Strategy guidelines

The reference. Anatomy, indicator catalogue, sizing, time periods, and the mistakes that ruin good ideas.

Anatomy of a strategy

Every strategy needs four parts. Anything else is decoration.

Universe

The list of tickers. Five to fifty US stocks works best.

Bars

Daily, weekly, or monthly. No intraday.

Entry

When to buy. Indicator + filter + price condition.

Exit

When to sell. Target, stop, time limit, or signal flip.

Patterns that work

Most edges fall into one of these. Pick one. Don't mix.

Trend follow

Buy strength. Sell weakness. Sit out the rest.

Mean revert

Buy panics in strong names. Hold for a snap-back.

Breakout

Buy when price clears resistance. Stop just below.

Pairs / spread

Long the cheaper of two correlated names, short the other.

Indicator catalogue

Tenachine supports the common TA-Lib indicators. Full list at ta-lib.org/functions.

IndicatorWhen to use
SMA / EMATrend filter. 50/200-day are the classic.
RSIMean reversion. Cross 30 = oversold, cross 70 = overbought.
MACDTrend confirmation. Histogram crossing 0 is the trigger.
ATRVolatility-based stops and position sizing.
Bollinger BandsVolatility envelope. Touch the band, then revert.
DonchianChannel breakouts. The original turtle entry.
ADXTrend strength filter. Above 25 = something to trade.
StochasticMean reversion alternative to RSI.

Sizing

Default: risk 1% of equity per trade. The position size is calculated from your stop distance, so a wider stop means a smaller position.

If you don't set a stop, we use a 2×ATR trailing stop. If you don't set a risk budget, we cap each position at 10% of equity.

Time periods & data

  • Bars: daily, weekly, monthly. No intraday.
  • History: back to ~2010 for most US stocks.
  • Universe: US-listed common stocks. ETFs supported, options not.
  • Costs: zero commission and 1bp slippage by default.

Common mistakes

Look-ahead

Using data from the close of bar N to make a trade on bar N. We use bar N+1's open for fills.

Tiny universe

Backtesting on five tickers gives noisy results. Use at least twenty if you're claiming an edge.

No exit

Buy rules without a clear exit are not strategies. Add a target, stop, or time limit.

Curve fitting

If your indicator parameters look like a phone number, you've overfit. Test robustness by varying them ±20%.