HARMONIC
The Academy
Derivatives Mastery
Core Curriculum
MODULE
08
Subscribe
Series Derivatives Mastery
Module 8 of 8
Topic Risk Management
Harmonic Academy
← Module 7: Technical Analysis
Core Curriculum Module 8 of 8 · Derivatives Mastery

Risk Management

How to size positions, manage drawdowns, and stay in the game.
Understand Apply Case Study

This is the final module, and arguably the most important. Every strategy covered in Modules 1 through 7, futures, options, carry, spreads, and technical analysis, only produces results if you are still in the market long enough to let those strategies work. Risk management is not a defensive add-on to trading. It is the primary discipline.

01
Understand

The mathematics are unforgiving.

Losses and the gains needed to recover from them are not symmetric. The deeper the drawdown, the more disproportionate the recovery required:

−10%
Requires +11% to recover
−25%
Requires +33% to recover
−50%
Requires +100% to recover

Avoiding large losses is more valuable than capturing large gains. Most retail traders who blow up accounts do not do so because their market analysis was wrong. They do so because they sized positions incorrectly relative to their risk tolerance and had no pre-defined rule for when to exit.

02
Understand, continued

The Kelly criterion, and knowing if your edge is real.

In 1956, physicist John Larry Kelly Jr. was working at Bell Labs alongside Claude Shannon, the mathematician who invented information theory. Using Shannon’s framework, Kelly published a paper answering a question gamblers had asked for centuries: given an edge, how much should you bet? Kelly died in 1965 at 41 and never used his own formula to make money, but it traveled from Bell Labs to Las Vegas to Wall Street and became the backbone of position sizing for some of the most successful traders in history.

Shannon introduced the formula to Ed Thorp, who used it to beat casinos at blackjack, then applied the same framework running Princeton-Newport Partners. Warren Buffett and Charlie Munger have cited proportional sizing as central to their capital allocation. Jim Simons employed Elwyn Berlekamp, who had worked with Kelly at Bell Labs, as a direct bridge to Renaissance Technologies and the Medallion Fund. The formula works because it is mathematically optimal for long-term capital growth given a genuine edge, and it protects against ruin, a property that makes it essential for leveraged derivatives trading.

The Kelly formula

f* = (p × (b + 1) − 1) ÷ b

p = probability of winning. b = net odds received (profit per unit risked).

Example: 60% win rate, 2:1 payoff → f* = (0.60 × 3 − 1) ÷ 2 = 0.40 → bet 40% of capital

In a coin flip where heads pays $2 and tails loses $1, win probability 50 percent, net odds 2:1, Kelly says bet 25 percent of capital per flip. At that fraction, long-term capital growth is maximized. Bet more and you grow faster short-term but risk ruin, bet less and you grow slower than you could. In trading the variables are less clean, you do not have a fixed probability or payoff on every trade, and overestimating your edge leads to overbetting, the most dangerous error. A Kelly fraction of 40 percent means risking 40 percent of capital on one position, psychologically and practically extreme for most traders.

Kelly’s classic form assumes a binary win or lose outcome with a fixed payoff, true for a simple directional futures bet but only an approximation for the spreads and multi-leg options strategies from Modules 5 and 6. A bull call spread or iron condor has a range of possible outcomes, not two, and its actual payoff distribution should be estimated from the structure’s own maximum gain, maximum loss, and probability of landing in each zone rather than plugged into the formula unmodified. Treat Kelly as a sizing discipline and starting discount, not a precise output, for anything more complex than a single directional bet.

Kelly answers how much to bet given an edge. Expectancy answers whether the edge is real in the first place, and it is the simpler tool most traders should run first. Measure every trade in R, multiples of your initial risk: a trade that risked $500 and made $1,000 is a +2R win, one that lost $500 is a −1R loss. Expectancy is the average R across a sample of trades:

Expectancy

E = (win rate × average win in R) − (loss rate × average loss in R)

Example: 40% win rate, average win +2.5R, average loss −1R → E = (0.40 × 2.5) − (0.60 × 1) = 1.0 − 0.6 = +0.4R per trade

A positive E means the system makes money over enough trades even with a losing win rate, a 40 percent win rate with 2.5:1 payoffs beats a 60 percent win rate with 1:1 payoffs. This is why professional standard is a minimum 2:1 reward-to-risk on entries, at 2:1 you only need to win 34 percent of trades to break even before edge; at 3:1, only 25 percent. Track expectancy in your journal alongside Kelly, if E turns negative or flat, no amount of position sizing discipline fixes it, the setup itself has stopped working.

