Build Strategies Across
the Multiverse of Price Paths

The historical data is one path. The future is another. Train your strategies across hundreds of synthetic price series — noise-adjusted, Monte Carlo permutated, and Vs Shifted — so they survive whatever comes next.

By David Bergstrom·Updated Feb 25, 2026·14 min read

Watch Demo Get Build Alpha

Part of the Robustness Testing Guide — 12+ stress tests for algo trading

Why One Historical Path Is Not Enough

All system traders aim to build more robust strategies and bullet-proof their portfolios. But most traders are stuck training on the historical data that every other participant has. We do not want strategies reliant on one market environment persisting forever — that simply is not the way things are.

A robust strategy can withstand changes to the market and continue to perform. It is not dependent on things remaining the same. Our goal: find strategies that handle varying amounts of noise, volatility, sequences of prices, or even price shocks. One way to better do that is by introducing synthetic data to the building process.

Multi-Path vs. Single Path

For decades, system traders have developed strategies on only historical price data. But historical data represents a single path of what actually happened (green below), while various paths were possible (black below) and can exist going forward.

Instead of building strategies on a single path, why not build strategies on the many paths that could have happened? A favorite trading quote: “do not develop for the knowable past but rather for the probable future.”

🌐 Alternate Realities

Think of these paths as parallel universes. In our universe, prices behaved how they behaved. But alternate realities could have existed and their prices may have played out differently. Furthermore, we may soon jump to one of these realities. What if the future takes path #37 but you only trained on the single historical green line?

REAL

🌍

Our Universe

The actual historical path. All-time highs at the known dates.

🪐

Universe #2

Noise-adjusted. Earlier ATH, different volatility clusters.

🌀

Universe #3

MC permutated. Same endpoint, entirely different path.

✨

Universe #4

Vs Shifted. Same market, different bar boundaries.

By creating strategies across multiple data series, we aim to capture the underlying signal rather than the noise (randomness) in any one specific universe.

Pre-Simulation vs. Post-Simulation

Build Alpha has always enabled traders to use synthetic data to test strategies from a robustness testing standpoint (post-simulation). Now you can include this data before the simulation engine or genetic algorithm begins — for the strategy discovery process itself.

Post-Simulation (Traditional)

Test After Building

Build strategy on historical data first, then validate by re-trading on synthetic series.

• Noise Test — detect overfitting
• Monte Carlo — stress drawdowns
• Strategy already locked in

Pre-Simulation (This Feature)

Build Across Synthetic Data

Add synthetic series to the genetic algorithm. Strategies trained across all paths simultaneously.

• Discovery accounts for alternate realities
• Rules must work across the multiverse
• Fundamentally more robust from the start

Both approaches have value, and using them together is ideal. Build on synthetic data first, then validate survivors with post-simulation Noise Test, Monte Carlo, and out-of-sample testing.

Synthetic Data Methods

How can we create synthetic data? Below are the methods Build Alpha supports, ranging from simple to advanced. Building strategies on a variety of methods diversifies away pitfalls from relying on any single generation approach.

⏱️

Vs Shifted Data

Shifts bar open/close times by 1, 2, 3+ minutes — creating data with the same characteristics but different OHLCV values, patterns, highs/lows, and indicator calculations.

Method: shift bar boundaries ±N min
Control: shift amount, any timeframe
Result: same market, different bars

📊

Noise-Adjusted Data

Randomly adds or subtracts volatility (as a % of ATR) to each bar, creating thousands of new synthetic price series. Helps avoid fitting to the specific noise of one data set.

Method: ±% of ATR → O, H, L, C
Control: bars affected + magnitude
Result: same structure, different noise

🔀

Monte Carlo Permutation

From Timothy Masters’ book — reshuffles the log of inter and intrabar price changes then exponentiates back into raw OHLC. Retains statistical properties but destroys patterns.

Method: shuffle log price increments
Control: permutation parameters
Result: same distribution, new path

Vs Shifted in Detail

Standard 15-minute data creates bars at 0:00, 0:15, 0:30, 0:45. But what if we shifted to 9:03–9:18, 9:18–9:33, etc.? Or 9:06–9:21? Each shift creates a new data series with the same characteristics but different OHLCV values. Patterns are the same but the specifics are not:

Noise-Adjusted in Detail

Noise-adjusted data randomly adds or subtracts volatility to each bar. The animated example below shows how adjusting the noise alters the same data — each frame is a different noise-adjusted series. For deeper dives see: Noise Test — When OOS Is Not Enough, Noise Test Strategy Example, and Noise Test Parameter Optimization.

Monte Carlo Permutation in Detail

From Timothy Masters’ Permutation and Randomization Tests for Trading System Development — the MCP method reshuffles log inter and intrabar price changes, then exponentiates them back into raw OHLC prices. The result: data with the same statistical fingerprint but completely different price paths.

You can read more about this test and other Monte Carlo methods in the complete Monte Carlo Simulation Guide which includes a free simulator.

Advanced Machine Learning (GANs & Deep Learning)

Many machine learning methods lend themselves to generating synthetic data. Generative Adversarial Networks (GANs) and deep learning approaches can learn the distribution of historical price data and generate entirely new series that exhibit similar statistical properties — regime clustering, fat tails, volatility persistence — without being a transformation of the original data. These are frontier methods that go beyond simple permutation or noise injection.

🧠 GANs for Synthetic Data

GANs train two neural networks against each other: one generates synthetic data, the other tries to distinguish it from real data. The result is synthetic price data that is statistically indistinguishable from actual market data — a powerful way to expand your training set beyond simple transformations of the historical path.

