Strategies
Paper strategies read the candles and emit order intents. Nothing is placed. They exist to make repeatable research probes — run the same rule over the same observed data and see how it behaves.
The roster
Conventional controls sit next to the house experiments so every result has something honest to measure against. The controls are well-known baselines; the experiments descend from the indicator lineage, and what each one computes stays under the hood. A rule that can't beat Buy&Hold on the same data hasn't earned attention.
| strategy | role |
|---|---|
| Buy&Hold | control — the baseline everything is measured against |
| Donchian | control — channel breakout |
| MaCross | control — moving-average crossover |
| LowVol | control — low-volatility allocation across majors |
| BandScale | house experiment |
| EmaCurve | house experiment |
| TrendSlope | house experiment |
| Terraced | house experiment |
| CurveScale | house experiment — dormant by design (5-minute only) |
| TBMC Alpha | flagship — the TBMC family in code |
| TBMC Beta | flagship — second TBMC variant |
What the leaderboard ranks
The leaderboard scores each paper strategy over a recent window by return, drawdown, and a Sharpe-style risk-adjusted number, defaulting to a Sharpe sort. The live pages let you rank the roster, follow one strategy family across pairs and timeframes, and mark on a chart where a rule would have acted.
Scores are comparisons inside one dataset over one window. A rolling risk-adjusted number thickens with weeks of observation — early readings are noisy by construction. Read the ranking as "which rule behaved better on what we saw," never as a forecast.
Reading a score without fooling yourself
The dataset is what synthCore observed, not the whole market. Paper fills assume the intent transacts at the bar; real slippage, fees, and depth are not modeled. Survivorship is bounded because the roster is fixed and public. None of this makes the numbers useless — it makes them what they are: a controlled comparison, useful for studying behavior, not a promise about money.
Back to synthCore docs · see also: Engine, Mint.