Auction Market Value Theory & Analytics

Lately, I was stuck in a dilemma of some sort

Is building market models enough to forecast price? If one model is not powerful enough, we then add more to our pipeline, is it just about getting the right pipeline?

In other words, I was starting to think of algorithms as a technical indicators on a chart. As a beginner would think "Maybe if I can get the correct and most advanced technical indicators and combine them together, I can get the most profitable strategy"

And yes, that is neither true or false. It might be possible to do that. In-fact some people might have done it already.

But what distinction do consistent traders have as compared to greedy retail traders? We often think about this statement and come to some sort of a conclusion.

My conclusion is that, consistent traders don't focus on the aspects that are the most attractive to a retail trader. That's an easy answer. But let's get more specific.

The consistent trader, has an instinct. Though, that's his major advantage, but it's also his level of knowledge, theory etc. All of them play a role, into something specific. An Idea.

Think about this, innovation does not happen in the absence of inspiration. Innovation occurs when you understand a subject soo deeply and have the perfect attributes (being fully committed, persistent) which creates a powerful mind that does not fear to explore but knows to stay close to reality simultaneously.

In a way, that's not just a description of a consistent trader, but any person who has been successful in any career.

I finally and fortunately got past the dilemma and focused on exploring instead.

I have built a much more powerful algorithm with the limited knowledge I have about automated systems.

I incorporated 2 things from Timothy Masters to criticize the model.

1. Walk Forward
2. Walk Forward Permutation Test

Walk forward is simply used so that the model has the least amount of chance to be overfitted.
The permutation test is used so that we can understand if the model is performing based on luck or actual patterns in the data.

Here are the results:



I don't want to market my results as if I have the "best algorithm ever". Rather, I want to teach people the importance of doing something correctly.

It's not about what you want to see. It shouldn't be guided by your fantasy. And you know that too.

I believe in my results for the two main things I mentioned above:

1. Walk Forward Testing: The model had a total data-set range from 2018 till 2024. The model trained on half a year and tested on a half year, it did this repeatedly for every year and got up to 12 folds in total. The results are the average of what we got in each fold or half year data set.
2. Walk Forward Permutation Test: By knowing the P-Values are less then 0.05, I can confirm that the model is statistically significant and uses actual patterns in the data rather than getting lucky through random chances.

Other metrics also show great numbers, but I don't want to go into too much detail.

My main goal was to tell people that yes, it's possible. However, it's also extremely dangerous to view it as a money-making machine.

It's the system's design that excels, not the data or complex mathematical formulas themselves!

Cheers.

New CME Gold contract just added for trading! Maintenance margin is only $165 bucks per contract. I decided to take a stab at shorting here. In most instances my breakout trades never meet the TP, because the market data forces a close sooner


Failed break. Hit my Quadrant stop. Previous day's TPO ratio was pretty lopsided, but since this was a newer contract I used data from the larger Gold contract to judge it. Oh well

Guys, I don understand one thing. Why mostly people from here trade forex using Daltons approach and AMT. His main tool for general thoguhts and analyzation is market profile. And as we know forex has uncentralized data. Market profile is only useful in futurse... Let me know what you all think about it.

While it is true that some tools work better in certain markets, but it's not true to limit a tool to only that market aswell.

Market Profile and its concepts can perfectly be applied to any market because it does not require special data like volume or order-book. Dalton's choice of futures market is his own preference. You could say, that his preference was biased from the original creator of the Market Profile Peter.

There's a whole forum related to your question, they apply market profile concepts to forex markets and it works fine. But, the problem with market profile is that it's very subjective which led to the creation of AMVT.

However, please read the first page on this forum. This forum does not intend to discuss market profile because new-comers get confused with market profile (AMT) with AMVT. Both are related, but different in practice.