Custom Data Import

Build Alpha also supports custom data imports — if you have an alternative generative method (your own GAN, deep learning model, or any other approach), you can add your own synthetic data to the build process via CSV or TXT file.

Why Not Just Use Other Markets?

You can! Building strategies on other markets is another great idea, discussed in Testing Across Markets and Trigger Signals. However, other markets often have different players, point values, volatility ranges, and hedging use cases. Despite heavy correlations, the underlying pattern may not exist across markets — resulting in an insufficient test.

The best approach combines all available methods in your portfolio construction:

• A single market
• A single market plus related markets
• A single market plus unrelated markets
• A single market plus synthetic data
• A single market plus related markets plus synthetic data

There is no single best approach. We constantly aim to equip ourselves with robust strategies built on different methods to best handle changing markets.

Setting Up Synthetic Data in Build Alpha

After selecting the market you want to trade (can be multiple markets), hit the Additional Data button to create as many synthetic data series as needed. Below shows SP500 Emini Futures with 9 additional data series — three from each method: Vs Shifted, Noise Adjusted, and Monte Carlo Permutation.

📝 Note on Multi-Symbol

If you select multiple symbols (e.g. @ES and @GC), the above configuration creates 18 synthetic data series — 9 for each symbol. The synthetic data scales with your symbol universe automatically.

Basket (BSK) Results

After simulating with added synthetic data, Build Alpha’s results window shows ‘BSK’ results — the strategies that performed best across the entire basket. Expand the BSK row to view how the strategy did on each individual data series:

Basket Equity Curve Graph

Alternatively, view the equity curve and hit the ‘Basket’ button to overlay the selected strategy’s performance on each synthetic series. Each generation method has its own color:

Orange = Vs Shifted

Blue = Noise Adjusted

Purple = Monte Carlo Permutation

Custom Rules: Filter Your Multiverse

Build Alpha’s no-code rule builder lets you qualify which synthetic series make it into training. Set rules so the data matches the stress scenarios you want to prepare for:

📐

Correlation Filter

Only use series with 0.25–0.75 correlation to original

📉

Larger Drawdown

Synthetic must have deeper drawdowns than history

⏳

Slower Recovery

Must take longer to recover from max drawdown

🌊

Higher Volatility

Require increased volatility regimes in synthetic data

🔗

Combine OR

Chain rules with OR — each gets its own color on the chart

📊

Summary Stats

View how each rule performed independently below the display

Synthetic Data in Walk-Forward Optimization

Build Alpha’s walk-forward optimizer can inject synthetic data into the process. Select Simulate with Synthetic Data, choose the symbol, timeframe, and data-generation methods, then set qualifying rules.

For example, Build Alpha creates 200 synthetic series but only includes those with correlation between 0.25 and 0.75 to the original. The synthetic results are included in pass/fail scoring — parameters must work across both the historical path and synthetic series to pass. See also Noise Test Parameter Optimization which leverages synthetic data for parameter selection.

✓ Why This Matters

Market shocks are what gets system traders in trouble. By generating synthetic data with your desired “shocks,” you can stress-test your strategy against the market you fear before it happens. This is preparing for the probable future, not optimizing for the certain past.

Integrating with the Full Workflow

Generate strategies across historical + synthetic data → Filter by OOS performance → Validate with Noise Test + Monte Carlo + Vs Random → Optimize parameters with NTO → Walk-forward with synthetic data → Deploy via automated code generation.

Build Alpha v3’s automated workflows run this entire pipeline. Set filters, hit simulate, and only see strategies that survive everything — including training across the multiverse.

Takeaways

✓ Key Takeaways

1. Training on synthetic data is pre-simulation — fundamentally different from post-simulation robustness tests
2. Three built-in methods: Vs Shifted, Noise Adjusted, Monte Carlo Permutation — plus GANs, deep learning, and custom imports
3. BSK results show performance across all paths; color-coded equity curves reveal method-specific behavior

💡 Need to Know

1. Building on a variety of synthetic methods diversifies away pitfalls from any single generation approach
2. Synthetic data integrates with walk-forward optimization — results count in pass/fail scoring
3. Combine with the full robustness testing suite for maximum confidence before live deployment

Synthetic Data FAQ

What is synthetic data in algorithmic trading?

Synthetic data refers to artificially generated price series that differ from actual historical data. Methods include noise adjustment, Monte Carlo permutation, Vs Shifted, GANs, and custom imports. They represent alternate realities — plausible price paths the market could have taken.

How is building on synthetic data different from the Noise Test?

The standard Noise Test is post-simulation — build first, then re-trade on noise-adjusted data. Building on synthetic data is pre-simulation — the genetic algorithm trains across the historical path AND synthetic series simultaneously from the start.

What data generation methods does Build Alpha support?

Three built-in methods: Vs Shifted (shifts bar boundaries), Noise Adjusted (adds/subtracts volatility), and Monte Carlo Permutation (shuffles log price increments). Custom data can be imported for advanced methods like GANs or deep learning.

Can I filter which synthetic data sets are used?

Yes. Build Alpha’s no-code rule builder lets you qualify synthetic data by correlation, drawdown depth, recovery time, volatility, and more. Rules can be combined or evaluated independently with color-coded results.

Can synthetic data be used in walk-forward optimization?

Yes. Build Alpha injects synthetic data into the walk-forward process. Synthetic results are included in pass/fail scoring — parameters must work across both the historical path and synthetic series to pass.

