Surviving drawdown
Learning objectives
- Define drawdown as peak-to-trough in R and compute the gain a percentage loss needs to recover.
- Write three loss limits as house rules (a daily stop, a weekly brake, a max-drawdown line) and say what each one does when it fires.
- Size those limits from your own method's drawdown distribution instead of copying someone else's numbers.
- Grade the lesson's claims: recovery maths (definition), loss limits (house rule), taking more risk after losses (measured in other markets, plausible for you).
It is Thursday, 11:40 ET (23:40 ICT, winter). Your method has made money for a year, but you are down 9R from your best day, three weeks running, and you have just taken a second loss this session. Two questions arrive together: is the method broken, and should you take one more trade to win some of it back? This lesson answers both before you are in that seat. Every method with a real edge spends long stretches below its last high. The job is to decide in advance what you do at each depth, so a normal drawdown stays normal and a broken method gets caught.
Key term
Drawdown (sụt giảm vốn) is the fall from the highest point your equity has reached to the lowest point after it, measured before a new high is made Definition. Measure it in R, the amount you risk per trade. “I’m down $800” is a feeling. “I’m 8R below my peak on a method that averages +0.35R a trade” is information: it tells you whether this is ordinary variance or something M7.07’s process review should look at.
One method, one path
A winning method still spends weeks underwater
ModelHow it works
The recovery maths
A loss of a fraction d of the account needs a gain of d ÷ (1 − d) to get back Definition. The two numbers drift apart quickly:
| Drawdown | Gain needed to recover |
|---|---|
| 10% | 11.1% |
| 20% | 25% |
| 30% | 42.9% |
| 50% | 100% |
In R the picture depends on how you size. If you risk a fixed dollar amount, an 8R hole takes 8R to fill. If you risk a fixed percentage of the account (M6.09), each R gets smaller as the account shrinks, so the same hole takes more than 8R. Either way, preventing a deep drawdown is cheaper than climbing out of one. The full treatment, with the distribution behind it, lives in RSK1.03.
Three limits, written before you need them
These three limits and their numbers are our house rules, not measured thresholds House rule. Each works on a different clock.
- Daily stop: −2R. At −2R on the day you go flat and stop for the session (M7.02). Its job is to cap how bad an ordinary day can get. It cannot make a bad day impossible: a fast market can fill your stop well past its price.
- Weekly brake: −6R. A −6R week triggers half size the next week until you have two green weeks, plus a review brought forward. It is a brake, not a verdict: it slows you down while you check whether the method, the market or your execution changed.
- Max-drawdown line: −15R from peak. Written before it ever happens. At the line you stop trading the method live and return to sim and the validation steps in M7.10. Reaching it is the plan doing its job, and trading past it is the failure.
Where the numbers should come from
The numbers above are a starting template. The right line for you comes from your own method’s drawdown distribution. Take your journal’s R results (or the win rate and payoff you have measured), shuffle them into many possible orders, and record the deepest drawdown each order produces. That gives you a range of drawdowns an unchanged method produces by chance alone Definition. Then put the max line beyond most of that range, so hitting it says “this is unlikely to be ordinary variance”.
For the method in the figure (45% at +2R/−1R, independent trades, no costs), our simulation for this lesson gives Model:
- over 200 trades, a median max drawdown of 9R, 14R at the 90th percentile and 16R at the 95th; about 10% of paths touch −15R;
- over 500 trades, a median of 12R, 17R at the 90th and 19R at the 95th; about 25% of paths touch −15R;
- with 10 trades a week, about 2% of weeks end at −6R or worse, so across a 50-week year the brake fires at least once on most intact paths.
So for this method a −15R line halts an intact method in roughly one year out of four if it trades 500 times a year. That may be a price you accept, because the line also catches a method that has stopped working. Or you may widen it, or reset it after each new equity high. Either choice is a house rule. What matters is that you choose it from your own numbers before the drawdown starts. Costs, a lower edge or trades that cluster on bad days all make the drawdowns deeper than this model.
Drawdown changes judgement
People and traders behave differently after losses. In experiments, people who are behind take bets they would refuse otherwise when the bet offers a chance to get back to even MeasuredOther market (Thaler and Johnson, 1990). Proprietary futures traders at the Chicago Board of Trade who lost money in the morning took more risk in the afternoon than those who had gained MeasuredOther market (Coval and Shumway, 2005; Treasury bond futures, 1998). That the same pull acts on you on NQ is plausible but not measured PlausibleOther market.
In a drawdown you will feel three urges: change methods (M7.07 says judge the process, not a short run of results), size up to “make it back” (M6.09 sets size by formula; M6.10 limits re-entries) and skip valid setups (M7.06). Write all three on your plan now, each followed by “I will not.”