The Volume Reference Point: Continuation Quantified
The primary reference point is volume. Volume is the best indicator of the psychology of the traders who make up the market. Growing volume indicates that market participants support what is happening, that trade is being facilitated. If volume is large relative to average values, even if it is declining, trade is most likely still being facilitated. When volume dries up in the short term, market participants are voting with their wallets against the current state of affairs. If volume continues to drop over a long time, the market itself is in danger. Very low volume markets are dangerous to trade. They are illiquid (trade is not being facilitated) and you run the risk of being the only paper on the floor.

Phase Reference point Interpretation
Bracketing: Volume decreasing positive for continuation
Bracketing: Volume increasing positive for change
Bracketing: Volume stable positive for continuation

Testing: Volume decreasing positive for continuing bracket
Testing: Volume increasing positive for trending
Testing: Volume stable positive for continuing bracket

Trending: Volume decreasing positive for change
Trending: Volume increasing positive for continuation
Trending: Volume stable positive for continuation if volume high
Trending: Volume stable positive for change if volume low

Trend end: Volume decreasing positive for change
Trend end: Volume increasing positive for continuation of trend
Trend end: Volume stable positive for continuation of trend high vol
Trend end: Volume stable positive for change of trend low vol




Volatility Reference Point: Additional Quantification of Continuation
The CISCO volatility (CV) is designed to measure short-term interest in the market. If volatility is too small it indicates lack of opportunity, there is little movement. Too high volatility characterizes high risk. Unlike baby bear's porridge, there is no 'just right' volatility. Rather, in most markets there is quite a wide range of perfectly acceptable volatilities. For instance, in swing trading T-bonds, CV greater than 12 is high, CV less than 4 is low. We rarely worry about too low volatility because those markets have a small trading range and we discard them for their lack of profit potential.

Phase Reference point Interpretation
Bracketing: Volatility increasing positive for continuation
Bracketing: Volatility decreasing positive for change
Bracketing: Volatility stable positive for continuation

Testing: Volatility decreasing positive for continuation
Testing: Volatility increasing positive for trending
Testing: Volatility stable positive for continuation if Volatility low
Testing: Volatility stable positive for trending if Volatility high

Trending: Volatility decreasing positive for change
Trending: Volatility increasing positive for continuation
Trending: Volatility stable positive for continuation if Volatility high
Trending: Volatility stable positive for change if Volatility low

Trend end: Volatility decreasing positive for change
Trend end: Volatility increasing positive for continuation of trend
Trend end: Volatility stable positive for continuation of trend if high
Trend end: Volatility stable positive for change of trend if low




Trade Facilitation Factor Reference Point
The Trade Facilitation Factor is based on the price events on standard half-hour bars, where there can be no more than a single occurrence (TPO) at any price per half-hour. TFF is defined as the number of TPO's divided by the number of price steps (the range). Limiting values of the TFF occur at perfectly un-facilitating markets (a locked limit market with trading throughout the day at that single price), and perfectly facilitating days (a trend in which each period moved through many prices. Totally un- facilitating markets will have a TFF that equals the number of periods. Maximum facilitation is one TPO for each price, or TFF = 1.

Markets that are not facilitating trade are often coalescing into a critical mass. Traders are sitting on their wallets -- unusually low volume is reported. Something is going to blow. Large TFF's are an alert. This is a very reliable tip-off. Direction is not indicated by the TFF.



Initial Balance (IB) Reference Point The Initial Balance is the period from the opening bell for about an hour that the floor traders (locals) seek out a balance region for their trading. We are interested in who is in control. If the market trades throughout the day within the initial balance, the floor is in control. If the close is above or below the Initial Balance, long time frame traders (public) are in control.

During a trend, the close will generally lie outside the initial balance, in the direction of the trend. As a trend pauses, the close can lie within the Initial Balance. A close on the opposite side of a trend is an alert that the trend is over.

The Initial Balance is a contributor to continuation analysis.