Related Deep Dives

📈 Noise Test🔧 Noise Test Parameter Optimization🔄 Walk-Forward Optimization🎲 Monte Carlo Simulation📊 Out-of-Sample Testing🌐 Testing Across Markets

David Bergstrom

Founder, Build Alpha

A decade-plus in professional trading as a market maker and quantitative strategy developer at a high-frequency trading firm with a CME seat. Self-taught C++, C#, and Python programmer specializing in data science, machine learning, and trading strategy development.

What Traders Say About Build Alpha

Train for the Unknown Future

Build strategies across hundreds of alternate price paths. Three generation methods. Custom stress rules. No coding required.

Get Build Alpha Watch Demo

Portfolio trading is combining many independent, uncorrelated trading strategies into one shared account to gain some net benefit from their combination. One should design automated trading systems to handle and capture different market effects and environments to maximize this benefit.

Next, one needs to figure out how to combine these trading strategies to best improve the overall risk-adjusted return (or whatever your portfolio’s objective may be). This is [successful] system trading 101.

A Portfolio backtest splits the starting capital among the various individual trading strategies to see how the interplay of strategies affects the shared capital pool (aka our account).

In this post, I will discuss how we rebalance capital, why we rebalance capital, when to rebalance capital, and why we need many uncorrelated trading strategies vs. a single “holy grail” strategy.

Why Portfolio Trading?

There is no “holy grail” strategy. Sorry. I have been using machine learning and artificial intelligence for a decade-plus and have yet to find it. What’s worse is that over-optimizing, constant tweaking, and adding complex rules produce great-looking backtests which fail in live markets. Trust me.

The real “holy grail” is combining multiple strategies into one capital-aware portfolio.

The goal isn’t one perfect strategy — it’s a team of [robust] strategies that work well together. Some perform in bull markets, others in bear markets; some thrive in calm conditions, others in crises. Since no one knows what tomorrow brings, your portfolio should be prepared for all environments.

To do this, strategies must be different from each other. If they’re too similar, you’re just taking the same risk twice while adding operational complexity (yuck). We measure difference with correlation: ideally, when one strategy zigs, another zags — like pistons in a car engine (some up, some down but car moves forward).

Mixing strategies across assets, timeframes, frequencies, and styles (effect capture) builds a far more robust portfolio.

Combining Strategies to Improve a Portfolio

Below are three simple strategies that no one would trade by themselves. However, the historical drawdowns of these three strategies do not happen at the same time so when the strategies are combined, we notice improved performance (green equity curve).

And notice the drawdowns! On its own, each strategy experienced a 40%+ reduction from its peak equity at one point, but the combined hypothetical example never experienced even a 5% haircut.

This, generally speaking, is what we aim to achieve with portfolio trading. The coveted “holy grail”.

Parrondo’s Paradox in Trading

Wikipedia defines the Parrondo’s Paradox in Game Theory as “a combination of losing strategies that becomes a winning strategy”. Below is a brief video explaining Parrondo’s Paradox and loosely explains what system traders aim to achieve by combining uncorrelated strategies together.

This video begs the question: can we combine two bad strategies into a winning strategy?

The answer is (mostly) yes, but it is not done through simply combining two bad strategies together. We cannot just mash any two strategies together and voila.

An example portfolio

Trading is never that simple. Take this simple proxy portfolio

• Trend Following (TF) proxy: 20% win rate, 8% average win, -2% average loss
• Mean Reversion (MR) proxy: 80% win rate, 4% average win, -15% average loss (occasional crash)

Both of these systems are losers on their own (negative geometric return). And if we give each 50% of our capital and mash together as a portfolio we still lose (green line).

But wait…

Portfolio Rebalancing is key

Rebalancing our capital every day turns the combined equity curve into a winner (red line). The “daily” 50/50 rebalancing mechanically sells what just went up and buys what just went down.

When strategies are sufficiently uncorrelated or negatively correlated, and volatile enough, this creates a rebalancing premium (sometimes called Shannon’s Demon).

The effect or positive return comes from harvesting the volatility when we rebalance the capital. In quant speak, the geometric return is roughly the arithmetic return – 1/2 of the variance.

As we combine strategies (and rebalance the capital) the portfolio variance drops a lot more than the arithmetic mean does. This can flip the geometric mean positive!

In English please…

That is, we must pool the capital together and reallocate periodically in order to experience Parrondo’s Paradox in our trading accounts.

Not so fast

So just rebalance every day? No, there’s trading costs. And why 50/50 rebalancing? And what about correlation? And what if mid-trade our 50/50 weighting becomes 70/30? And what if, and what if, and what if, and.. hold on.

Rebalancing isn’t magic. It only helps when the strategies take turns leading (low/negative correlation) and have enough volatility/dispersion to harvest — and even then, trading costs can erase it.

Rebalancing can also hurt when one strategy is the clear long-run winner: you’re constantly selling the winner to buy the laggard, which reduces terminal wealth. We rarely know if we’ll have one long-run winner or which it will be, though.

In other words, if the strategies don’t “take turns” (correlated or very different long-run growth rates), frequent rebalancing can act like a “sell your best idea” tax.

In short, it is vital to combine uncorrelated, unrelated strategies to capture this effect! Not similar strategies.

So where does that leave us now?

New Questions

We know we want to combine unrelated strategies but now have new variables and a bunch of new questions. For instance, we now ask:

• what strategies to combine
• what weights to give them
• what correlation is acceptable
• how to size strategies
• show we rebalance capital
• how often to rebalance
• what about in-trade adjustments (e.g., 50/50 becomes 70/30)

Yep, an explosion of new things to test. “Hey, I’m barely working my way through my strategy ideas and now I have to answer portfolio questions? Come on, man. Can BA help?”

