What is signal confluence?
Signal confluence means requiring several independent technical conditions to agree before acting, instead of trusting any one indicator. QuantMedia's engine scores each stock against 30 binary checks — trend, momentum, volume, volatility and 52-week position — and flags a BUY only when at least 22 are simultaneously bullish, roughly 73% agreement. The purpose is to filter noise, and the cost is that confluence systems are structurally late at turning points.
The reasoning
Any single technical indicator produces frequent false positives. RSI drops below 30 often in a downtrend that continues; moving-average crossovers whipsaw in ranges. Requiring agreement across indicator families that respond to different market properties raises the bar for a signal to fire.
The important caveat is that confluence is not statistical independence. Technical indicators computed from the same price series are heavily correlated: if price is above its 5-, 10-, 20- and 50-day averages, those four checks pass together almost by construction. So a score of 22 out of 30 does not represent 22 independent confirmations. It represents broad agreement among overlapping measurements — useful, but a weaker claim than the raw count suggests.
How the score is built
| Family | Checks | What it tests |
|---|---|---|
| Trend / moving averages | 10 | Price vs SMA(5/10/20/50), MA ordering, EMA(12) vs EMA(26), MACD level and signal cross |
| Momentum oscillators | 4 | RSI(14) in the neutral band, RSI(7) vs RSI(14), RSI(21) level, RSI recovery cross |
| Rate of change | 4 | 5/10/20-day returns, plus a check that short-term momentum is not overextended |
| Bollinger Bands | 3 | Position inside the bands and a bounce off the lower band |
| Volume | 3 | Current vs 20-day average, 5-day vs 20-day, up-volume vs down-volume |
| Stochastic | 2 | %K level and %K vs %D |
| 52-week position | 2 | Position within the 52-week range |
| Volatility regime | 2 | ATR as a share of price, inside a workable band |
| Total | 30 | BUY at ≥ 22 |
Each check returns 1 or 0. There is no weighting, no optimisation of the threshold against past returns, and no discretionary override. That keeps the system un-fitted to any particular period — and equally means the threshold carries no claim of being optimal.
Why 22, and what that choice costs
22 of 30 is about 73% agreement. Requiring roughly three-quarters of checks to align means a stock generally has to be in a broad, confirmed uptrend to qualify. The direct consequences are worth stating plainly:
- Signals cluster in rising markets and thin out in falling ones. That is the design working, not a malfunction.
- The engine is structurally late at inflections. It cannot flag a bottom, because at a bottom almost no trend check passes.
- When the median score sits near 22, the BUY flag stops discriminating — half the universe qualifies. On those days the score itself carries more information than the flag. Signal Breadth exists to make that visible.
Limitations
- Correlated inputs. As above, 22 checks are not 22 independent pieces of evidence.
- Purely technical. No earnings, guidance, litigation or index-change awareness. A stock can score 30/30 the day before a profit warning.
- No published track record. QuantMedia does not publish live or audited performance for this engine, because no verified out-of-sample record exists to report.
- End-of-day only. Signals reflect the prior close and can be invalidated by overnight news or a gap.