The method.
What the desk is built on, in plain words. Three moves, four short parts.
- 1Watch a market on three clocks, and name its state.
- 2Judge every change by rules set before each run, inside its target state where it has one.
- 3Keep what survives, and let the line go on.
Markets are chaos.
A price is the sum of every reason anyone had to act, and most are never known. No model holds that in full, and Strat0 does not pretend to. What it holds is smaller and firmer: chaos has states, and most states last long enough to be read.
A state that has been confirmed can be named. Named, it can be judged in. Judged in, it can be built for. Everything on this desk follows from that one reading.
The watch.
The old stories gave the watch to a titan whose light saw the whole field at once. Strat0’s classifier reads a market on three clocks, hourly, four-hourly and daily, and names what it sees with one of eight states: how strongly it trends, crossed with how far it swings.
A state is written down only once it has held, and labels are computed without looking ahead. The watch makes no prediction. It says what has held, and when it turned.
The grid is the classifier’s own; where the clocks stand is the seeded run.
The watch, measured.
On Bitcoin’s 4-hour clock, every state’s typical stay was between one and six days.
| State | Time spent (4H) | Typical stay, days | Went next, most often (4H) | |
|---|---|---|---|---|
| 4H | 1D | |||
| Uptrend | 14.1% | 3.5 | 20 (21) | Quiet uptrend 44% |
| Quiet uptrend | 15.9% | 3.5 | 16 (26) | Sideways 48% |
| Volatile uptrend | 5.3% | 2.2 | 14 (18) | Uptrend 48% |
| Sideways | 33.0% | 5.4 | 29 (27) | Downtrend 33% |
| Sideways volatile | 1.3% | 1.0 | too few (7) | Sharp sell-off 39% |
| Downtrend | 24.1% | 5.2 | 34 (21) | Sideways 59% |
| Volatile downtrend | 1.6% | 1.5 | too few (4) | Downtrend 66% |
| Sharp sell-off | 4.7% | 2.8 | too few (7) | Downtrend 82% |
On 4-hour bars a state ended 78.6 times a year; about one in nine lasted a day or less (six bars).
The home page’s clock ladder uses the three-state summary label, so its medians differ.
The oracle.
At Delphi the oracle answered every question put to it, and the answer was final. Strat0’s oracle is a set of rules fixed before each run: hurdles on return and on drawdown, judged inside the state a strategy was built for.
Research agents propose changes. Six rules govern every judgement; three more come before any capital moves.
v22 · remove a filter · return clears, drawdown deeper · not kept
16 of 21 changes stopped
- 1One change at a time. Each change is of one named kind, so its effect can be read.
- 2Falsifiable in advance. Every change states, before it is tested, what result would prove it wrong.
- 3Judged where it should work. A change built for a market state is judged inside that state, and over its full history.
- 4Evidence is graded. Thin evidence is marked as thin in the record, and a result with no trades never passes.
- 5Five endings. A change is kept, kept and promoted, tried again from the same version, or archived with its reasons; a change that would make a different strategy stops for human review.
- 6Agents explain; code decides. They write the explanation of every verdict; they never decide pass or fail, and no single ratio does either.
- 7Costs are charged in paper trading. The backtest that judges a change charges only the commission its code declares. Paper trades then pay a fee, and the modelled ledger adds slippage and funding every eight hours.
- 8A held-out earlier span. A change that passes is run again on data from before the period it was built on, with its settings locked. If that run fails, it is not promoted. Where earlier data runs short, the run is skipped and the change is not held back.
- 9Rehearsal comes before money. No capital is designed to move until a Vault’s switching between strategies has been rehearsed on history and passed. To pass, it must beat both the best single strategy held throughout and holding the asset.
The hurdle values are not published.
The work.
The old work was to turn base matter into gold by patience and fire. Strat0’s work is to turn an idea into code, code into a strategy, and a strategy into one that survives its own history, one change at a time.
The next cycle is designed to read what failed before it proposes again. Volatility is then something a strategy was built for rather than something it endures. What survives becomes the line.
v20 · add a filter · kept
This is the whole of the method. Inside: the Desk, the full reference and a dated status.