Build Alpha Portfolio Mode

Build Alpha’s portfolio mode enables traders to save strategies, combine them, analyze asset allocation methods, optimize rebalancing frequencies, and more. Let’s go through some of the core features below.

Fundamental Data and Factor Analysis

Build Alpha signed a data deal with the Nasdaq to offer fundamental data for building trading and investment strategies. This allows equity traders to filter symbol universes to only trade when Free Cash Flow or Revenues or Price to Book or PEG ratio or any other fundamental data point is to your liking.

This also enables Build Alpha to build and test various factor models such as market, size, value, profitability, investment, momentum, etc. The price data provided by Nasdaq is split and dividend adjusted.

Combining investment strategies built on different factors is one of the most time-tested methods for portfolio construction.

Custom Filters and Ranks

Filters are rules of what is trade-able and what is not. For example, only trade stocks or ETFs with an average daily volume greater than $10,000,000 or only trade futures with an average true range (ATR) at or above the 70th percentile or growth mutual funds above $5 or whatever.

Ranks are how you determine what symbol gets an allocation when you have more entries than available slots.

We can rank on anything: relative strength, volatility, market cap, momentum or any custom ranking you can imagine. Use the no-code editor to build rankings and review current rankings using the Rank Inspection tool (pictured below).

Create a library of filters and ranks, and let Build Alpha show which ones work best for your strategy.

The results above show different combinations of filters and ranking methods applied to the core strategy. The selected row shows how the strategy performs when stocks under $5 or below $500M in market cap are excluded, and the remaining stocks are ranked by 252-day rate of change.

Portfolio Analysis and Visuals

Build Alpha portfolio mode comes with all standard visuals and metrics such as equity curve comparison to buy and hold, performance metrics, correlation matrix, minimum variance test, and the institutional-grade robustness tests you all know from Build Alpha. Below shows an equity curve and correlation matrix.

A Quick Note About Other Portfolio Backtesters

Most portfolio tools just merge trade lists (txt or csv files) from multiple strategies. This cannot model real portfolio behavior because it does not know how trades would change when strategies share capital.

Example: a strategy that sizes positions based on stop-loss distance might only take 3 trades in its own backtest because capital was fully used. In a real portfolio, profits from other strategies could free up enough capital to take a 4th trade—but a trade-list simulator will never add this 4th trade, so the results are unrealistic.

Build Alpha solves this by re-simulating the strategies together, tracking capital in real time and letting strategies interact the way they would in live trading—something static trade lists can’t do.

Build Alpha – An Adanced Portfolio Backtesting Tool

Build Alpha is built for both independent traders and professional managers (CTAs and RIAs) running multiple strategies across multiple client accounts. It’s flexible enough to support a wide range of workflows and requirements—while also avoiding the common portfolio-simulator pitfalls mentioned above.

Backtesting Portfolio Strategies with Various Exit Strategies

Many equity investors and Registered Investment Advisors utilize strategies that rank a symbol universe periodically and only trade the top N signals. However, what happens when a position grows to be too large a percentage of your portfolio? Should we immediately reduce size? Or what happens if the stock falls out of the top N ranked stocks? Should we immediately close the position and add a position in the newly ranked stock?

New questions mean new tools.

Build Alpha has various exit modes for portfolio strategies to give traders flexibility:

(Terms: Standard exit = stop-loss, trailing stop, or signal exit. Rank = position’s standing vs your universe. Fit to target = enforce position-size target E (e.g., 10%). “Down only” trims but never tops up.)

• Standard — Close the position only on a standard exit. Ignore rank and size while held. Lets winners run (no trimming if it appreciates or drops in rank).
• Respect Rank, No Fit — Close on any standard exit or when the stock falls out of the top N (your max positions). Example: Max positions = 10; if a holding becomes rank 12, it’s closed.
• Respect Rank, Fit to Target Down — Same as above plus enforce max exposure E: trim if the position grows above E; do not add if it falls below E.
• Respect Rank, Fit to Target — Same as above, but rebalance both ways: trim if >E and top up if <E.
• Ignore Rank, Fit to Target Down — Close only on standard exit; ignore rank changes. Trim positions that exceed E; never add when below E.
• Ignore Rank, Fit to Target — Close only on standard exit; ignore rank. Rebalance to E whether the position is above or below target.

These provide the trader or investor the ability to create constant volatility targeted portfolios or ones that hold for the long-term aiming to capture large trends. This flexibility is paramount with many use cases.

In-Trade Adjustments and Rebalance Periods

We have all been in a trade when volatility drastically changes. That is, we may size based on current volatility but a few bars later the volatility has collapsed, and we do not have enough exposure or a few bars later volatility explodes, and we have too much exposure!

Build Alpha allows traders to enable In-trade adjustments where Build Alpha will re-size positions if they deviate from the desired or current size by a certain percentage threshold. Below would re-size a position whenever the current exposure deviates 10% or more from your desired position size.

This constant or continuous adjustment helps smooth returns and helps traders and investors target certain volatility levels.

This is akin to a Hold ’em Poker player changing his bet as new cards (data) are unveiled. A poker player doesn’t determine his bet size at the beginning of the hand and wait until the end of the hand (trade’s exit). He adjusts his exposure mid-hand, multiple times often. However, in trading, we can adjust our exposure both up and down – how cool!

Here is an example TSLA trade with many in-trade adjustments (buys and sells to resize the position as needed).

