Multi-timeframe stochastic voting as one mechanical rule

Independent lookbacks, not one blended line

A classic stochastic %K can be computed separately on several calendar or bar-count intervals and then reduced to a single pair of integer vote totals rather than one blended oscillator line. Each reading is a bounded %K from ordered closes versus a lookback high-low range, computed independently on that sampling interval.

Editorial: TradersWeek treats this construction as a committee of independent lookbacks. Each horizon casts one oversold or overbought vote, and the combined count is the object that can be written as a single mechanical procedure.

How each interval votes

One published implementation uses a daily chart and six rounded trading-day intervals of 63, 31, 15, 21, 10, and 5 bars, plus a same-length daily %K, as the seven voting series. The stochastic length used for every interval in that implementation is a single shared input with a default of 21.

Each interval contributes one buy vote when its %K is below an oversold threshold (default 20) and one sell vote when its %K is above an overbought threshold (default 80). A momentum strategy converts those multi-horizon %K readings into buy-pressure and sell-pressure counts and acts only when the counts meet explicit thresholds.

One mechanical long procedure

A mechanical trading system is a fully specified rule set that maps the vote counts into entries, pyramiding, and exits without discretionary override. A mechanical long procedure can be written as: enter when buying pressure is at least 5, add at the next open whenever selling pressure is greater than 0, and exit at the next open when selling pressure equals 0.

The same voting construction is not limited to the stochastic oscillator. Contributors note it can be applied to other oscillators that admit overbought and oversold readings, including RSI and ADX.

The sampled intervals are part of the rule

Changing the set of sampled intervals changes the vote series. One platform substituted 60-minute, daily, weekly, monthly, quarterly, and yearly 21-period readings because it lacked native bi-weekly and half-quarterly chart scales.

A quarter-quarterly interval is described as roughly 3.25 weekly bars or 16.25 daily bars, which one implementer argued may add little information beside bi-weekly (10 trading days) and monthly (22 trading days) scales.

An Excel reconstruction maps every horizon’s vote onto weekly bars so a monthly or quarterly vote is repeated on each week that belongs to that longer period.

Recodings need not match

Independent recodings of the same voting idea on the same symbol can produce buy and sell vote paths that are similar but not identical, because the functions that define each timeframe are not uniquely specified.

Editorial: if the procedure is to stay mechanical, the timeframe definitions belong in the written rule set with the vote thresholds.