Range (RNG) Reference Point Daily price range tends to expand during trends and contract during balances. When it is very narrow during a balance it alerts to an explosive move. The wider the RNG the more confident the traders. During a trend, the run-pause character of the market insures narrow RNG days during the pauses. Normally, elevated activity continues into the trend end.

Point of Control (POC) Reference Point The Point of Control tends to follow the market. It is the point of most trading for the day. POC tends to be a coincident indicator with the Value Area.



Value Area (VA) Reference Point Each day a market finds value (those prices that are accepted). Value encompasses the central 70% of trading volume (or TPO's). The high point, peak volume is the POC. Value Area, like trading range, expands with confidence, with trends. In quiet markets, T-bond's Value Area is of the order of a dozen or so ticks. As the market heats up, the number of ticks in the VA will increase. We use this data in the same manner as the Range.


-Here is some wisdom from CISCO

After much work and hardships, I got to a turning point. Discretionary Analysis merged with Dynamical Machine Learning Models is feat that can be taken only if enough resources are present.

My main limitations are that I don't have a good enough system or infrastructure to support the training of Machine Learning systems. If one has to do something the right away, there's usually a cost to it. In my context, the cost is investment in good GPU's, development complexities and education required to understand key concepts.

However, this has not stopped me, it's merely a turning point. To me, this tells me there's an opportunity to share my knowledge with others. As I have not seen any activity in this forum for some time, it's best that I share my analysis with all of you and maybe we can spark conversations and learn from each other.

Just like the good ol' days of this forum, where MzVega and others used to share trades and ask questions to learn, I hope we can revive that spark in this forum if people do contribute.

I will be sharing my analysis for the trades I want to partake in (or not and are worth sharing anyway). Let's start !

Todays Date: 9/7/2025, Friday
Market: GBPUSD

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Todays Date: 9/7/2025, Friday
Market: EURUSD

Attached Image (click to enlarge)

Market: NZDUSD

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Heads Up!

All Markets range is un-usally large. Which means Greater Risk, but also Greater Profit Potential. The market is behaving a bit radically and irrationally, which can introduce public/retail traders. This leads to frequent stop-loss hits if not placed properly.

Trades require special care after entering.

Results: 9/8/2025
Market: NZDUSD (WIN)

Attached Image (click to enlarge)

Market: GBPUSD (Loss)

Attached Image (click to enlarge)

Market: EURUSD (Order Modified After NZD Breakout & GPUSD Loss)

Short Discussion on Market Correlation

The losses on GPBUSD did not make sense to me. From the perspective of AMVA (Auction Market Value Analytics), it didn't make sense to buy the pair. At-least in my opinion. Admittedly, the activity was all over the place, it was also not clear either the market will be going down. But, we learn and we pray.

The activity and direction was much much clear on NZDUSD pair. And the results say so. However, I fear that AMVA does not take into account the correlation of different symbols. AMVA although is great in terms of analyzing and knowing entry/exit points. It's not so great for market-to-market comparison. At-least, in the aspect of correlation.

Mathematically, I have no proof as to if NZD/USD is actually correlated, but the recent move makes it a gut-feeling. Even eye-balling previous movement of NZDUSD and GBPUSD even EURUSD, we can see the market moves almost simultaneously.

For now, I will avoid taking opposite positions on symbols that have the same last currency. But, this can be further researched and a very good opportunity to understand market correlation in terms of AMVA.

Todays Date: 9/9/2025, Tuesday
Market: AUDCHF

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Attached Image (click to enlarge)

Market: EURCHF

Attached Image (click to enlarge)

Like what you’re doing here. I don’t have any experience trading cross pairs, but recognize every instrument has its own unique characteristics. For example, on the S&P and Dow contracts I’ve had to use AMVA as a momentum filter, using past 3 days of data before deciding to open a trade. Also I found it not profitable to trade breakouts in the extended hours and instead start the logic screening for trades at the IB.

Some contracts I use a break even based on time, others not, and all contracts I’ve adopted a trailing SL trigger at various profit points. Manually I just pluck at the market, mostly still in backtesting phase discovering what actually works.

