RTS5Pattern V.20.00

RTS5Pattern-profi v20.00 is an interactive pattern-matching indicator for MetaTrader 5. Select any price segment on the chart (first and last candle), the tool searches historical windows for the closest matches, projects their continuations, and proposes ENTRY, TP, and SL with clear on-chart visuals. Workflow remains click-to-analyze, fully integrated with the chart.

Area	Version 10.50	Version 20.00 (Current)
Language & UX	Mixed CZ/EN texts	Full English texts & labels, consistent UX
Time anchor & repaint	Basic redraw	Stable time-anchor (repaints segment reliably)
Top matches	Top-3 drawn initially	Top-3 stored in ERR_TOP and reused for repaint
AStat calculation	Weighted mean	Weighted mean with EMPTY_VALUE safety guards
Pattern similarity	OHLC min–max normalization	Same core logic; code streamlined for clarity
TP/SL hit test	M1 comparison	Precise M1-based order check (TP vs SL) with markers
Object lifecycle	Occasional residue	Consistent cleanup (lines, labels, vertical markers)
Controls	Buttons & label	>> / << navigation, Reset RTS, status label
Performance	Optimized vs 7.50	Same footprint; predictable runtime on ≥1000 bars
License / expiry	Various dates	Time-lock enforced in OnInit & OnCalculate — valid through 31 Dec 2025

Core Functionality

• Pattern Selection — Click the first and last candle to define the reference pattern.
• Similarity Search — Normalized OHLC differences across historical windows (optional volume penalty).
• Projection Generation — Three candidate paths (Predict, Predict1, Predict2) + AStat consensus.
• Trade Proposal — Automatic ENTRY/TP/SL plotted on chart; M1-based hit order (TP vs SL).
• Navigation — Browse matches with >> / <<, reset selection with Reset RTS.

Inputs

Parameter	Description
PredictionLen	Forward projection horizon (bars).
VolumeCalculate	Include volume differences in similarity error.
MSQErr	Maximum error threshold for accepting a match.
Plot	Toggle graphical output.

Outputs & Objects

• Predict, Predict1, Predict2 — Candidate projection curves.
• AStat — Weighted consensus (safe handling of missing points).
• Load — Visual reconstruction of the selected pattern vs best match.
• ENTRY / TP / SL — Horizontal levels with labels and hit markers.
• RTSPatternsLabel — “current / total” match indicator.

License and Time Restriction

RTS5Pattern-profi v20.00 is licensed for use until 31 December 2025 (inclusive). Time-lock is enforced in initialization and runtime. After expiry, the indicator displays a notice and stops. For extensions, contact Remind.cz.

Who is it for? Traders using price action and classical S/R who want historical validation for visually spotted patterns and a

consistent, on-chart projection workflow.

Why Analog-Based Pattern Analysis Beats “Indicator-Only” Trading

Executive Summary

Analog-based pattern analysis searches large price histories for segments that look like the one you’ve just selected and projects what typically followed. Done well, it turns subjective price-action reading into evidence-based decision making. The edge comes from three realities: (1) markets often recycle structure across regimes, (2) volatility and microstructure generate recurring motifs, and (3) humans respond to similar contexts in predictable aggregates. The output is not a single path but a distribution—letting you size trades based on probability and expectancy, not hunches.

---

1) What “analog” pattern analysis is (and isn’t)

Is: A systematic search for historical look-alikes of a user-defined pattern; a projection of future outcomes derived from the continuations of those look-alikes; a way to quantify P(TP before SL) and expected value (EV) for concrete trade planning.

Isn’t: Curve-fitted astrology or a generic indicator mash-up. It doesn’t predict the future deterministically; it conditions your decision on the closest available historical evidence.

---

2) Why the approach can work

1. Structural recurrence: Market participants recycle behavior around liquidity pockets, session boundaries, and risk events. Price compressions, breakouts, and mean-reversions recur with similar shape even if absolute price levels differ.
2. Path dependence: Order flow and inventory dynamics produce local path similarities (squeezes, pullbacks, failed breaks).
3. Regime persistence: Trends and vol regimes often persist long enough that short-term analogs carry information about near-term odds.

---

3) The critical shift: from a single “call” to a distribution

Traditional indicators emit a binary or scalar signal. Analog analysis yields a set of plausible futures. From the top-K matches you can compute:

• Median Path (P50) and Uncertainty Bands (P10 / P90)
  Probability of Target before Stop: P(TP < SL)
  Expected Value (EV):
  EV = Σ (wᵢ × Rᵢ), where weights wᵢ are proportional to exp(−λ × errorᵢ),
  and Rᵢ is the R-multiple derived from the continuation of each analog.

This lets you optimize TP/SL and position size based on objective odds.

---

4) Advantages vs. indicator-only and black-box ML

Compared to indicator-only:

• Context awareness: Uses the entire local path (shape), not just pointwise indicator values.
• Scenario thinking: Quantifies multiple futures instead of betting on one.
• Decision auditability: You can show the exact historical analogs that informed the trade.

