, earns the spread between a buy and the offsetting sell, and takes no directional view. It is a different family, and its failure mode is accumulation.

If it fails. There is no partial failure available. Because the operator cannot override the engine without reintroducing the exact behavior the design removes, this premise is either load-bearing or the system should not be running.

2.4 Patient retention beats timing

The assumption. Holding an asset through its drawdowns outperforms repeatedly attempting to time entries and exits.

Evidence. Quality assets with durable fundamentals have appreciated across extended periods despite severe bear markets (Bianchi, 2020), and the same holding-period premium appears in broad equity indices, where patient accumulation beats repeated timing attempts (Bogle, 1999). Positioning orders at predefined intervals without directional prediction is long-established practice in high-volatility markets (Pardo, 2008), and grid and ladder systems are a standard expression of it.

Why the design depends on it. The accumulation mechanism compounds the traded asset out of realized trading gains, and the engine has no mechanism that realizes the position for a profit. reserved_base grows monotonically unless an asset permanently impairs. That is only a sound design under a retention premise.

If it fails. Accumulation concentrates losses in precisely the assets that fall hardest, because the reserve grows as prices decline. Fundamental screening is the only mitigation available, and no sizing model hedges permanent impairment. Section 9.1 covers this.

2.5 What the engine does not assume

The engine places no dependency on yield. It does not execute staking, dividends, coupons, or lending, and nothing in the sizing or execution path changes based on whether an asset pays one. Those instruments remain available to an operator on accumulated holdings, and are treated as a separate decision made outside the engine.

Yield is often assumed to belong to a carry or accumulation strategy; here it is a portfolio choice layered on top of the engine, and nothing in the engine depends on it.