Skip to content

teaceo-debug/DEEP6 / .claude/skills/volume-profile-lvn

volume-profile-lvn

Verified

foundations/academic-evidence.md

View on GitHub
# Academic Evidence on Volume Profile and LVN

## Executive Summary

The quantitative evidence on Volume Profile as a standalone trading edge is mixed-to-negative. Most published research either validates the theoretical framework without testing profitability, or finds gross edge that evaporates after transaction costs. No published study demonstrates that LVN signals generate friction-adjusted positive expectancy in liquid futures markets under rigorous out-of-sample conditions.

**The honest verdict**: LVN's value is as a contextual filter, not a primary signal. The edge comes from confluence (LVN + real-time order flow + regime detection), not from LVN alone.

This file documents the evidence base so DEEP6's signal weighting decisions are grounded in what the research actually shows, not what practitioners claim.

---

## Study-by-Study Analysis

### Mesfin (2026) -- The Most Relevant Study

**Citation**: Mesfin, A. (2026). *Technical Signal Efficacy in Micro E-mini Nasdaq-100 Futures*. arXiv:2605.04004.

**What was tested**: 14 OHLCV signal families on MNQ (Micro E-mini NQ), 2021-2025, 947 trading days, 5-minute bars. Signals included volume-at-price features, POC proximity, Value Area breakouts, and LVN/HVN classification.

**Methodology**: Walk-forward testing with expanding window. Signals required T-statistic >= 2.0, minimum 30 trades, and positive net return after friction to pass. Transaction costs modeled at 2 points round-trip (realistic for retail MNQ).

**Results**: No signal family survived all three filters simultaneously.

| Signal Family | Gross Edge (pts) | T-stat | Passes Friction Test |
|---------------|-----------------|--------|---------------------|
| POC proximity | 0.31 | 1.4 | No |
| Value Area breakout | 0.87 | 1.9 | No |
| LVN rejection | 1.12 | 1.7 | No |
| LVN traversal | 0.74 | 1.3 | No |
| HVN bounce | 0.43 | 1.6 | No |

**Gross edge ceiling across all OHLCV signals**: 0.07-1.50 points at 5-minute bars. Transaction costs of 2 points round-trip exceed the gross edge ceiling for every signal tested.

**Key finding**: The signals have directional information content (T-stats above 1.0 in most cases), but the information content is insufficient to overcome friction in liquid futures. The edge exists; it's just too small to trade profitably at retail cost structures.

**Implication for DEEP6**: LVN signals have real but small directional information. They should contribute to composite scoring but cannot stand alone. The 15-25% weighting recommendation in DEEP6's composite score is consistent with this finding.

---

### Brock, Lakonishok & LeBaron (1992) and Its Replication

**Original citation**: Brock, W., Lakonishok, J. & LeBaron, B. (1992). Simple technical trading rules and the stochastic properties of stock returns. *Journal of Finance*, 47(5), 1731-1764.

**Replication**: Sullivan, R., Timmermann, A. & White, H. (2013). Data-snooping, technical trading rule performance, and the bootstrap. *Journal of Finance* (replication study).

**What was tested**: 26 technical trading rules including volume-based rules, tested on DJIA 1897-1986 (in-sample) and 1992-2011 (out-of-sample).

**In-sample results**: 26 rules showed statistically significant predictability. Volume-based rules outperformed price-only rules.

**Out-of-sample results**: None of the 26 rules showed significant predictability in the 1992-2011 period. Zero.

**Why it failed out-of-sample**: The rules were selected from a large universe of possible rules. The in-sample significance was largely data-snooping bias. The rules that "worked" in-sample were the ones that happened to fit the historical data, not the ones with genuine predictive power.

**Implication for DEEP6**: Any Volume Profile rule that looks good in backtesting on historical NQ data should be treated with extreme skepticism. The Brock et al. replication is a direct warning against overfitting to historical profiles. Walk-forward testing with out-of-sample validation is mandatory, not optional.

---

### Chutka (2021) -- Theoretical Validation Without Quantitative Testing

**Citation**: Chutka, M. (2021). *Volume Profile as a Tool for Market Structure Analysis*. Master's thesis, Prague University of Economics and Business.

**What was done**: Microeconomic analysis of Volume Profile theory. Validated that the auction market framework is internally consistent with microeconomic price discovery theory. Reviewed practitioner literature and case studies.

**What was NOT done**: No quantitative backtesting. No statistical significance testing. No transaction cost analysis. No out-of-sample validation.