Walk Forward Test to Backtest Portfolio Strategies

Walk Forward Testing splits the historical data into multiple training and testing partitions where we optimize strategy parameters during the training data (in-sample data, blue below) and test the best parameter settings on the unseen (out-of-sample, green below) data.

This process is repeated until we “walk forward” through the entire historical data set which ultimately allows the trader to combine all out-of-sample data (green boxes) into one truly out-of-sample equity curve.

Build Alpha enables you to do this with a single strategy or a portfolio of strategies and you get this results window which shows performance and risk metrics for every out-of-sample period for the portfolio. Expand any row and view the individual strategies’ performance during any OOS period.

One can gain insights into how often one should re-optimize a strategy or how much ideal parameters have/do change over time. I wrote more about Walk Forward Optimization here.

Portfolio Suggest – AI Asset Allocation

Build Alpha’s Portfolio Suggest feature is a single button that enables Build Alpha’s AI, that was trained on thousands of strategies and market environments, to sift through your saved strategies and suggest the best sizing method, exit modes, combination of strategies, weightings, and more so you do not have to test every combination yourself.

Below are results of the Portfolio Suggest feature. The blue lines are various combinations, weightings, sizing methods that Build Alpha simulated based on the saved strategies in the portfolio. The red line is Buy and Hold (S&P 500).

See what settings were applied to create each equity curve. The Generate button opens a new results window for the selected settings for further analysis.

Position Update Report

Build Alpha runs a daily report to update all your saved strategies. This overview gives an update on every saved strategy. The report includes:

• What to enter
• What to exit
• Current Positions
• Current Rankings
• Open P&L
• Funds Available
• Total Exposure

This is the simplest way to track portfolio strategies and get daily updates. This is also for all of the new Build Alpha members that are not ready for full automation! The data updates overnight so this feature is best run before the US market opens.

Multiple Portfolios in one place

Color code, name, and organize all your saved strategies into different portfolios. This gives a one screen summary to view risk metrics, compare strategies, and keep tabs on all your hard work. There is a dropdown in the upper left that will allow you to move between each portfolio or view the combined Common Portfolio (pictured below).

Build Alpha Portfolio Mode

What’s Included

• Multi-strategy portfolio simulation
• Portfolio Suggest AI
• Multi-strategy walk-forward testing
• Portfolio filters and ranks
• In-trade volatility adjustments
• Fundamental data access (Nasdaq)
• Ranking inspector
• Save multiple portfolios
• Daily position update report

Please email me with any questions and check out the Build Alpha V3 Update to see what else is new in the latest release.

The Commitment of Traders Report is a weekly report released by the Commodity Futures Trading Commission (CFTC) that shows the open interest for futures and options on futures in markets with 20 or more traders holding positions above or equal to the reporting levels established by the CFTC. The report shows open interest data as of Tuesday but is not released until Friday afternoon.

The reports are based on open interest position data supplied by reporting firms such as FCMs, clearing members, foreign brokers and exchanges. The data is categorized by the trader’s dominant business purpose self-reported by traders on the CFTC Form 40 and is subject to review by the CFTC.

The report is broken down into three different types of traders:

• Commercial traders (hedgers): Traders in this category are typically considered to be market professionals, such as producers, processors, users, merchants, and swap dealers. These traders use futures markets to hedge against price changes in the underlying commodities. They are considered to have a direct commercial interest in the underlying commodities.

• Non-commercial traders (large speculators): This category of traders is considered to consist of speculators who trade futures contracts for financial purposes, rather than to hedge against price changes in the underlying commodities. These traders include hedge funds, Commodity Trading Advisors (CTAs), individual traders, and other investment firms.

• Nonreportable (small traders): These traders are defined as those holding positions below the reporting levels established by the CFTC. Traders in this category are considered to have less impact on the market and provide valuable insight into market sentiment and behavior. Their positions are classified as non-reportable positions in the COT report.

Smart Money vs Dumb Money

The COT report is broken down into three categories of traders and often the commercial traders and to a lesser extent the non-commercial traders (large speculators) are considered “smart money”. The nonreportable category is often called small speculators or retail traders which unfortunately gets the moniker “dumb money”.

Watching the divergence between smart and dumb money can often leave clues to direction or market sentiment which can help influence trading approaches. The most popular COT trading style is watching when these two categories diverge to extremes as large divergences between smart and dumb money is typically when or how the market moves. Someone has to be someone’s liquidity!

Is there Free COT Data?

Yes, COT data is freely available on the CFTC’s website at cftc.gov and then navigating to the Commitment of Traders report link here: Commitments of Traders | CFTC

However, the historical data from COT reports are not formatted for easy use and often needs modification to make useful futures markets signals. More on how Build Alpha does this later.

How can traders use COT data?

Simply using the commitments of traders reported data as a trade signal is not a great idea! The data is shared as long and short positions and is not normalized for Tuesday’s open interest. Instead, the values are just raw values and rather meaningless on their own. For instance, is a number of 10,374 contracts long a lot or a little? Well, it depends on what open interest is! And how many contracts are short!

There are two simple ways of deriving signals from COT data. First, the normalized net position which is expressed as a percent. Positive 20% would mean the net position is long and 20% of open interest. The second way is the COT Index.

What is the COT Index?

The COT Index or Commitment of Traders (COT) Index is a popular trading indicator showing where current net positioning is in relation to historical positioning of the same trader group over the past three years (36 months).

A COT Index reading of 100 means the group is the most bullish compared to any point in the past 36 months and a reading of 0 means the group is the most bearish compared to any point in the past 36 months. The COT Index can be calculated on Commercials, Non-Commercials, or Non-Reporting traders.