Fractional Kelly: the practical application
Full Kelly
Mathematically optimal growth rate. Volatile path, maximum drawdowns can be severe. Requires extremely accurate probability and payoff estimates. Used by very few practitioners.
Half Kelly
Half the optimal fraction. Grows at roughly 75% of the full Kelly rate, drawdowns roughly halved. The most common professional choice, Ed Thorp’s preferred approach.
Quarter Kelly
Conservative. Minimal drawdowns. Appropriate when probability estimates are uncertain, which is most of the time in trading.
Fixed 1–2%
Risk a fixed 1 to 2 percent of capital per trade regardless of estimated edge. Simplest approach, no estimation error. The most practical starting point for retail derivatives traders.
03
Understand, continued

Liquidation mechanics and avoidance.

Liquidation is the forced close of a leveraged position when margin falls to the maintenance threshold, the most destructive event in derivatives trading because it removes both capital and the ability to participate in any subsequent recovery. The liquidation price is calculable before you enter, know it before you open the position, and your stop-loss must sit comfortably above it, not adjacent to it.

Liquidation price, approximate

Long ≈ Entry − (Margin × (1 − Maintenance rate)) ÷ Position size

Short ≈ Entry + (Margin × (1 − Maintenance rate)) ÷ Position size

Example: long BTC at $60,000 with $20,000 margin (3x leverage), 0.5% maintenance rate. Liquidation ≈ $60,000 − ($20,000 × 0.995) ÷ 0.333 BTC ≈ $40,200, a 33% move against you. At 10x leverage, that liquidates near $54,000, a 10% adverse move.

Leverage vs. liquidation buffer · the buffer shrinks non-linearly
Liquidation buffer as a percentage of entry price, by leverage Approximate liquidation buffer shrinks sharply as leverage rises: roughly 33 percent at 3x, 18 percent at 5x, 10 percent at 10x, and under 5 percent at 20x. 0% 35% 33% 3x 18% 5x 10% 10x <5% 20x

The drop from 3x to 20x is not linear. Most of the buffer disappears between 5x and 10x, where a single volatile crypto candle can already reach the liquidation threshold.

Avoidance tactics
Use low leverage
Stay at 3 to 5x maximum. At 5x your liquidation buffer is roughly 18 percent, at 20x it is under 5 percent. A single volatile candle in crypto can cover 5 percent.
Isolated margin
Always use isolated margin on individual positions. Cross margin pools your entire account balance, one bad position can drain everything else.
Top up margin early
If a position moves against you and your thesis is still intact, add margin before you approach the liquidation threshold. Defending a position is cheaper than re-entering after liquidation.
Monitor funding
High positive funding combined with choppy sideways price action is the highest-risk environment for leveraged longs. Reduce size or close.
Watch liq. clusters
Large liquidation clusters above or below price act as magnets. Know where they are before entering, avoid stops placed directly at cluster levels.
04
Understand, continued

Stop-loss discipline and the psychological risks.

A stop-loss is not a prediction that price will not go lower. It is a pre-committed decision about when you are wrong. The best stops sit at levels that invalidate the original thesis, from Module 7: below the most recent higher low for trend following, one to two ATR below the breakout level for breakout, beyond the Bollinger Band extreme for mean reversion. Stops feel like losses. They are not, a stop-loss is the cost of being wrong in a business where being wrong is inevitable.

Stop types
Fixed stop
Set at a specific price level determined before entry. Does not move. Placed at the level that invalidates the technical thesis.
Trailing stop
Moves with price in the direction of the trade, locking in profit as the position moves in your favor. Stays fixed when price moves against you. Best for trend following.
Conditional stop
Triggers on a condition beyond price: a time stop (close if thesis not confirmed within X hours), a volatility stop (close if ATR expands beyond threshold), or a funding stop (close if funding turns negative on a long carry position).

Every psychological failure in trading shares the same root cause: emotion overrides the pre-committed rules set before the trade. FOMO entries happen when price is moving fast and you fear missing the move, entering without a setup, always at the worst price. Revenge trading increases size after a loss to recover quickly, turning a manageable loss into an account-threatening one. Moving stops extends the loss threshold because you believe the position will recover, converting a defined loss into an undefined one. Averaging down adds to a losing position without new confirming information, the first loss is the best loss. Overbetting after wins mistakes variance for skill and inflates size precisely when mean reversion is most likely.

05
Understand, continued

Your risk framework.

Max risk per trade
No more than 1 to 2 percent of total capital on any single position. At 1 percent, fifty consecutive losses reduce capital to 60 percent of starting value, survivable.
Max open exposure
Total risk across all open positions should not exceed 5 to 10 percent of capital. Correlated positions, multiple crypto longs, count as one exposure.
Daily loss limit
If you lose more than 3 to 5 percent of capital in a single day, stop trading, come back tomorrow. Forced breaks prevent the revenge-trading spiral.
Tiered drawdown response
0 to 5% drawdown: normal variance, no change. 5 to 15%: cut position size in half and trade only your highest-conviction setups. Beyond 15 to 20%: stop trading entirely and review the system before resuming at reduced size.
Pre-trade checklist
Before every trade: write down the entry, stop-loss, target, and maximum dollar loss. If any of the four is unclear, do not enter.
Trade journal
Record every trade with the setup, entry, exit, and outcome. Review weekly. Identify which setups are working and which are not, the journal is the only reliable feedback mechanism in trading.