Compared to black-box ML:

• Transparency: No opaque feature engineering; similarity and outcomes are inspectable.
• Data efficiency: Works well with modest data and non-stationarity; less reliant on stable labels.
• Robustness to drift: Re-runs similarity on current data rather than trusting a static model fit months ago.

---

5) Measuring “similarity” the right way

Even a basic setup (normalized OHLC with optional volume) beats eyeballing, but precision improves with:

• Return-space normalization: Use log returns and Z-scores to reduce scale effects.
• Time alignment: Allow small warps (e.g., DTW with a band) to tolerate tempo differences.
• Multi-channel features: Combine close-returns, wick/real-body ratios, and ATR structure; weight channels.
• Affine fit on overlap: Small gain+offset alignment stabilizes projections without overfitting.
• Regime penalties: Down-weight analogs from different trend/volatility regimes.

(Your current EN v20.00 uses normalized OHLC plus optional volume and a stable repaint via time anchor; these fundamentals already deliver practical value. The above ideas are incremental upgrades if/when you want more precision.)

---

6) Turning analogs into trades

A disciplined workflow:

1. Select pattern (first → last candle).
2. Collect top matches; compute continuation curves.
3. Form AStat (weighted consensus) and uncertainty band.
4. Compute P(TP before SL) and EV over a TP/SL grid.
5. Choose the pair (TP, SL) maximizing EV under your risk constraints (e.g., cap SL distance, ensure P(TP) ≥ threshold).
6. Validate context (trend alignment, S/R, news risk).
7. Execute and monitor; use the pre-computed expected time-to-hit for managing patience vs. opportunity cost.

---

7) Risk management: from odds to size

• R-multiple framing: Compare EV across setups independent of instrument volatility.
• Kelly-lite: If ppp is calibrated and edge is small, size a conservative fraction of Kelly to control drawdowns.
• Vol-aware SLs: If projections span a wide band, widen SL but reduce size to keep risk per trade constant.
• Portfolio view: Prefer setups with uncorrelated analog contexts to avoid clustered losses.

---

8) Validation and reliability

• Walk-forward evaluation: Freeze your similarity settings, replay history, and collect out-of-sample results.
• Hit-order truth: Always check which came first on a finer timeframe (e.g., M1) to avoid bar-aggregation bias.
• Calibration: Adjust the weighting temperature λ\lambdaλ so that predicted probabilities match observed frequencies (reliability).
• Leakage control: Exclude any future bars from the similarity window; anchor strictly to the user’s end bar.

---

9) When to stand down

• Regime breaks and macro events: CPI, NFP, rate decisions, earnings—history can be a poor guide during shocks.
• Illiquid sessions: Thin books deform analog continuity.
• High divergence: If top-K futures disagree wildly (wide band, multimodal outcomes), skip or reduce risk.

---

10) Performance considerations

• Shortlisting: Pre-filter candidates with cheap metrics (e.g., downsampled returns) before detailed comparison.
• Window lengths: Keep pattern length meaningful (15–40 bars is a good starting band).
• Caching: Reuse computed errors when panning/zooming; store top-K indices for stable redraw (as your v20.00 does).
• Complexity vs. latency: Add sophistication (e.g., DTW) only if it materially improves out-of-sample performance.

---

11) Practical checklist before entry

• Pattern length is sufficient and context-relevant.
• Top analogs are plausible on visual inspection (no regime mismatch).
• AStat aligns with higher-timeframe bias or S/R context.
• P(TP≺SL)P(\text{TP} \prec \text{SL})P(TP≺SL) and EV are above your cutoffs.
• TP/SL are practical (not inside obvious liquidity or micro-noise).
• News calendar clear; liquidity normal.

---

12) Frequently asked questions

Q: Isn’t this just curve fitting to history?
A: Only if you over-tune similarity and ignore out-of-sample checks. With fixed rules and walk-forward validation, you’re conditioning on empirical frequencies, not fitting parameters to maximize past PnL.

Q: What if two analogs conflict?
A: That is information—dispersion signals uncertainty. Either pass or cut size. Systems fail when they force conviction where the data shows ambiguity.

Q: Why not just use an LSTM or transformer?
A: You can—but analog methods offer interpretability, lower maintenance, and resilience to regime drift. If you later add ML, feed it analog features (error ranks, dispersion, regime flags) to boost signal quality.

Q: Does volume matter?
A: Sometimes. Include it as a soft penalty, not a hard filter, to avoid discarding good price analogs when volume data is noisy or broker-specific.

---

Closing

Analog-based pattern analysis won’t predict the future; it prices it—by showing the range of likely paths and their frequencies. That shift—from a single opinion to a distribution with calibrated odds—is the core advantage. It lets you trade expectancy, not anecdotes; process, not impulse.

https://www.youtube.com/watch?v=Up960fCgvsw&t=1s

---

Author: Tomáš Bělák