Above is an image showing the SP500 price fluctuations with the non-commercial trader (large speculators) positioning via COT Index over the past 36 months. Traders can isolate periods of extreme positioning to identify the start or end of a trending environment.

In Build Alpha, traders can test any COT Index value or range of values from the point-and-click interface. Here’s a sample strategy that goes long $10,000 per trade in QQQ (Nasdaq ETF) when Nasdaq futures non-commercials COT Index is less than or equal to 50. Simply being long QQQ when large speculators are less bullish (blue line) has drastically outperformed QQQ buy and hold (red line) over the past 20 years.

This signal can of course be improved by requiring this signal and having Build Alpha search thousands of other signals to find complimentary filters.

Simpler COT Trading Signals

The simplest COT trading signal is to look at the net position – subtract the shorts from the longs – so that a positive number means that particular category of trader is net long and a negative number means that category of trader is net short. If the net position is zero, long and short traders’ positions reported equal each other. It is common when there is no underlying bias or directional sentiment that traders hold positions equal to each other or close to it.

Further, we can then divide the net number by the total open interest to express a normalized value that can provide apples to apples comparisons over time or across markets. A value of -0.20 would be a 20% net position or 20% of open interest is positioned short.

COT Report Trading Signals in Build Alpha

Build Alpha has COT data back to 2000 included in the 5,000+ built-in signal library. You can search for them in the top filter using “COT” as pictured below.

These signals can be used for entries, exits or general strategy filters. Build Alpha has COT data for the following markets:

• S&P500
• Nasdaq
• Dow Jones
• Vix Futures
• AUDUSD
• Dollar Index
• EURUSD
• 2-Year Note Interest Rates
• 5-Year Note Interest Rates
• 10-Year Note Interest Rates
• 30-Year Bond Interest Rates
• Natural Gas
• WTI Oil
• Silver

COT Report Trading Strategy Example

Index traders take a look at this one! This simple SP500 futures market strategy demonstration was built using Build Alpha’s COT signals. I only used the COT signals and did not use any other price, volume or non-price based entry or exit signals. This strategy enters whenever the following is true:

• Nasdaq Commercials Net positioning 15% or less of open interest
• Five-Year Note Commercials Net positioning 15% or less of open interest
• AUDUSD Commercials Net positioning 10% or less of open interest

The strategy simply exits after one bar. Its performance below is from 2007 through 2023 using only one contract per trade.

COT Indicator or Filter

A better use of non-price data like COT report or other sentiment data is often to use the data as a filter to a price action or price and volume-based strategy. Price strategies often lack context about the market environment and as we all know patterns and signals act differently in different market conditions.

COT data and other non-price based data acts as a way to relay information about the environment to our trading systems.

Below is a simple example of how to find or improve a simple price action strategy with COT report data as a filter using the required signal functionality in Build Alpha.

Cool Ideas for using COT Data with BA

You can pair COT data with economic news events. For instance, perhaps it is only worth trading a news event when smart money is in the same direction. Perhaps intraday trading systems could benefit from a filter from larger players.

Summary

The Commitment of Traders COT Report is a weekly report released by the Commodity Futures Trading Commission which discloses open interest and market positioning in futures markets and options on futures markets for three different types of traders: Commercials (hedgers), non-commercials (large speculators), and nonreportable speculative traders (small trader).

The data gives insights into how market participants are positioning for future trends and give an asset manager or commercial traders an idea on market timing or pending market pressure. The report is released Friday but only contains data up until Tuesday.

Traders reports and other non-price-based data can be valuable contextual information to relay market environment or regime details to automated trading strategies. Build Alpha now contains COT report data for 15+ futures markets that can be used to create automated trading strategies without any coding.

Author

David Bergstrom – the guy behind Build Alpha. I have spent a decade-plus in the professional trading world working as a market maker and quantitative strategy developer at a high frequency trading firm with a Chicago Mercantile Exchange (CME) seat, consulting for Hedge Funds, Commodity Trading Advisors (CTAs), Family Offices and Registered Investment Advisors (RIAs). I am a self-taught programmer utilizing C++, C# and python with a statistics background specializing in data science, machine learning and trading strategy development. I have been featured on Chatwithtraders.com, Bettersystemtrader.com, Desiretotrade.com, Quantocracy, Traderlife.com, Seeitmarket.com, Benzinga, TradeStation, NinjaTrader and more. Most of my experience has led me to a series of repeatable processes to find, create, test and implement algorithmic trading ideas in a robust manner. Build Alpha is the culmination of this process from start to finish. Please reach out to me directly at any time.

What You’ll Learn

• How to generate NinjaTrader strategy code with Build Alpha
• How to import/compile in the NinjaScript Editor
• How to verify backtest matching and avoid common mismatches

How to Download NinjaTrader

First, create a free NinjaTrader account at NinjaTrader.com and download the NT8 NinjaTrader8 platform. Often Ninjatrader.com needs to review the security of your connection and may request you verify you are human by completing a simple captcha.

NinjaTrader 8

Currently NinjaTrader is on version 8. Build Alpha has supported NinjaTrader since NinjaTrader7 but now proudly supports Ninjatrader8. Make sure you download the latest version for the best experience.

NinjaTrader Brokerage Account

Ninjatrader is also their own broker, meaning you can open a real money account and trade directly through the Ninjatrader platform.