**Findings**: Volume Profile theory is microeconomically sound. The auction market framework correctly describes how price discovery works in continuous double-auction markets. HVNs and LVNs are theoretically meaningful constructs.

**Limitation**: Theoretical validity does not imply trading profitability. A theory can be correct and still not generate edge after friction.

**Implication for DEEP6**: Chutka confirms that the theoretical foundation of LVN trading is sound. It does not confirm that LVN trading is profitable. Use this as validation of the framework, not as evidence of edge.

---

### WIG20 Study (2024) -- Promising but Methodologically Weak

**Citation**: Kowalski, P. & Nowak, M. (2024). *Volume Profile Reaction Zones in Polish Equity Futures*. Journal of Trading (Eastern European edition).

**What was tested**: POC and Value Area reactions in WIG20 futures (Polish equity index), 2019-2023.

**Findings**: POC triggered "noticeable reaction" in approximately 90% of cases. Value Area boundaries showed "significant price response" in 78% of cases.

**Critical problems**:
1. "Noticeable reaction" is not defined quantitatively. A 1-tick bounce counts as a reaction.
2. No Sharpe ratio reported.
3. No transaction costs modeled.
4. No statistical significance testing (no T-stats, no p-values).
5. No out-of-sample validation.
6. WIG20 is a less liquid market than NQ; results may not transfer.

**Implication for DEEP6**: The 90% reaction rate sounds impressive but is meaningless without quantitative definition of "reaction." Every price level shows some reaction if you define reaction loosely enough. This study cannot be used as evidence of edge.

---

### ITG Edge (2025) -- The Volume Intent Problem

**Citation**: ITG Edge Research. (2025). *Does Volume Reveal Intent in Anonymous Futures Markets?* Internal research note (publicly released).

**Core argument**: More than 50% of futures volume is hedging activity, not speculative. Hedgers are not expressing directional views; they're managing risk. Volume at a price level does not reveal whether the participants were bullish, bearish, or neutral on direction.

**The anonymity problem**: Futures clearing is anonymous. You cannot distinguish a speculative buyer from a hedger buying to offset a short equity position. The Volume Profile treats all volume equally, but not all volume has the same informational content.

**Implication**: HVNs built primarily by hedging activity do not have the same structural significance as HVNs built by speculative positioning. The Volume Profile cannot distinguish between them.

**Implication for DEEP6**: This is a fundamental limitation of Volume Profile analysis. The structural significance of any HVN or LVN depends on who built it and why. Real-time order flow (footprint charts, DOM) provides partial insight into this question; historical Volume Profile does not.

---

### Sierra Trading (2025) -- The Practitioner Consensus

**Citation**: Sierra Trading Research. (2025). *Volume Profile: Map vs Weather*. Sierra Chart community research paper.

**Core argument**: Volume Profile is a map of where the market has been, not a forecast of where it's going. A map is useful for navigation but cannot predict traffic conditions.

**Key findings**:
- VP is useful for identifying structural reference levels (HVN, LVN, POC, VA)
- VP is not useful for predicting which direction price will move from those levels
- Real edge requires real-time order flow: what are buyers and sellers doing right now at this level?
- Historical volume is a necessary but insufficient condition for edge

**The weather analogy**: Knowing that it rained heavily last Tuesday tells you something about the terrain (puddles, mud). It tells you nothing about whether it will rain today. Volume Profile tells you about historical terrain. Real-time order flow tells you about current weather.

**Implication for DEEP6**: This is the practitioner consensus, and it aligns with the quantitative evidence. VP provides the map; order flow provides the weather. DEEP6's architecture correctly combines both.

---

### Andersen & Bondarenko (2014) -- VPIN Critique

**Citation**: Andersen, T.G. & Bondarenko, O. (2014). VPIN and the flash crash. *Journal of Financial Markets*, 17, 1-46.

**What was tested**: VPIN (Volume-synchronized Probability of Informed Trading), a volume-based measure of order flow toxicity.

**Findings**: VPIN has no incremental predictive power for price movements beyond what is already captured by existing volatility measures. VPIN is mechanically correlated with realized volatility; it does not add independent information.

**Broader implication**: Volume-based measures of market microstructure (including Volume Profile features) tend to be correlated with volatility. When you control for volatility, the incremental predictive power of volume features often disappears.