I personally have not traded anything else then forex pairs, but it's very much a try. Some concepts of market profile/AMVA like IB can not be really used for forex and has different meaning market to market. Sure, IB can be used per session, but that's also something not been done previously in AMVA.

I hope it works out for you, keep us updated I am very curious to the application of AMVA to markets other then forex

My fixed lot test results are extremely promising. But to really make the cheddar you need to implement some form of position sizing. I’ve adopted a modified version of Williams MM and plan to run at least four contracts at the same time. When I go live then I can post a myfxbook account for MT5

I see..

Are you interpreting the S&P and other indices the same way you did forex? I also saw you tried commodities like Gold. I am interested in how you interpreted the reference points. Was it the same way? Or did you had any internal ideas?

I say this because Donald L. Jones, emphasized knowing who and where is participating. In commodities and indices, as it's not OTC market, it can actually be analyzed where the public, commercial and floor traders operate. Have you tried this?

I’ve looked at the chapters on commercial volume in VBPT, and I’ve read his writings in multiple pages concerning commercial capping, and don’t use it like that.

I can see large volume spikes and recognize completed or incomplete auctions but it’s only one reference point. Other references have to align with that. I use only Total Volume in two ways. When scanning for rotation back into the brackets, declining TPOs, volume, and volatility are big tip offs. Sometimes TFF plays a factor. And for the breakout, volume and volatility should be increasing

Ohh okay, so it's very similar to how I interpret reference points but for forex

Great discussion happening here. AMV is a powerful lens, but I’ve been experimenting with a system that extends it using real-time news-order flow fusion and adaptive auction theory. Here’s a high-level look:
Auction Reactivity via News & Microstructure

1. The system detects imbalance shocks in the LOB (bid/ask asymmetry > |0.25|) and correlates them with NLU-processed news events (earnings, macro shifts).
2. Instead of static value areas, it uses regime-switching Kalman filters to adapt POC/VA levels in real-time based on order flow pressure.
3. If news sentiment and order imbalance align (e.g., +sentiment & +imbalance), it interprets it as auction acceleration and adjusts short-term bias.

Risk-Aware Auction Participation

1. No static stops or targets—every trade uses contextual sizing based on:
   1. Volatility-adjusted brackets (ATR-derived).
   2. News-induced entropy (prevents overexposure during event chaos).
   3. Regime confidence (only auctions with >80% confidence are traded).

2. Each trade is guarded by formal invariants (e.g., no trade if VWAP divergence > 2σ).

Backtest Insight
In ~6 months of paper trading:

1. 72% win rate when auction theory + news momentum align.
2. Drawdown < 8% thanks to invariant-based safety locks.
3. Works best during RTH and high-impact news auctions.

Why It Works
It doesn’t just react to auctions—it anticipates auction redirection using:

1. News as a leading indicator (not lagging).
2. Order book imbalance as conviction signal.
3. Causal models (AIPW) to filter out noise.

I’m still refining the adaptive sizing model, but the fusion of AMV + news + microstruct seems promising. Happy to exchange ideas—always learning.
Keep the thread alive!
i can give you guys a real one but i'm still building it! i will crash the market with this project! trust me! is new and in PhDs level and research!

I have always appreciated innovation, but do know, that there is a thin-line between professionalism and illusion. I am not trying to say that you have a system that does not work, however be really careful when testing out systems or models because there's a tone of stuff you need to take into consideration.

Other than that, I would be really interested to know more about this system. How have you incorporated market conditions? What's the overall entry criteria? How are you testing the system? The 72% win rate sounds attractive, but was it fully mechanical, or did you had some discretion to it? Can you share some examples?

It would be great to hear from you and hope this works out well. If it does not, I would highly recommend that you start simple. Paper trade the basic AMVT strategy and see what can be improved. From there, the path will be much clearer and instead of adding complexity, you will add necessary improvements.