Machine Learning OSC [ZynAlgo]
Stop re-optimising your oscillator every few weeks. Machine Learning OSC scores dozens of parameter combinations against recent market behaviour and runs the one that fits current conditions.
Key benefits
- Self-tuning: periods, smoothing and thresholds are selected by score, not set once and forgotten.
- Trend protection: a hidden structural filter blocks crossings that point against the wider market.
- Clean readings: raw oscillator data is smoothed before anything is evaluated.
How it works It sweeps a grid of settings across a lookback window, scores each on what price did next, promotes the best one, then marks threshold crossings that agree with the trend filter. The full walkthrough is in the Manual.
What's included / Platforms NinjaTrader 8 indicator, lifetime or membership access.
Best for Traders running one oscillator across several instruments or regimes who want the settings maintained automatically.
Get started See it on zynassets.com/marketplace.
Disclaimer: Trading involves risk. Adaptive selection changes with new data and does not guarantee future results.
Machine Learning OSC [ZynAlgo]
What Machine Learning OSC does differently
Every oscillator has a settings problem. A 14-period lookback with thresholds at 70 and 30 works acceptably in some conditions and poorly in others, and the market never tells you which one you are in.
This tool does not use fixed settings. It continuously tests a grid of parameter combinations across a recent window, scores each one on how the market actually behaved afterwards, and promotes the best-scoring combination to be the live configuration.
You also choose which family of oscillator it adapts: a standard momentum reading, a volume-weighted variant, or a deviation-based one. Whichever you pick, the raw output is passed through a smoothing filter first, so the line you see has no jagged noise in it. Signals appear as markers on the price chart while the line stays in its own pane, on NinjaTrader 8.
Inside Machine Learning OSC: the adaptive engine
The word adaptive gets used loosely. Here it means something specific: the thresholds and the period are not the ones you set, they are the ones that scored best recently.

- The active line is the smoothed output of the currently selected configuration.
- The overbought and oversold bands belong to that configuration, which is why they are not always at the same place.
- The state tab reports the active oscillator and the market structure the engine has classified: trending, choppy or noisy.
- That classification is a by-product of the tuning. If the engine has settled on short periods and narrow bands, the market is behaving one way; long periods and wide bands mean another.
How the configuration is selected

- The engine sweeps a grid built from several periods, several smoothing amounts and several threshold pairs, giving dozens of combinations.
- Each is scored on forward returns within the lookback window, so the measure is what happened next, not how pretty the line looked.
- A switching margin is required before the active configuration changes. A new combination has to be meaningfully better, not marginally better, which stops the settings flipping every bar.
The honest consequence: the active configuration can change as new data arrives, and a marker that would have appeared under yesterday's configuration may not appear under today's.
When to act, and when not to

A marker requires three conditions together:
- The active line crosses its active threshold, upward through oversold for a buy, downward through overbought for a sell.
- The hidden trend filter agrees with that direction. This filter is a depth-configurable structural read that is deliberately not drawn on the chart.
- The spacing rule has cleared, meaning enough bars have passed since the last marker in the same direction.
The trend filter is the one that surprises people. You will see the line cross a threshold with no marker printed, and the reason is almost always that the crossing pointed against the wider structure. That is the tool refusing to catch a falling market, which is the behaviour it was built for.
Use the line and its bands to read pressure, but take only the markers. The crossings without markers are the ones the filter already rejected.
Who benefits most from Machine Learning OSC
- Traders who keep re-optimising oscillator settings and want that work done continuously instead of by hand.
- Anyone trading several instruments where one fixed setting was never going to fit all of them.
- Traders who want an oscillator that refuses counter-trend signals rather than one that fires at every extreme.
An oscillator that changes its own thresholds is only useful if it also knows when to stay silent. Most of the value here is in the crossings that never become markers.
Browse the full ZynAlgo catalogue at zynassets.com/marketplace.
Trading involves risk. The adaptive selection changes as new data arrives, and markers are analytical events that can fail in strong or noisy markets.
- Machine Learning OSC [ZynAlgo] v1Jul 17, 2026
Initial release


