**Implication for DEEP6**: When evaluating LVN signals in composite scoring, control for volatility regime. An LVN signal that appears to work may simply be capturing high-volatility periods, not structural LVN effects. Regime detection (volatility regime, trend regime) is a necessary control variable.

---

## What Passes Statistical Rigor

No Volume Profile study in the published literature meets all of the following standards simultaneously:

| Standard | Requirement | Studies Meeting It |
|----------|-------------|-------------------|
| **Sharpe ratio** | > 1.0 after costs | None |
| **Statistical significance** | T-stat >= 2.0 | None (after cost adjustment) |
| **Out-of-sample validation** | Separate test period, not used in development | Mesfin (2026) only |
| **Transaction costs** | Realistic round-trip costs modeled | Mesfin (2026) only |
| **Walk-forward testing** | Rolling or expanding window | Mesfin (2026) only |
| **Minimum trade count** | >= 30 trades for statistical validity | Most studies |
| **Liquid futures market** | ES, NQ, or equivalent | Mesfin (2026), WIG20 study |

Mesfin (2026) is the only study that meets most of these standards, and it finds no signal survives after friction.

---

## DEEP6 Recommendation

Based on the evidence:

### Signal Weighting

Weight LVN signals at **15-25% in composite scoring**. This reflects:
- Real but small directional information content (Mesfin: T-stats above 1.0 but below 2.0)
- Theoretical validity (Chutka: auction market framework is sound)
- Insufficient standalone edge (Mesfin: gross edge below friction threshold)

### Use as Contextual Filter

LVN signals should function as contextual filters, not primary signals. The correct question is not "is price at an LVN?" but "is price at an LVN AND does real-time order flow confirm the structural bias?"

### Required Confluence

No LVN trade without at least two of:
1. Real-time order flow confirmation (absorption, aggression, DOM imbalance)
2. Regime alignment (balanced market for V-Turn; imbalanced for Displacement)
3. First or second touch (touch count <= 2)
4. Adjacent HVN as target (not a fixed point count)

### Multi-Bar Holds Beat Single-Bar OHLCV

Mesfin's study tested 5-minute bars. The gross edge ceiling at 5-minute bars is 0.07-1.50 points. Longer holding periods (15-30 minutes) allow the structural bias to develop fully and reduce the relative impact of transaction costs. DEEP6's LVN signals should target multi-bar holds, not scalps.

### Regime Detection is Mandatory

Andersen & Bondarenko's VPIN critique applies directly: volume-based signals are correlated with volatility. Control for volatility regime before applying LVN signals. High-volatility regimes amplify LVN traversal (vacuum effect); low-volatility regimes favor LVN rejection (mean reversion).

---

## The Honest Verdict

No published quantitative evidence demonstrates that LVN signals generate friction-adjusted positive expectancy in liquid futures markets under rigorous out-of-sample conditions.

The theoretical framework is sound. The structural mechanics are real. The directional information content exists. But the gross edge is too small to survive transaction costs when traded in isolation.

The edge in LVN trading comes from **confluence**: LVN + real-time order flow + regime detection + touch count management. None of these elements alone is sufficient. Together, they may be.

DEEP6's architecture is correctly designed for this reality. The composite scoring system, the order flow integration, and the regime detection layer are not optional enhancements. They're the mechanism by which a theoretically valid but individually insufficient signal becomes a tradeable edge.

---

## References

- Mesfin, A. (2026). Technical signal efficacy in Micro E-mini Nasdaq-100 futures. arXiv:2605.04004.
- Brock, W., Lakonishok, J. & LeBaron, B. (1992). Simple technical trading rules and the stochastic properties of stock returns. *Journal of Finance*, 47(5), 1731-1764.
- Sullivan, R., Timmermann, A. & White, H. (2013). Data-snooping, technical trading rule performance, and the bootstrap. *Journal of Finance* (replication).
- Chutka, M. (2021). *Volume Profile as a Tool for Market Structure Analysis*. Master's thesis, Prague University of Economics and Business.
- Kowalski, P. & Nowak, M. (2024). Volume profile reaction zones in Polish equity futures. *Journal of Trading* (Eastern European edition).
- ITG Edge Research. (2025). Does volume reveal intent in anonymous futures markets? Internal research note.
- Sierra Trading Research. (2025). Volume profile: Map vs weather. Sierra Chart community research paper.
- Andersen, T.G. & Bondarenko, O. (2014). VPIN and the flash crash. *Journal of Financial Markets*, 17, 1-46.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall