When technical indicators disagree

Disagreement is the normal state

The common assumption is that indicators mostly agree, and that conflict is a warning sign. The opposite is closer to the truth.

Technical indicators measure different properties of the same price series. Momentum, trend strength and volatility are genuinely different things, and there is no reason for them to point the same way. A momentum oscillator saying "oversold" while a trend measure says "strong downtrend" is not a contradiction — it is two accurate statements about different questions.

The practical problem is not that indicators disagree. It is that most people resolve the disagreement by choosing whichever one supports the position they already wanted, and do it inconsistently from one day to the next.

What each family actually measures

Read that table again with the failure column in mind: each family fails in the conditions where another family is most informative. Momentum breaks down in strong trends, which is exactly when trend measures are reliable. Trend measures break down in ranges, which is where momentum works well.

That is why using several is worthwhile, and it is also why they conflict. The conflict carries information about which regime you are in.

The four common conflicts

Oversold in a downtrend

An oscillator reads oversold; a trend measure reads strong and downward.

The oscillator is telling you the move is stretched. The trend measure is telling you that stretched conditions can persist. Historically this is one of the most expensive setups to trade on the oscillator alone, because "oversold" is not a reversal signal, it is a description of the current state.

Resolution: in a confirmed trend, the trend reading dominates. Treat the oscillator as a note about timing within the trend, not as a reason to fade it.

Breakout with no volume

Price clears a level; participation does not increase.

Resolution: volume is a confirming input, not a leading one. Absence of confirmation lowers confidence rather than reversing the direction. This is a case for a smaller published confidence, not a suppressed signal.

Volatility expanding, direction ambiguous

Bands widen, ATR rises, and no directional indicator has a clear view.

Resolution: this affects position sizing and stop placement, not direction. It is the clearest case where the correct output is a wider stop and a smaller size rather than a different call.

Momentum turning while trend is intact

The most genuinely ambiguous case, and the one where most systems overclaim.

Resolution: it is early. Either wait for trend confirmation or publish at low confidence and say so. Any system that reports high confidence here is telling you about its optimism rather than about the market.

Four ways to resolve conflict, ranked

1. Discretion. A person weighs the indicators case by case. Flexible, capable of incorporating context no system has, and completely inconsistent — the same setup gets read differently on a Tuesday than on a Friday, and it leaves no record.

2. Hierarchy. A fixed precedence order: trend beats momentum, always. Consistent and easy to explain. Rigid, and wrong in the regimes where the priority indicator is the one failing.

3. Voting. Count how many point each way. Simple and transparent, but treats all indicators as equally informative, which they are not, and double-counts correlated ones. Two momentum oscillators are close to one opinion cast twice.

4. Regime-conditional weighting. Determine the regime first — trending or ranging — then weight the indicators that are reliable in that regime. More complex, and the only approach that reflects how the indicators actually behave.

The fourth is the right answer, and the first is what most people do.

Turning conflict into a confidence score

The useful output of a multi-indicator system is not a binary call. It is a call plus an honest statement of how much the inputs agreed.

A workable structure:

1. Classify the regime first. Trending or ranging, from a trend-strength measure. Everything downstream depends on it.
2. Weight by regime. Trend-following inputs carry more weight in a trending regime, mean-reversion inputs in a ranging one.
3. Group correlated indicators so that two measures of the same property count once, not twice.
4. Score direction from the weighted inputs.
5. Score confidence from the degree of agreement, adjusted for regime certainty. Unanimous in a clear regime is high. Split in an ambiguous regime is low.
6. Apply a publication floor. Below a threshold, publish nothing. Silence is a legitimate output and most systems will not produce it.
7. Attach the reasoning. Which indicators supported it, which dissented, and what each read at the time.

Step 7 is what turns a signal into something a person can disagree with. A call with no visible rationale can only be accepted or ignored.

How PSX Invest handles it

PSX Invest runs seven checks per symbol across all four families — RSI, MACD, Bollinger Bands, ADX, ATR and volume — computed consistently across 500+ symbols after the session closes rather than assessed by eye.

Three design choices follow directly from the conflict problem:

• The indicator rationale travels with the call. You can see which inputs supported it and what each read, which means you can disagree with the reasoning rather than just the conclusion.
• A confidence score reflects agreement, and it is calibrated against real outcomes rather than asserted.
• A minimum 2:1 reward-to-risk applies to anything published, with the entry range, target and stop set before the call goes out. Risk is framed at publication, not after a position is open.

The system publishes a shortlist and stops. It places no orders, and every entry and exit is a manual decision by the trader.

The transferable point

This is a specific instance of a general pattern in decision systems: multiple imperfect signals, each reliable in some conditions and misleading in others, needing to be combined into one ranked output with an honest uncertainty attached.

The market case is unusually clean because outcomes resolve mechanically and quickly. Most domains have the same structure with slower, messier feedback — which makes the discipline harder to maintain and more valuable when it is.