NT8 is a large reputable futures broker and were recently acquired by Kraken in a $1.5B deal. You can read about the deal here: Kraken to Acquire NinjaTrader: Introducing the Next Era of Professional Trading | NinjaTrader I imagine the trend of merging traditional finance and crypto will continue.

Futures Trading NinjaTrader

NinjaTrader is arguably the best futures broker available at the moment for retail traders and serious independent traders. NinjaTrader has a wide variety of futures data connections and has competitive margin requirements. You can trade all assets through NinjaTrader but it is particularly good for futures traders.

NinjaTrader Margin Requirements

NT8 has competitive margin requirements which can be found here: Margins for Available Futures Markets | NinjaTrader | NinjaTrader

NinjaTrader Automated Trading Systems

NinjaTrader also permits full automation of trading strategies using their proprietary NinjaScript programming language. The programming language is C# based and extremely flexible.

There are tons of existing functions, tools, indicators and strategy logic blocks for those that want to dive into the C# code. For those that do not want to write their own code, there is of course Build Alpha!

NinjaTrader Automated Trading Systems with No Code

Build Alpha generates fully operational NinjaTrader code at the click of a button with perfectly matching results across the two platforms. Build Alpha enables traders to build and execute their own trading ideas with no code and then easily transfer the generated code over to NT8 for live (or simulated) trading.

Build Alpha vs NinjaTrader Strategy Builder vs Coding

NinjaTrader has their own basic Strategy Builder which may be good for basic rules but can be quite limiting. It also only tests a single idea at a time and has no artificial intelligence to help arrive at better strategies for you.

NinjaScript – NinjaTrader’s proprietary coding language – gives maximum flexibility but requires coding time. It is C# based so it is not trivial for a non-coder to pick up.

On the other hand, Build Alpha generates thousands of strategies per second improving them using a genetic algorithm and artificial intelligence (my background in machine learning). BA also has the most advanced suite of robustness tests and completely exportable NinjaScript code.

How to Automate Build Alpha strategies in NinjaTrader with no coding

First, let’s generate a simple strategy so we can talk about the same example. I selected a few thousand signals in Build Alpha, hit simulate, and selected a random strategy to review.

Next, hit the ‘Generate NT Code’ button in the lower right of the Strategy Simulation results window pictured above. NT stands for Ninjatrader8. Please note there are also Build Alpha code generators for

• Tradestation

• Multicharts

• Metatrader4/5

• TradingView

• Python (Interactive Brokers)

• ProRealTime

Below is a picture of the Build Alpha generated code for NinjaTrader8. Copy all the code by hitting CTRL + A (to highlight all) and CTRL + C (to copy all).

In NinjaTrader, go to New >> NinjaScript Editor

In the NinjaScript Editor window, find Strategies and right-click to create a New Strategy.

The Strategy Wizard will open. Give your strategy a name and hit Generate. I chose ‘MyCustomName’ as the name for my strategy.

In the new window that opens, hit CTRL + A to highlight and then Delete to remove all default strategy text provided by NinjaTrader.

After completing the action, simply copy and paste the Build Alpha code into this blank file. Next, rename ‘SlaveY’ class name and ‘StrategyName’ to ‘MyCustomName’ which was the name I provided for the strategy – which can also be seen in the bottom tab of the NinjaScript editor (all in orange boxes below). Finally, compile the strategy by clicking the button in the green box at the top of the second image below. This saves your strategy in Ninjatrader8.

Finally, open a NT8 chart and add the strategy to the chart (enable it in NT8 settings). Below I’ve added our strategy onto the SP500 Futures emini chart and pulled up the NinjaTrader strategy performance report to show the backtest results matching in both Build Alpha (left) and NinjaTrader (right).

Reminder: Please check your data connection before proceeding from the NinjaTrader Connections menu or you will not be able to load a strategy onto a chart.

Build Alpha’s generated code will include position sizing contracts, time of day, market conditions, entry and exit signals, risk management, session information, etc. In other words, the complete strategy developed or produced in Build Alpha with all settings you decide to use will be reflected in the generated NT code.

Market Data, Trade Entries and Trade Execution

Now that you have a strategy loaded on a NinjaTrader chart you can watch the live market data and the strategy will show live trade executions (sim or live) on the NinjaTrader chart. You have now built and automated a trading strategy in Ninjatrader8 with no coding using Build Alpha.

NinjatraderEcoSystem – Proud Partner

Build Alpha is also a proud member of the NinjaTrader EcoSystem. Meaning NinjaTrader has reviewed and approved Build Alpha as certified member of the NT8 developer community.

To learn more about Build Alpha please check these resources

Everything you need to know about Monte Carlo Simulations

Failing strategies? Read the Robustness Testing Guide

NinjaTrader Automation FAQ

Does Build Alpha generate NinjaTrader8 strategies or indicators?

Build Alpha generates fully tradeable strategies. However, the generated code can be modified to serve as an indicator, if desired.

Do the results match exactly?

If you use the same data in both platforms, session templates, transaction costs then yes, the results will match. An example is pictured above.

Can the generated code be used to paper trade?

Yes, you can use the generate code for both paper trading and live trading.

Interactive Brokers ($IBKR) is both a retail and professional trading broker and software platform that can be downloaded directly from the Interactive Brokers website assuming you have an existing Interactive Brokers account. If you do not have an existing IBKR account create one here: Open Interactive Brokers Account

If you have an account, please download and install their software for desktops here: Interactive Brokers Download

To learn how to navigate the IB platform, check out this video: IBKR Short Video – TWS for Beginners – Getting Started

Finding Trading Strategies with Build Alpha – No Code

Build Alpha is a no code strategy builder that enables traders to build, backtest, and generate code for tons of automated trading strategies. This presents an option for those that do not want to learn how to program but still want to do algorithmic trading.

Generate Python Code for Interactive Brokers with Build Alpha

After generating automated strategies, simply highlight the trading algorithm you like and hit the ‘Generate Python Code’ button to get the full python code to use with Interactive Brokers. Later I will show step by step how to automate trading with Interactive Brokers using this generated code.

Build Alpha currently generates code for the following platforms

• Tradestation
• Ninjatrader
• Multicharts
• Metatrader4
• Metatrader5
• Interactive Brokers (python)
• Pro Real Time
• TradingView

Additional Notes about Build Alpha

For those interested in adding custom signals to the existing 7,000+ built-in signal library, Build Alpha has both a no code custom editor and the ability to add custom signals via python. Build Alpha also has tons of alternative, non-price-based data to add context to automated trading strategies. Read more about that here: Build Alpha Breadth and Sentiment Data

To see more Build Alpha features please click here: Build Alpha features

Automate Trading with Interactive Brokers Python API

The generated strategy code will be able to access live and historical data to assess trading opportunities, calculate position sizes, weigh available capital, determine new buy and sell decisions, and ultimately send orders (market, bracket orders, advanced order types) directly to IBKR for execution for paper trading or live trading.

Step One: Configure IBKR TWS

In your Interactive Brokers Traders Workstation please go to Settings >> API >> Settings and do the following

• Enable ActiveX
• Untick Read-Only API
• Make sure socket port is 7496 (live) or 7497 (simulated)

Step Two: Configure the Python code

In Build Alpha, select the strategy you want then hit Generate Python Code (instructions above). Next, go to Tools >> Python Development Environment (pictured below)

Then copy and paste your code into a new python file and save it. Make sure to modify any contract expiration dates or symbols that may be different then when you originally generated the code.

Step Three: Run the Python Code

Once your strategy is configured and saved in a new file, you can simply hit ‘Run Module’ or F5 and the code will connect to IB, start monitoring data and issuing orders based on your exact strategy logic and settings from Build Alpha. Not difficult and all with no code!

Step Four: Monitor Your Live Strategies

Once your strategy or strategies are running live then you can see your orders, resting limit order status, filled trades log, profit and loss, overall performance, and account changes directly through the visual interface provided by Interactive Brokers’ Trader Workstation (pictured below from Getting Started with TWS | Trading Lesson | Traders’ Academy).

All of this is enabled through Interactive Brokers by utilizing their application programming interface or API. The Build Alpha generated python code includes logic so your strategy can communicate with IBKR through the Interactive Brokers’ API.

Above is a brief IB API Python example but for full depth please consult the Build Alpha User’s Guide or the Interactive Brokers API Python Documentation here: IBKR TWS API Documentation

What is an API?

In simple terms, an API or application programming interface is a set of instructions that one code developer shares with another code developer so that their respective codes can effectively communicate with each other. To learn more about APIs please check this out: Wikipedia API page

Some APIs require payment or a private api key for use. Some APIs are free and do not require additional money. Interactive Brokers API is free and has good security.

More Info on Interactive Brokers

Interactive Brokers is a professional trading brokerage and has great documentation. For instance, their step-by-step guide for those that want to try connecting their own python scripts to IBKR is very robust: Interactive Brokers Python API (Native) – A Step-by-step Guide. Additionally, IB is great for a handful of other reasons and compliments Build Alpha’s advanced tools quite nicely.

Paper Trading with IBKR

All IBKR accounts have the ability for a paper account or practice account which is recommended when initially learning how to connect code, trying new strategies or even just getting familiar with a new brokerage platform. Take it slow and understand what you are doing before putting real capital at risk.

All Asset Classes and Timeframes

IBKR and Build Alpha both support all timeframes and asset classes in one place. Accessibility is key in today’s markets. You can track intraday data and place orders for stocks, etfs, futures, mutual funds, forex and even crypto currency through a single IBKR account. This is a large advantage for those that want to master systematic trading and expand their trading portfolio.

IBRK Stock

Interactive Brokers is a publicly traded company. It was founded in 1978 by Thomas Peterffy, who remains chairman. The company is headquartered in Greenwich, Connecticut and has over 2 million users and nearly 3,000 employees worldwide. The stock trades under the ticker $IBKR hence my common use of the IBKR abbreviation throughout this article.

IBRK margin rates

Interactive Brokers is often lauded for their low percentage margin rates which can be found here: Margin Rates and Financing | Interactive Brokers LLC. Please do your own due diligence and caution when utilizing margin. The rates do get lower as you have a larger account – some power in having more capital.

Does IBKR have a trading simulator?

No, Interactive Brokers does not have a trading simulator, but you can utilize their paper trading account with live data. However, Build Alpha allows traders to generate, backtest, and build individual strategies and portfolios of strategies on any asset class, timeframe and bar type. The only no code multi-strategy portfolio testing software.

Summary

There are many factors to consider when reviewing all current material and aiming to complete research on trading brokers such as Interactive Brokers. Do not immediately open an account or automate any trading at any broker without doing your own due diligence. However, please be aware that many professionals have at least one Interactive Brokers account due to its reliability, professionalism and financial stability.

Compare and contrast the value with the other brokerage platforms Build Alpha supports (Tradestation, Ninjatrader, Metatrader, etc.) in order to take full advantage of both the Build Alpha software and your fees when you execute trades in a live account.

Please reach out to me anytime through email or the contact form on the site regarding an Interactive Brokers account or anything else in this post.