Based on Curtis Faith, Way of the Turtle: The Secret Methods that Turned Ordinary People into Legendary Traders (2007)
In 1983, legendary commodities trader Richard Dennis made a bet with his partner William Eckhardt about a deceptively simple question: Can trading be taught, or is it an innate talent?
Dennis believed that he could take a group of ordinary people — people with no trading experience — teach them his rules, give them money to trade, and they would become profitable traders. Eckhardt disagreed, arguing that there was something intrinsic in Dennis's abilities that could not be transferred through instruction alone.
To settle the bet, Dennis recruited two classes of "Turtles" (named after turtle farms he had visited in Singapore, where he remarked "We are going to grow traders just like they grow turtles"). The first class was recruited in late 1983 via a newspaper ad in the Wall Street Journal. The second class followed in 1984.
Over 1,000 people applied. Dennis and Eckhardt selected approximately 23 Turtles across both classes. The backgrounds were deliberately diverse: a card player, a game designer, an accountant, a security guard, a Dungeons & Dragons player, and others. The selection criteria focused on:
Dennis was right. Over the approximately four-and-a-half-year period of the program, the Turtles collectively earned over $175 million in profits. Curtis Faith, who was the youngest Turtle at age 19, was the most successful, personally earning over $31 million for Dennis.
The Turtle experiment proved that:
For decades, the Turtle rules were secret. When they were eventually published online (first partially by former Turtle Russell Sands, then more completely), Faith felt the context was missing. This book provides not just the rules but the rationale, the psychology, and the system design philosophy behind them. The rules alone are worthless without understanding why they work and what makes them hard to follow.
The Turtles were taught to think about trading not as a series of individual trades, but as a statistical process. Any single trade is essentially random — you cannot know in advance whether it will win or lose. What you can know is the expected value of a large number of trades.
This requires a fundamental shift in thinking:
| Novice Thinking | Turtle Thinking |
|---|---|
| "Will this trade work?" | "What is the expected value over 100 trades like this?" |
| "I've lost 5 in a row — the system is broken" | "5 losses in a row is within normal statistical variance" |
| "I should skip this signal; it looks weak" | "I must take every signal to capture the statistical edge" |
| "I'll take a partial profit to lock in gains" | "I'll let the system's exit rules determine when to close" |
| "This market feels wrong" | "My feelings are irrelevant to the system's edge" |
An edge is a statistical advantage that, over a large number of occurrences, will produce a net profit. It does not mean every trade wins — not even close. A casino's edge on roulette is only about 5.26%, yet it generates billions. The Turtles' edge was similar in concept: small per-trade, but relentlessly positive over hundreds of trades.
Expectation is the mathematical expression of edge:
Expectation = (Probability of Win × Average Win) - (Probability of Loss × Average Loss)
For the Turtles, a typical profile was:
This means: E = (0.38 × 4.5R) - (0.62 × 0.8R) = 1.71R - 0.50R = +1.21R per trade
A positive expectation of +1.21R means that for every dollar risked, the system returns $1.21 on average — an enormous edge.
The edge only manifests over a large sample of trades. In any short sequence — 10, 20, even 50 trades — randomness dominates. The Turtles were taught that:
This is why consistency and discipline matter more than the specific parameters of the system.
Faith devotes significant attention to why most people fail even when given a profitable system. The key biases:
| Bias | Description | How It Destroys Trading |
|---|---|---|
| Loss aversion | Losses hurt ~2.5x more than equivalent gains feel good | Traders cut winners short and let losers run — the exact opposite of what works |
| Recency bias | Overweighting recent events | After 3 losses, traders skip the next signal — which is often the big winner |
| Outcome bias | Judging a decision by its result, not its process | A profitable trade from a bad process reinforces bad habits |
| Anchoring | Fixating on a reference price | Refusing to sell at a loss because of the purchase price |
| Sunk cost fallacy | Holding losers because of what's already invested | Adding to losing positions to "average down" |
| Hindsight bias | Believing past events were predictable | Leads to overconfidence and under-appreciation of risk |
| Gambler's fallacy | Believing past losses make future wins more likely | Increasing size after losses, decreasing after wins |
Trend following is psychologically brutal because:
Dennis himself said the rules could be published in the newspaper and nobody would follow them. This turned out to be largely true — even among the Turtles, the performance variation was enormous, primarily because of varying levels of discipline rather than different rules.
The Turtles traded two complementary breakout systems simultaneously. Both are trend-following systems based on Donchian Channel breakouts (named after Richard Donchian, the pioneer of channel breakout trading).
Concept: Capture shorter-term trends using a 20-day breakout with a filter to avoid trading immediately after a winning breakout.
| Parameter | Long Entry | Short Entry |
|---|---|---|
| Entry trigger | Price exceeds the highest high of the last 20 days | Price falls below the lowest low of the last 20 days |
| Filter | Skip if the previous 20-day breakout in this market was a winner | Skip if the previous 20-day breakout in this market was a winner |
| Exit | Price falls below the lowest low of the last 10 days | Price exceeds the highest high of the last 10 days |
| Breakout period | 20 days | 20 days |
| Exit period | 10 days | 10 days |
The Filter Rule (Critical Detail):
The filter exists because breakout systems suffer from false breakouts in choppy markets. After a profitable breakout, the probability of the next breakout being a false signal (whipsaw) is higher. So System 1 skips the next breakout signal if the preceding breakout would have been profitable.
However, there is a critical safety mechanism: if a filtered (skipped) breakout turns out to have been a truly large move, the Turtles would enter on the 55-day breakout as a failsafe. This prevents the filter from causing them to miss a major trend.
Definition of "previous breakout was a winner": The preceding 20-day breakout in that market would have been profitable if traded — meaning the price moved at least 2N in the favorable direction before triggering the 10-day exit. This evaluation is theoretical (the breakout may or may not have actually been traded).
Concept: Capture longer-term trends using a 55-day breakout with no filter.
| Parameter | Long Entry | Short Entry |
|---|---|---|
| Entry trigger | Price exceeds the highest high of the last 55 days | Price falls below the lowest low of the last 55 days |
| Filter | None — take every signal | None — take every signal |
| Exit | Price falls below the lowest low of the last 20 days | Price exceeds the highest high of the last 20 days |
| Breakout period | 55 days | 55 days |
| Exit period | 20 days | 20 days |
System 2 is simpler and more robust. Because 55-day breakouts are rarer, they tend to represent stronger trends. The absence of a filter means the Turtles never miss a major move on this timeframe.
The two systems are complementary:
A breakout occurs when the current price exceeds (for longs) or falls below (for shorts) the channel boundary. The boundary is calculated from the closing prices of the lookback period. The signal is generated intraday — you do not wait for a close above the level. The moment price touches or exceeds the 20-day high (System 1) or 55-day high (System 2), the order is triggered.
In practice, the Turtles placed stop orders at the breakout levels at the start of each trading day, so execution was automatic.
Position sizing is the single most important component of the Turtle System. Curtis Faith is emphatic about this: the entry and exit rules are almost secondary compared to how much you trade. The Turtles used a volatility-based position sizing method built around a concept they called N.
N is a measure of a market's daily volatility, equivalent to the Average True Range (ATR) calculated as a 20-day exponential moving average.
True Range for a single day is defined as:
True Range = MAX of:
(a) Today's High - Today's Low
(b) |Today's High - Yesterday's Close|
(c) |Today's Low - Yesterday's Close|
The absolute values in (b) and (c) capture gap moves. If the market gaps up and then trades in a narrow range, the True Range reflects the gap.
N is the 20-day exponential moving average of True Range:
N_today = (19 × N_yesterday + TR_today) / 20
This is an EMA with a smoothing period of 20. It adapts to changing volatility — when a market becomes more volatile, N increases; when it quiets down, N decreases.
To compare volatility across markets with different contract sizes, the Turtles converted N into dollar terms:
Dollar Volatility = N × Dollars per Point
Where "Dollars per Point" is the dollar value of a one-point move in the futures contract. For example:
| Market | Contract | Dollars per Point | If N = | Dollar Volatility |
|---|---|---|---|---|
| Crude Oil | 1,000 barrels | $1,000 per $1 move | 1.20 | $1,200 |
| Gold | 100 oz | $100 per $1 move | 8.50 | $850 |
| S&P 500 | $250 per point | $250 per 1 pt move | 12.00 | $3,000 |
| Soybeans | 5,000 bushels | $50 per 1 cent move | 14.00 | $700 |
| T-Bonds | $100,000 face | $1,000 per 1 pt move | 0.70 | $700 |
| Japanese Yen | 12,500,000 yen | $12.50 per 0.01 move | 0.0032 | $400 |
The fundamental building block of Turtle position sizing is the Unit. One Unit represents the position size at which a 1N move in price equals 1% of the trading account.
Unit Size = 1% of Account Equity / Dollar Volatility
Or equivalently:
Unit Size = (Account Equity × 0.01) / (N × Dollars per Point)
Worked Example:
Account equity: $1,000,000
Market: Heating Oil
Dollars per Point: $42,000 per $1 move (42,000 gallons per contract)
N: 0.0141 ($1 = 1 cent move)
Dollar Volatility: 0.0141 × $42,000 = $592.20
Unit Size = ($1,000,000 × 0.01) / $592.20
= $10,000 / $592.20
= 16.88
Round down to 16 contracts per Unit.
Why this matters: By defining position size in terms of volatility, the Turtles ensured that every market contributed roughly the same amount of dollar risk per Unit. A 1N move in heating oil costs the same as a 1N move in gold or soybeans — approximately 1% of account equity. This is the mechanism that makes diversification work: no single market can dominate the portfolio.
The 1% figure was chosen so that:
To prevent excessive concentration, the Turtles operated under strict position limits:
| Limit Type | Maximum Units | Rationale |
|---|---|---|
| Single market | 4 Units | Prevents over-concentration in any one market |
| Closely correlated markets | 6 Units total | E.g., gold + silver together cannot exceed 6 Units |
| Loosely correlated markets | 10 Units total | E.g., all metals (gold, silver, copper) cannot exceed 10 Units |
| Single direction (all longs or all shorts) | 12 Units | Prevents excessive directional bias |
| Total portfolio | 24 Units | Absolute maximum across all markets and directions |
Correlation groupings used by the Turtles:
Closely correlated:
Loosely correlated:
The Unit calculation uses current account equity, which means:
Scenario: Account equity is $2,000,000. We are evaluating three markets on the same day.
Market: Gold (COMEX)
Price: $450.00/oz
20-day ATR (N): $6.50
Contract size: 100 oz
Dollars per Point: $100/point (1 point = $1)
Dollar Volatility: $6.50 × $100 = $650
1% of Account: $20,000
Unit: $20,000 / $650 = 30.77 → 30 contracts
Market: Crude Oil (NYMEX)
Price: $58.00/bbl
20-day ATR (N): $1.30
Contract size: 1,000 barrels
Dollars per Point: $1,000/point (1 point = $1)
Dollar Volatility: $1.30 × $1,000 = $1,300
1% of Account: $20,000
Unit: $20,000 / $1,300 = 15.38 → 15 contracts
Market: Japanese Yen (CME)
Price: 0.8900 ($/100 yen)
20-day ATR (N): 0.0074
Contract size: 12,500,000 yen
Dollars per Point: $12.50 per 0.0001 move
Dollar Volatility: 74 ticks × $12.50 = $925
1% of Account: $20,000
Unit: $20,000 / $925 = 21.62 → 21 contracts
Notice how the Unit sizes differ dramatically (30 gold vs 15 crude vs 21 yen) but the dollar risk per N-move is equalized at approximately $20,000 (1% of account) for each market.
A long entry is triggered when the current price exceeds the highest high of the previous N days (20 for System 1, 55 for System 2). A short entry is triggered when the current price falls below the lowest low of the previous N days.
In practice:
The filter rule for System 1 states: Do not take the current 20-day breakout if the previous 20-day breakout in this market would have been profitable.
Profitable is defined as: the breakout would have moved at least 2N in the favorable direction before being closed by the 10-day exit.
Implementation logic:
For each market, track the LAST 20-day breakout signal (regardless of whether it was traded):
1. Record the breakout price.
2. After the breakout, track the maximum favorable excursion (MFE).
3. Track when the 10-day exit would have been triggered.
4. If MFE ≥ 2N before the 10-day exit → classify as WINNER.
5. Otherwise → classify as LOSER.
Current signal decision:
If previous_breakout == WINNER → SKIP the current System 1 signal.
If previous_breakout == LOSER → TAKE the current System 1 signal.
The failsafe: If a System 1 signal is skipped due to the filter, but price continues to trend and eventually triggers a 55-day breakout (System 2 entry), the Turtles would enter on that signal. This prevents the filter from causing a total miss of a major trend.
Once an initial position of 1 Unit is established, the Turtles added additional Units as the price moved in their favor. The adding rule:
Add 1 Unit for every 0.5N move in the favorable direction from the previous entry price.
Maximum: 4 Units per market.
Example — Long Gold, N = $6.50:
| Unit | Entry Trigger | Price | Contracts | Cumulative Contracts |
|---|---|---|---|---|
| 1st Unit | 20-day breakout | $450.00 | 30 | 30 |
| 2nd Unit | Entry + 0.5N | $453.25 | 30 | 60 |
| 3rd Unit | Entry + 1.0N | $456.50 | 30 | 90 |
| 4th Unit | Entry + 1.5N | $459.75 | 30 | 120 |
The adding levels are based on the actual fill price of the previous Unit, not the theoretical price. If the 2nd Unit was filled at $453.50 (slippage of $0.25), the 3rd Unit trigger is $453.50 + $3.25 = $456.75.
The 0.5N spacing was chosen to balance two goals:
The Turtles placed their orders before the market opened each day. All entry orders were stop orders (buy stops for longs, sell stops for shorts) placed at the breakout levels. This ensured:
Every Turtle position had a hard stop-loss at 2N from the entry price. This is the maximum loss per Unit.
Long stop = Entry Price - 2N
Short stop = Entry Price + 2N
Since 1 Unit is sized so that a 1N move = 1% of equity, a 2N stop means:
When additional Units are added, the stops for all existing Units are raised (for longs) or lowered (for shorts) to maintain a 2N distance from the most recent entry.
Example — Long Gold, N = $6.50:
| Event | Stop Calculation | Stop Price |
|---|---|---|
| 1st Unit enters at $450.00 | $450.00 - $13.00 | $437.00 |
| 2nd Unit enters at $453.25 | $453.25 - $13.00 | $440.25 (all Units) |
| 3rd Unit enters at $456.50 | $456.50 - $13.00 | $443.50 (all Units) |
| 4th Unit enters at $459.75 | $459.75 - $13.00 | $446.75 (all Units) |
Risk at maximum position (4 Units):
| Unit | Entry | Stop | Risk per Contract ($) | Risk per Unit (30 contracts) |
|---|---|---|---|---|
| 1st | $450.00 | $446.75 | $3.25 | $9,750 |
| 2nd | $453.25 | $446.75 | $6.50 | $19,500 |
| 3rd | $456.50 | $446.75 | $9.75 | $29,250 |
| 4th | $459.75 | $446.75 | $13.00 | $39,000 |
| Total | $97,500 |
On a $2,000,000 account, this is 4.875% of equity — well within acceptable limits and significantly less than the theoretical maximum of 8% because the earlier Units have already moved in favor.
The 2N stop is wide enough to survive normal daily volatility but tight enough to be stopped out by genuine reversals. However, in choppy markets, the system will produce whipsaws — entering on a breakout, getting stopped out, and then watching the market resume the trend.
The Turtles' approach to whipsaws:
The Turtles had an absolute rule: stops are never moved further away from price. They can only be tightened (moved closer to price) as the position adds Units. This is non-negotiable. Moving stops wider is the single most destructive habit a trader can develop, because it converts a small, defined loss into an uncontrolled, potentially catastrophic one.
Longs: Exit when price touches or falls below the lowest low of the last 10 days.
Shorts: Exit when price touches or exceeds the highest high of the last 10 days.
This is a trailing exit — the 10-day low/high moves with price. In a strong uptrend, the 10-day low rises steadily, protecting profits. In a stalling trend, the 10-day low catches up to current price, triggering the exit.
Longs: Exit when price touches or falls below the lowest low of the last 20 days.
Shorts: Exit when price touches or exceeds the highest high of the last 20 days.
The longer exit period allows System 2 to ride trends longer, but also means it gives back more open profit before exiting.
Faith argues that exits are far harder psychologically than entries because:
Giving back profits. A 10-day trailing exit means that by the time you exit, the price has already dropped significantly from its peak. Watching $100,000 in open profit shrink to $60,000 before the exit triggers is emotionally devastating, even though $60,000 is an excellent outcome.
Early exit temptation. The urge to "lock in" profits by tightening the exit is overwhelming. But tightening exits destroys the system's edge — the large winners that compensate for the many small losers only materialize because the exit is wide enough to hold through normal retracements.
Late exit regret. After a large trend ends and the trailing exit closes the position, the market often bounces back slightly, making the trader feel they exited too late. This is an illusion — the exit price was unknowable in advance.
Faith notes that exits were the single biggest source of deviation among the Turtles. Those who manually overrode exits — taking profits early or exiting late — consistently underperformed those who followed the rules mechanically.
The Turtles traded a diversified portfolio of liquid futures markets across five sectors:
| Sector | Markets |
|---|---|
| Interest Rates | US T-Bonds (30-year), US T-Notes (10-year), Eurodollars (90-day) |
| Currencies | British Pound, Deutsche Mark, Swiss Franc, Japanese Yen, Canadian Dollar, French Franc |
| Commodities — Grains | Corn, Soybeans, Soybean Meal, Soybean Oil, Wheat |
| Commodities — Metals | Gold, Silver, Copper, Platinum |
| Commodities — Energy | Crude Oil, Heating Oil, Unleaded Gasoline |
| Commodities — Softs/Meats | Coffee, Cocoa, Sugar, Cotton, Live Cattle, Lean Hogs |
This gave them exposure to approximately 20-25 markets at any time.
Trend following depends on catching large trends, which are rare and unpredictable in any single market. Diversification ensures:
Faith argues that diversification is the only free lunch in trading. You can increase returns without proportionally increasing risk simply by adding more uncorrelated markets.
The Turtles only traded markets with sufficient liquidity, defined by:
As a rough guideline, a market needed enough volume that the Turtles' maximum position represented less than 1-2% of daily volume to avoid slippage problems.
The Turtles used three measures of equity:
| Type | Definition | Used For |
|---|---|---|
| Core Equity | Starting equity + or - all closed trade P&L | Base reference |
| Total Equity | Core Equity + open trade P&L (mark-to-market) | Reporting |
| Reduced Equity | Core Equity minus a drawdown-reduction factor | Position sizing during drawdowns |
For position sizing, the Turtles used Core Equity under normal conditions. This prevents open profits from inflating position sizes prematurely (a volatile open winner could temporarily inflate equity, leading to oversized positions that reverse painfully).
When the account suffered a drawdown from its equity peak, position sizes were reduced:
For every 10% drawdown from equity peak, reduce Unit size by 20%.
| Drawdown | Unit Size Reduction | Effective Unit Size |
|---|---|---|
| 0% to -10% | No reduction | 100% of normal |
| -10% to -20% | 20% reduction | 80% of normal |
| -20% to -30% | 40% reduction | 60% of normal |
| -30% to -40% | 60% reduction | 40% of normal |
| -40%+ | 80% reduction | 20% of normal |
Implementation:
Drawdown = (Peak Equity - Current Equity) / Peak Equity
Reduction Factor = FLOOR(Drawdown / 0.10) × 0.20
Adjusted Unit = Normal Unit × (1 - Reduction Factor)
Example: Account peaked at $2,000,000 and is now at $1,650,000.
The drawdown reduction rule serves two critical functions:
Survival. A 50% drawdown requires a 100% gain to recover. By reducing size during drawdowns, the system ensures the account can survive even severe adverse periods. Without this rule, a normal drawdown in trend following (20-40%) could become a terminal one.
Anti-martingale logic. The system inherently trades larger when winning and smaller when losing. The drawdown rule amplifies this effect during severe drawdowns, acting as an additional safety net.
Faith provides guidance on what to expect from a properly implemented Turtle-style system:
| Statistic | Typical Range |
|---|---|
| Maximum historical drawdown | 30-50% |
| Average annual drawdown | 15-25% |
| Longest drawdown duration | 6-18 months |
| Recovery time from max drawdown | 12-36 months |
| Frequency of 10%+ drawdowns | 1-3 per year |
| Frequency of 20%+ drawdowns | Once every 2-3 years |
These numbers are sobering. A 40% drawdown lasting 12 months is psychologically excruciating, even if it is statistically normal for the system. This is why mindset and discipline are so heavily emphasized — the system works, but only if you can survive the drawdowns without deviating from the rules.
Faith recounts that several Turtles failed not because the system stopped working, but because they could not tolerate the drawdowns:
The lesson is stark: the drawdown IS the system. You cannot have the 100%+ winning years without the 30-40% drawdown years. They are inseparable.
The Turtle edge is expressed as:
E = (PW × AW) - (PL × AL)
Where:
For the Turtle system:
| Component | Typical Value |
|---|---|
| PW (Win Rate) | 35-40% |
| PL (Loss Rate) | 60-65% |
| AW (Average Win) | 4-5R |
| AL (Average Loss) | 0.5-1.0R |
E = (0.38 × 4.5) - (0.62 × 0.8)
= 1.71 - 0.50
= +1.21R per trade
The key insight is the asymmetry between winners and losers:
Average Win / Average Loss = 4.5R / 0.8R = 5.6:1 reward-to-risk ratio
This means each winner is worth roughly 5.6 losers. Even losing 62% of the time, the winners more than compensate. The math:
For every 100 trades:
38 winners × 4.5R = 171.0R gained
62 losers × 0.8R = 49.6R lost
Net: = 121.4R profit
Per trade: = 1.21R
An R-multiple expresses a trade's result as a multiple of its initial risk. If a trade's initial risk is $2,000 (the 2N stop distance):
| Trade Result | Dollar P&L | R-Multiple |
|---|---|---|
| Stopped out at 2N | -$2,000 | -1.0R |
| Stopped out at 1N (tightened stop) | -$1,000 | -0.5R |
| Breakeven exit | $0 | 0R |
| Exited with 2:1 profit | +$4,000 | +2.0R |
| Rode major trend | +$20,000 | +10.0R |
| Monster trade | +$50,000 | +25.0R |
The Turtle system's edge comes from the fat right tail — the occasional +10R to +25R winners that make up for the many -0.5R to -1.0R losers.
Faith advocates using Monte Carlo simulation to understand the range of possible outcomes. By randomly resampling from actual trade results, you can generate thousands of hypothetical equity curves and assess:
For the Turtle system, Monte Carlo analysis typically shows:
Faith stresses that statistical edge only manifests over large samples. The practical implications:
Faith is deeply critical of curve-fitting — optimizing system parameters to fit historical data. The Turtle rules use round numbers (20-day, 55-day, 2N) precisely because they are not optimized. The system works not because 20 days is the perfect lookback period but because trend following works across a wide range of parameters.
The robustness test: A system is robust if it remains profitable across a wide range of parameter values. If changing the breakout from 20 days to 18 or 22 days causes a significant change in results, the system is likely curve-fit. The Turtle system is profitable with breakout periods ranging from approximately 15 to 30 days (System 1) and 40 to 70 days (System 2).
Curve-fitting works in backtest but fails in live trading because:
Faith recommends never evaluating a system solely on the data used to develop it. Proper testing requires:
A system that works in-sample but fails out-of-sample is curve-fit. The Turtle system was developed in the early 1980s and continued to work for decades — the ultimate out-of-sample test.
The complete Turtle system can be described in a few pages. There are no exotic indicators, no pattern recognition, no fundamental analysis. This simplicity is deliberate and essential:
Faith echoes the principle from Modern Portfolio Theory: diversification is the only way to improve risk-adjusted returns without requiring additional edge. The Turtles diversified across:
Trade every signal. No cherry-picking, no second-guessing, no "this one doesn't feel right." The system's edge depends on capturing the full distribution of outcomes, including the rare outlier winners that are impossible to predict in advance.
Never override the system. If the system says buy, you buy. If the system says hold, you hold. If the system says exit, you exit. Your feelings are noise; the system is signal.
Be consistent across time. The system will have terrible months and even terrible quarters. Consistency means trading the same way in month 18 of a drawdown as in month 1 of a winning streak.
Accept losses as the cost of doing business. Every loss is a small premium paid for the option of catching a large trend. You cannot get the large trends without paying these premiums.
Never risk more than the system dictates. The position sizing rules are there to ensure survival. Exceeding them for any reason — even a "high conviction" trade — violates the statistical foundation of the system.
Dennis's famous observation — "I could publish the rules in the newspaper and no one would follow them" — proved remarkably prescient. The rules were eventually published. They are simple. Yet very few people trade them successfully because:
A system with a 38% win rate will, with reasonable probability, produce:
During these streaks, every cognitive bias screams at the trader to stop, to change something, to abandon the system. The Turtles who performed worst were the ones who listened to these impulses.
Changing the breakout from 20 to 19 days because the backtest shows a 3% improvement is almost certainly capturing noise. The correct response to parameter sensitivity analysis is: if the system works across a broad range, pick a round number and leave it alone.
This is the most destructive mistake. After 4-5 losses in a row, the trader decides "this market isn't working" or "I'll wait for conditions to improve." The very next signal is often the one that turns into a 10R winner. By skipping it, the trader has destroyed the mathematical edge.
Reducing position size after losses (beyond what the drawdown rules dictate) means that when the winning trades finally arrive, they are too small to recover the accumulated losses. The Turtle drawdown rule handles size reduction systematically — ad-hoc reductions do more harm than good.
After a losing period, the temptation to add filters, indicators, or conditions to "improve" the system is overwhelming. Common additions that hurt performance:
Each addition reduces sample size, increases complexity, and provides more opportunity for curve-fitting.
Trend following systems have extended periods of poor performance — typically during low-volatility, range-bound markets. These periods are not evidence that the system is broken. They are statistically expected. The system "breaks" only if the fundamental nature of markets changes (i.e., markets stop trending) — which has not happened in recorded history.
Taking profits before the exit signal fires is one of the most common deviations. It feels good to "lock in a winner," but it systematically removes the large outlier wins that drive the system's profitability. A trader who takes profits at +2R instead of letting the exit run to +8R has destroyed 75% of that trade's value.
System: System 2 (55-day breakout)
Market: Crude Oil (NYMEX)
Account equity: $1,000,000
Date: March 15
N (20-day ATR): $1.50
Contract size: 1,000 barrels
Dollars per Point: $1,000 per $1 move
55-day high: $62.00
55-day low: $53.00
Current price: $61.80
Dollar Volatility = $1.50 × $1,000 = $1,500
1% of Account = $1,000,000 × 0.01 = $10,000
Unit Size = $10,000 / $1,500 = 6.67 → 6 contracts per Unit
March 17: Crude oil breaks above $62.00, triggering the 55-day breakout.
Entry: $62.10 (filled with minor slippage)
Size: 6 contracts (1 Unit)
Stop: $62.10 - (2 × $1.50) = $59.10
Risk: ($62.10 - $59.10) × 6 × $1,000 = $18,000 (1.8% of account)
The add trigger is 0.5N above the first entry: $62.10 + $0.75 = $62.85.
March 20: Price reaches $62.85.
Entry: $62.90 (filled)
Size: 6 additional contracts (Unit 2)
Stop: $62.90 - $3.00 = $59.90 (all 12 contracts now have this stop)
Updated risk profile:
| Unit | Entry | Stop | Risk/Contract | Risk/Unit |
|---|---|---|---|---|
| 1 | $62.10 | $59.90 | $2.20 | $13,200 |
| 2 | $62.90 | $59.90 | $3.00 | $18,000 |
| Total | $31,200 (3.12%) |
Trigger: $62.90 + $0.75 = $63.65.
March 24: Price reaches $63.65.
Entry: $63.70 (filled)
Size: 6 additional contracts (Unit 3)
Stop: $63.70 - $3.00 = $60.70 (all 18 contracts)
Updated risk:
| Unit | Entry | Stop | Risk/Contract | Risk/Unit |
|---|---|---|---|---|
| 1 | $62.10 | $60.70 | $1.40 | $8,400 |
| 2 | $62.90 | $60.70 | $2.20 | $13,200 |
| 3 | $63.70 | $60.70 | $3.00 | $18,000 |
| Total | $39,600 (3.96%) |
Note: Unit 1 is now only risking $1.40 from its entry — less than 1N. The trailing stop is protecting early profits.
Trigger: $63.70 + $0.75 = $64.45.
March 28: Price reaches $64.45.
Entry: $64.50 (filled)
Size: 6 additional contracts (Unit 4 — maximum)
Stop: $64.50 - $3.00 = $61.50 (all 24 contracts)
Updated risk:
| Unit | Entry | Stop | Risk/Contract | Risk/Unit |
|---|---|---|---|---|
| 1 | $62.10 | $61.50 | $0.60 | $3,600 |
| 2 | $62.90 | $61.50 | $1.40 | $8,400 |
| 3 | $63.70 | $61.50 | $2.20 | $13,200 |
| 4 | $64.50 | $61.50 | $3.00 | $18,000 |
| Total | $43,200 (4.32%) |
Position is now at maximum size: 24 contracts across 4 Units.
Over the next several weeks, crude oil continues higher:
April 10: Price $68.00 | 20-day low = $63.50 | Stop: $61.50
April 20: Price $71.00 | 20-day low = $65.00 | Stop: $61.50
May 1: Price $74.50 | 20-day low = $68.50 | Stop: $61.50
May 10: Price $73.00 | 20-day low = $69.00 | Stop: $61.50
Note: The 2N stop ($61.50) has not been tightened further because the exit is governed by the 20-day low rule for System 2. The stop only serves as an emergency backstop at this point — the 20-day exit will trigger first.
May 15: Crude oil pulls back. The 20-day low is $69.00. Price touches $69.00.
Exit price: $69.00
All 24 contracts exited simultaneously.
| Unit | Entry | Exit | Profit/Contract | Contracts | Profit |
|---|---|---|---|---|---|
| 1 | $62.10 | $69.00 | $6.90 | 6 | $41,400 |
| 2 | $62.90 | $69.00 | $6.10 | 6 | $36,600 |
| 3 | $63.70 | $69.00 | $5.30 | 6 | $31,800 |
| 4 | $64.50 | $69.00 | $4.50 | 6 | $27,000 |
| Total | 24 | $136,800 |
Profit as % of account: 13.68%
R-multiple (Unit 1): $41,400 / $18,000 initial risk = 2.30R
R-multiple (total): $136,800 / $43,200 max risk = 3.17R
This is a moderate winner — not the kind of outlier that makes a Turtle's year, but a solid profitable trade that demonstrates the system working as designed. The really large Turtle profits came from moves where crude oil (or other markets) trended for months, producing +10R to +25R outcomes.
| Characteristic | System 1 | System 2 |
|---|---|---|
| Breakout period | 20 days | 55 days |
| Exit period | 10 days | 20 days |
| Filter | Yes — skip after winning breakout | No — take every signal |
| 55-day failsafe | Yes — enter on 55-day if System 1 filtered | Not applicable |
| Signal frequency | Higher (more trades per year) | Lower (fewer trades per year) |
| Win rate | Lower (~32-38%) | Higher (~38-42%) |
| Average win size | Smaller (shorter holding period) | Larger (longer holding period) |
| Average loss size | Smaller (tighter exit) | Larger (wider exit) |
| Holding period | Days to weeks | Weeks to months |
| Whipsaw frequency | Higher (shorter breakout, more false signals) | Lower (longer breakout, more significant moves) |
| Performance in choppy markets | Poor — many false breakouts | Less poor — fewer signals generated |
| Performance in trending markets | Good — catches trends early | Good — catches larger portion of major trends |
| Psychological difficulty | Higher (more frequent losses) | Lower (fewer trades, less noise) |
| Best suited for | Markets that trend frequently in shorter bursts | Markets that produce sustained multi-week/month trends |
| Capital requirement | Lower per-market (tighter stops) | Higher per-market (wider stops) |
| Gives back less open profit | Yes — exits faster (10-day) | No — exits slower (20-day) |
| Gives back more open profit | No | Yes — wider trailing exit |
| Expectation per trade | Slightly lower | Slightly higher |
| Annual return contribution | More trades × smaller avg win | Fewer trades × larger avg win |
Combined advantage: Running both systems simultaneously provides diversification across timeframes. When System 1 is whipsawing in a choppy market, it generates small losses but System 2 may not trigger at all (avoiding those same losses). When a major trend develops, System 2 catches the big move while System 1 may have been filtered out or exited early.
"The secret of trading success was never about finding a 'secret' system. It was about learning to trade a good system well."
"I could publish my trading rules in the newspaper and no one would follow them. The key is consistency and discipline." — Richard Dennis
"Good trading is not about being right, it is about trading right. If you want to be successful, you need to think of trading as a statistical game."
"Trade with an edge, manage risk, be consistent, and keep it simple. That is the whole secret of the Turtle way."
"The Turtles who did the best were the ones who followed the rules most consistently. The Turtles who did the worst were the ones who tried to be clever."
"A losing period does not mean that a system is broken; it means that market conditions are temporarily unfavorable. The system will recover when conditions change — but only if you are still trading it."
"Most traders would rather feel good than make money. They want to be right on every trade. The Turtles learned to accept being wrong on individual trades in exchange for being right over the long run."
"The markets I traded in 1984 were not all that different from those available today. The principles of trend following are timeless because they are based on human nature, which does not change."
"Exits are much more important than entries. You can enter randomly and still make money with a good exit strategy. You cannot make money with a good entry and a bad exit."
"Position sizing is the most important part of any trading system. It determines the risk of the portfolio and the long-term compounding of returns. Yet most traders spend 90% of their time on entries and 0% on position sizing."
"The emotional difficulty of following a system is the main reason it works. If it were easy, everyone would do it, and the edge would disappear."
"Don't worry about individual trades. Worry about whether you are following your rules. The results will take care of themselves over time."
"The best traders I know are the simplest. They have a few straightforward rules, they follow them, and they make a fortune. The worst traders I know are the most complex — always looking for the perfect indicator, the perfect timing, the perfect entry."
End of implementation specification.