CAR25 Trading: Risk-Normalized System Evaluation

By Dr. Ken LongDrawn from 30+ years of teaching

Most traders pick a system based on its compound annual rate of return and call it a day. That single number tells you how fast the account grew, but it says nothing about the pain you endured to get there. A system that returned 40% per year but cratered 60% along the way is not the same animal as one that returned 25% with a 10% peak-to-trough drawdown. CAR25 trading solves this problem by giving you a single, risk-normalized profit metric that weighs return against the realistic worst-case drawdown at the 95th percentile of a Monte Carlo distribution. Developed by Dr. Howard Bandy as part of his integrated approach to quantitative technical analysis, CAR25 forces you to ask the only question that matters: how much did you earn relative to how much you could have lost?

If you have ever watched a backtest produce eye-popping returns only to see the live account bleed capital for months, you already know why raw return numbers lie. CAR25 reframes the conversation. It pairs a compound annual rate of return with a drawdown figure pulled not from a single historical path but from thousands of simulated equity curves. The result is a score that penalizes systems with ugly risk profiles, no matter how impressive their top-line numbers look.

This metric sits at the center of a broader framework that includes safe-f for dynamic position sizing, Monte Carlo simulation for stress testing, and walk forward analysis for validation. In the Owl Group Trading method taught by Dr. Ken Long — a forty-year systematic trader and founder of Tortoise Capital Management — CAR25 is the headline scoring metric every backtested system must clear before it earns capital. The CAR25 framework originates with Dr. Howard Bandy; Dr. Long teaches it as the integration point between backtesting, position sizing, and R-based reporting in every cohort session.

Key Takeaways

CAR25 measures risk-normalized profit by dividing compound annual return by worst-case drawdown at the 95th percentile of Monte Carlo simulation.
Safe-f and dynamic position sizing work alongside CAR25 to adjust trade-by-trade risk based on current system performance, not static assumptions.
Backtesting, optimization, and walk forward analysis each serve a distinct role in system development, and skipping any one of them leads to unreliable results.

How CAR25 Fits Into Risk-Controlled System Design

CAR25 is not a standalone number you slap onto a backtest report. It functions as the objective function inside a layered evaluation process that connects your personal risk tolerance, your position sizing method, and your system's live performance into a single feedback loop. The pieces that matter most are safe-f for dynamic sizing, Monte Carlo simulation for stress testing, and a clear separation between the trading system itself and the trading management layer that governs how much capital you commit on each trade.

What CAR25 Measures And Why Traders Use It

CAR25 takes the compound annual rate of return of a trading system and divides it by the maximum drawdown at the 25th percentile of the drawdown distribution from a Monte Carlo analysis. That "25" in the name is not arbitrary. It targets the drawdown level that only 25% of simulated paths exceeded, giving you a return figure that has already been punished for risk.

You use it because it answers a question no raw return metric can: "If I traded this system a thousand times, how does the return look when I account for the realistic bad outcomes?" A system with a CAR25 of 2.0 earned twice its worst likely drawdown. A system with a CAR25 below 1.0 earned less than the pain it put you through. That distinction matters more than any equity curve screenshot.

How CAR25 Differs From Standard Return Metrics

Standard metrics like total return, CAGR, or even the Sharpe ratio treat drawdown as a secondary concern. CAGR tells you how fast the money grew. The Sharpe ratio adjusts for volatility, but volatility is not the same thing as the gut-wrenching equity drop that makes you abandon the system at the worst possible moment.

CAR25 puts drawdown at the center of the calculation. It does not care how smooth your monthly returns were. It cares about the deepest hole the account fell into across a realistic range of outcomes. Two systems with identical CAGR can have wildly different CAR25 scores because one survived the simulated stress tests cleanly while the other did not.

The Link Between Safe-f, Position Sizing, And Drawdown Tolerance

Safe-f is the fraction of your account that is "safe" to risk on the next trade, recalculated dynamically based on recent trade results. It is not a fixed percentage like the classic 1% or 2% rule. Instead, safe-f rises when the system is performing well and drops when the system enters a cold streak.

This matters because your drawdown tolerance is personal. You might accept a 15% peak-to-trough drawdown. Someone else might quit at 8%. Safe-f connects your tolerance to your position size in real time. When the system's recent trades suggest elevated risk, safe-f shrinks your exposure automatically. When conditions improve, it lets you press. CAR25 then measures whether that dynamic sizing process actually produced a return worth the risk you allowed.