A more institutional way to frame all of the above: think of your total capital as a risk budget, not a pile of dollars. Every open position consumes a slice of that budget in proportion to its estimated volatility, not its dollar notional, a small position in a high-volatility altcoin perpetual can consume more of your risk budget than a larger position in a lower-volatility spot holding. Sizing to a target book volatility, or a value-at-risk figure you are comfortable losing over a given window, keeps the 1 to 2 percent per-trade rule and the 5 to 10 percent open-exposure rule from being defeated by unequal volatility across your positions, two trades sized the same in dollars are not the same size in risk.

The math of Kelly is simple. The hard part is honest estimation of your edge. Most traders overestimate their win rate and their payoff. Start with half Kelly or fixed 1 to 2 percent. As your track record develops and your estimates become more reliable, you can size up. Never the other way around.

06
Apply

Full risk framework checklist.

1

Check expectancy first. From your trade journal, calculate average R won versus average R lost across this setup. If expectancy is zero or negative, stop, no sizing method fixes a setup with no edge.

2

Calculate your Kelly fraction (win rate, average payoff to average loss ratio), apply f* = (p × (b + 1) − 1) ÷ b, then halve the result. Without at least 50 journal entries, use fixed 1 percent sizing instead.

3

Convert to dollar terms. Multiply the half-Kelly fraction by your total trading capital, that is your maximum position size. If it feels uncomfortably large, use quarter Kelly instead.

4

Calculate your liquidation price using the exchange calculator. Your stop-loss must sit above it with meaningful buffer, at minimum 10 to 15 percent of entry price.

5

Apply the daily loss limit and tiered drawdown rule. Before entering, confirm you have not already exceeded them. If you have, close the screen and come back tomorrow.

6

Write it down: entry price, stop-loss level, target, maximum dollar loss. These four numbers exist before you click confirm, no exceptions.

7

After the trade closes, record it in your journal: win or loss, the setup, what you saw, and what happened. The journal is how you get better.

07
Case study

Ed Thorp: from blackjack tables to 20% annualized returns.

Ed Thorp: Princeton-Newport Partners ~20% annualized, no losing years

Ed Thorp is the most instructive case study for Kelly in practice, his career spans casino blackjack to running one of the most successful hedge funds in history, built on the same framework of edge identification, probability estimation, and Kelly-based position sizing. He developed a card-counting system that gave him a statistical edge over the house, then applied Kelly to determine the optimal fraction of his bankroll to bet given that edge. Casinos eventually banned him, which Thorp took as the clearest possible confirmation that his edge was real.

He applied the same framework to financial markets at Princeton-Newport Partners: identify a genuine edge, estimate probability and payoff with honesty, size positions using fractional Kelly to maximize growth while controlling drawdowns. The fund ran for nearly two decades and produced approximately 20 percent annualized returns net of fees with no losing year. The most important lesson is not the specific strategies he used, it is the discipline of the framework. He never bet more than his edge justified, used half Kelly as his default, kept meticulous records, and exited when the thesis was invalidated, not when the loss was painful enough. The strategies changed as markets evolved. The framework never did.

Thorp’s rule, paraphrased: never risk ruin. A strategy that can end in ruin, no matter how positive its expected value, is not a good strategy. Kelly sizing is the mathematical implementation of that rule.

08
Summary

Risk management is the discipline that makes everything else work.

Key takeaway

Expectancy tells you whether your edge is real. Kelly, and fractional Kelly in practice, tells you how much to bet given that edge. Thinking in terms of a risk budget rather than dollar notional keeps position sizes honest across positions of unequal volatility. Liquidation avoidance, stop discipline, tiered drawdown response, and a trade journal give you the structure to execute all of it consistently. Modules 1 through 7 tell you what to trade and when. This module tells you how much. Get the sizing right and the strategies work. Get it wrong and no strategy saves you.

Risk warning: no position sizing system eliminates the possibility of loss. Markets are unpredictable and edge estimation is inherently uncertain. Kelly sizing assumes your probability and payoff estimates are correct, errors in those estimates lead to overbetting. Always use fractional Kelly and never risk capital you cannot afford to lose.

This content is produced by Harmonic for educational purposes. It is strategy education, not investment advice.

Get the full Harmonic week in your inbox.

Both issues, every week, delivered as clean PDFs you can read anywhere:

Opener, the week ahead
Sun night
Closer, the week in review
Thu night
No spam · unsubscribe anytime.

Plain English by Harmonic · harmonicsolutions.io