Trendline Detector

Both approaches work. Wicks capture full price rejection (where buyers/sellers actually stepped in). Bodies focus on where price 'agreed' to close. Many traders use wicks for more accuracy. Most important: be consistent in your approach.

A trendline is valid until broken. Signs of validity: price continues to respect the line (bounces), no close beyond the line. Signs of invalidity: price closes beyond line with volume. A trendline that's been broken is no longer valid for bounce trades.

Sustainable trends typically have 20-45 degree angles. Too flat (<10°) may be horizontal S/R instead. Too steep (>60°) is unsustainable and will break. Moderate angles indicate healthy, tradeable trends.

Yes, and you often should. You might have a long-term trendline (months), medium-term (weeks), and short-term (days). Multiple trendlines show the trend at different scales. Higher TF lines are more significant.

Trendlines are manually drawn through swing points at fixed angles. Moving averages are calculated averages that automatically adjust. Both can act as dynamic S/R, but trendlines show the 'natural' trend angle while MAs show average price.

1) Wait for close beyond line (not just wick), 2) Confirm with above-average volume, 3) Enter on close or wait for retest of broken line, 4) Stop on other side of broken line, 5) Target based on measured move or prior structure.

After a trendline breaks, price often comes back to test the broken line. Broken support becomes resistance (short the retest); broken resistance becomes support (buy the retest). This gives better entry and confirms the break was real.

1) Require candle close beyond line (not just wick), 2) Look for above-average volume, 3) Wait for follow-through (next bar continues direction), 4) Consider waiting for retest. No method is perfect, but these filters help.

Use wider tolerance for touches (1-2% vs 0.5%). Focus on obvious swing points. Accept that lines may not be as clean. Consider using closing prices only for cleaner lines. Higher timeframes help filter noise.

Redraw when: new swing point provides cleaner line, current line is clearly being violated. Don't redraw just to avoid admitting a break. If you're constantly redrawing, the trendline probably isn't valid.

Algorithm: 1) Detect swing points (fractal method - high/low higher/lower than N bars each side), 2) Generate all line candidates between same-type points, 3) Count touches within tolerance, 4) Filter by angle, touches, duration, 5) Score and output top lines.

Scoring factors: Touches (more=better), Duration (longer=better), Timeframe (higher=better), Recency (recent touches=better), Clean touches (minimal penetration). Formula: Score = Touches × TF_multiplier × Recency_bonus. Rank by score.

Convex hull = outer boundary containing all points. For swing lows, compute convex hull; the lower boundary segments form trendlines. Advantage: objective, no parameter tuning. Disadvantage: may miss internal trendlines.

Features: slope, touch count, duration, angle, volume at touches, bounce magnitude trend, higher TF trend. Target: break within N bars (binary). Train classifier on historical trendlines. Output probability of break. Focus on high-hold-probability lines.

Architecture: 1) Stream price data, 2) On new bar: update swing detection, check trendline touches/breaks, 3) State machine for each trendline (forming→confirmed→active→tested→broken), 4) Fire alerts on state changes, 5) Push to dashboard/mobile.