Trading System Versus Trading Management

This distinction is one of the most overlooked ideas in system development. The trading system is the set of rules that generates buy and sell signals. The trading management layer decides how much capital to allocate to each signal.

You can have a mediocre signal generator paired with excellent trading management and outperform a brilliant signal generator with reckless sizing. CAR25 evaluates the combined result of both layers working together. If you change your position sizing method, your CAR25 changes even if you never touch a single entry or exit rule. That is the point. The metric forces you to treat risk management as a first-class component of performance, not an afterthought.

Why The 95th Percentile Matters In Monte Carlo Analysis

Monte Carlo simulation takes your historical trade list, shuffles the order thousands of times, and generates thousands of possible equity curves. Each curve has its own drawdown. The 95th percentile drawdown is the level that only 5% of those simulated paths exceeded.

Why not use the worst single path? Because the absolute worst case is an outlier driven by an unlikely sequence. Why not use the average? Because the average drawdown understates the real danger. The 95th percentile sits in the zone where you are saying, "I accept that 5% of the time things could be worse, but I am building my system evaluation around the realistic bad outcome, not the fantasy or the catastrophe." That is the drawdown figure CAR25 uses to normalize your return, and it is why the metric punishes systems that look great on average but carry hidden tail risk.

Building And Interpreting A Reliable Evaluation Process

A CAR25 score is only as trustworthy as the process that produced it. The chain runs from backtesting through optimization, validation, and finally dynamic sizing in live markets. Each link has a specific job, and each one can mislead you if applied carelessly. Dr. Howard Bandy's body of work on quantitative technical analysis provides the clearest framework for assembling these links into a process you can actually trust.

Using Backtesting Without Overtrusting Historical Results

Backtesting shows you what would have happened if you had traded a set of rules on past data. It does not show you what will happen. The gap between those two statements is where most traders lose money.

The main danger is curve fitting. If you test enough parameter combinations on the same data, you will always find one that looks spectacular. That result tells you more about the noise in the data than about any real edge. Treat backtesting as a filter, not a proof. It eliminates bad ideas quickly. It does not confirm good ones. Confirmation requires out-of-sample testing, and that is where the next step comes in.

Where Optimization And Walk Forward Analysis Belong

Optimization searches for the parameter values that performed best over a training window. Walk forward analysis then tests those parameters on data the optimizer never saw. You repeat this process across multiple windows, sliding forward through time.

The result is a series of out-of-sample performance segments stitched together. If the system holds up across all of them, you have evidence that the edge is real and not just a historical artifact. If it falls apart on the out-of-sample windows, the optimization found noise. Walk forward analysis is the single most important validation step in trading system development, and skipping it is the most common reason systems fail in live markets.

How Dr. Howard Bandy Frames Quantitative Technical Analysis

Dr. Howard Bandy treats trading system development as an engineering problem, not a guessing game. His framework starts with a discussion and quantification of four distinct layers of risk: your personal tolerance, the price fluctuations of the instrument, the risk added by your system rules, and the trade-by-trade risk experienced during live execution.

CAR25 emerged from this framework as what Bandy calls "as near a universal objective function as I have found." It integrates all four risk layers into a single evaluation score. His broader body of work, spanning multiple books and decades of research, emphasizes the non-stationary nature of financial data. Markets change. A system that worked in 2019 may not work in 2026. Every evaluation method you use must account for that reality, or the numbers it produces are decoration.

The Role Of Empirical Bayesian Analysis In Dynamic Sizing

Empirical Bayesian analysis is the statistical engine behind safe-f. Instead of assuming a fixed distribution of future trade outcomes, it updates its estimate of the distribution as new trade data arrives.

In practical terms, this means your position sizing reacts to what the system is actually doing right now, not what it did three years ago. If the last 20 trades show tighter wins and wider losses than the historical average, the Bayesian estimate shifts, safe-f drops, and your next trade is smaller. This is not a discretionary judgment call. It is a mathematical recalculation grounded in the most recent evidence. When combined with CAR25, it creates a feedback loop where the system's own performance governs how aggressively you trade it.

Common Misreads When Comparing CAR25 Across Strategies

Comparing CAR25 scores between two different systems is useful, but only if the inputs are comparable. Here are the mistakes that trip people up most often.

Different Monte Carlo settings. If one system used 5,000 iterations and the other used 500, the drawdown estimates are not equally reliable. Standardize your simulation parameters before comparing.
Different drawdown tolerances. Safe-f adjusts sizing based on your stated maximum drawdown. If you used 20% for one system and 10% for another, the CAR25 scores reflect different risk budgets, not different system quality.
Different time periods. A system tested across a bull market will produce a different CAR25 than the same system tested across a full cycle. Always compare systems tested on overlapping data windows.
Ignoring the trading management layer. Two systems with identical signals but different position sizing methods will produce different CAR25 scores. When you compare, make sure you know whether you are evaluating the signal, the sizing, or both.

Frequently Asked Questions

Is this platform legitimate and how does it compare to other used-car marketplaces?

CAR25 in the trading context is not a used-car platform. It is a risk-normalized performance metric developed by Dr. Howard Bandy for evaluating trading systems. If you arrived here looking for a vehicle marketplace, the term refers to something entirely different in quantitative finance. The "Car 25" mobile app for automotive sales professionals is a separate, unrelated product.

What buyer protections and escrow options are available for purchases and deposits?

This article covers CAR25 as a trading system evaluation metric, not a transactional marketplace. There are no buyer protections or escrow services associated with it. If you are evaluating trading systems or software that references CAR25, verify the vendor's reputation independently and never deposit funds without confirming the provider's track record.

How do I verify a vehicle's history, title status, and accident records before buying?

CAR25 in trading has no connection to vehicle history reports or title verification. The metric measures compound annual return divided by drawdown risk from Monte Carlo simulation. For vehicle history checks, consult dedicated automotive services. For trading system verification, demand walk forward analysis results and out-of-sample performance data before committing capital.

What fees, taxes, and shipping costs should I expect when buying a car in the USA online?

No fees, taxes, or shipping costs apply to CAR25 as a trading metric. It is a mathematical formula, not a commercial transaction. If you are purchasing trading education or software that uses CAR25 in its evaluation framework, expect standard digital product pricing. Always confirm what is included before purchasing.

Where can I find physical locations or partner dealerships to inspect vehicles in person?

CAR25 does not involve physical locations or dealerships. It is a quantitative tool used at the trading desk. Traders working with this metric typically operate from home offices or institutional trading floors. Firms like Owl Group Trading that teach systematic, risk-based evaluation methods work with traders remotely through coaching and mentoring programs.

What are the best ways to find reliable used cars under $10,000 or under $5,000 in the USA?

This question falls outside the scope of CAR25 as a trading evaluation metric. The term "CAR25" in finance refers exclusively to a risk-normalized return calculation. For used vehicle searches, dedicated automotive listing sites serve that purpose. For finding a reliable trading system, apply the CAR25 framework discussed in this article: run Monte Carlo simulations, validate with walk forward analysis, and let the risk-adjusted numbers guide your decision instead of raw returns.

About Owl Group Trading and Dr. Ken Long

This essay is part of the Owl Group Trading educational library. Dr. Ken Long — a forty-year systematic trader, founder of Tortoise Capital Management, retired U.S. Army Lieutenant Colonel, and developer of the Markets–Systems–Self framework, the Plan-Prepare-Execute-Assess (PPEA) discipline, the RLCO (Regression Line Crossover) chart lens, the Nine-Box Market Model for regime classification, and the 2R Battle Drill for managing winning trades — has refined these methods across more than 1,000 weekly cohort sessions since 2018. The CAR25 framework originates with Dr. Howard Bandy; Dr. Long teaches CAR25 as the canonical risk-normalized go/no-go gate every Owl backtest must clear.

Risk acknowledgment

Trading involves substantial risk of loss and is not suitable for every investor. The frameworks, formulas, and historical examples in this essay are educational. Backtested or live past performance does not guarantee future results. CAR25 captures realistic worst-case drawdown within a Monte Carlo distribution but cannot anticipate regime shifts that invalidate the underlying trade distribution. Before risking capital, validate any framework against your own data, your own broker fills, and your own response under live conditions.

Improve Your Craft Every Morning

Daily commentary from Dr. Ken Long — what he's seeing in markets, how he's framing trades, and what's worth practicing today. Free.