What the data says
The limits do not make the method better. They decide how much of the account a bad stretch can take before you stop and look, and they take that decision away from the version of you who is in the middle of it.
Worked example
A $6,000 account trades a method that averages +0.35R a trade before costs. Risk is $60 a trade (1%): 2 MNQ ($2 a point each, $4 a point together) with a 15-point stop. Synthetic numbers, not a real account.
With the limits. Over three weeks the account falls 8R from its peak (−$480, 8% of the account). That is ordinary for this method: our simulation puts the median 200-trade drawdown at 9R. The daily stop holds on every day (worst day −1.8R), no single week reaches −6R, and the max line is far away. The trader keeps executing. Week four ends +5R and week five +4R, and the account makes a new high. At a fixed $60 risk the 8R hole took 8R to fill; at 1% of a $5,520 account the risk would be $55.20, so the same $480 would have needed about 8.7R.
Without them. Same market, same signals. There is no daily stop, so a revenge session ends at −5R. The trader then doubles size to “get it back”, so each loss now costs 2R in the original units. Ten days later the account is down 14R (−$840, 14%) and needs +16.3% to recover ($840 ÷ $5,160), traded by someone who is now afraid to take valid setups. In this constructed example the variance was the same; the difference came from the rules.
The failure case
The limits also fail in the other direction. On the 500-trade horizon about one intact path in four reaches −15R. A trader who halts there stops trading a method that still works, loses the weeks spent in sim and may give up on a good method entirely. That is why the line should come from your own distribution, and why reaching it sends you to review and sim, not to a new method.
Prop-firm note
If you trade a prop-firm evaluation or funded account, the firm sets its own daily and maximum loss limits. Those are firm rules with their own dates and definitions (PROP track), not the same thing as your house limits. Set your house limits tighter than the firm’s, so your rule fires before theirs.
Common mistake: Treating the daily stop as a guideline that resets when a great setup appears at −1.9R. The stop works because it is absolute. A −2R limit you override “just this once” is a suggestion you have already shown you will ignore, which means your real daily stop is wherever tilt takes you.
Common mistake: Copying someone’s −15R line and reading a hit as proof the method is broken. The same method reaches −15R by chance at a rate that depends on its win rate, payoff and how many trades you take. A line copied without your own numbers is a guess.
Common mistake: Measuring drawdown in dollars after changing size. A −$900 week at double size and a −$450 week at normal size are the same −7.5R. Compare drawdowns in R, or a size change will hide or invent a problem.
Drill
Take your last 100 trades in R from your journal. If you have fewer, use a 45% win rate at +2R/−1R. In a spreadsheet, shuffle the order 20 times (for example, sort by a column of RAND()), and for each order record the deepest peak-to-trough drop. Sort the 20 numbers. Where would you put your max-drawdown line, and how often would an unchanged method reach it?
Solution
With 20 shuffles the 2nd-deepest of 20 sits near the 90th percentile and the deepest near the 95th; the estimate is rough, so more shuffles give a steadier answer. For the 45% at +2R/−1R method over 100 trades, our 200,000-path simulation gives a median of 8R, 12R at the 90th percentile and 14R at the 95th, and about 4% of 100-trade paths reach −15R. A −15R line therefore sits beyond most of this method’s ordinary drawdowns over 100 trades. Over more trades the drawdowns deepen, so a line set for 100 trades fires more often across a year. Write down the line, the horizon you sized it for, and the rate at which it would fire on an intact method.
What we don’t know yet
- Whether daily, weekly or max-drawdown limits improve real traders’ long-run results, and which numbers work best: not yet measured.
- How much more risk NQ day traders take after a losing morning: not yet measured; the studies above are in experiments and Treasury bond futures.
- How much trades cluster in time for a given method (several losers on the same kind of day). Clustering deepens drawdowns beyond the independent-trade model above; measure it in your own journal.
Recap
- Drawdown is peak-to-trough in R. A percentage loss d needs d ÷ (1 − d) to recover: −50% needs +100%.
- The three limits (−2R daily, −6R weekly, −15R max) are house rules, each on its own clock: session, week, method.
- Size the limits from your own drawdown distribution; an intact method still hits each one at a known rate.
- After losses, people tend to take more risk to get back to even (measured in other markets, plausible for you), so write the three urges and refuse them in advance.
Further reading
- Thaler, R. H. and Johnson, E. J. (1990), “Gambling with the house money and trying to break even: the effects of prior outcomes on risky choice”, Management Science 36(6): the break-even effect behind “one more to get it back”.
- Coval, J. D. and Shumway, T. (2005), “Do behavioral biases affect prices?”, Journal of Finance 60(1): CBOT futures traders took more afternoon risk after morning losses.
- Our lab: loss limits have not been tested on real trading results; the numbers in this lesson are a synthetic model, and RSK1.08 shows how to derive your own.
Connects to: M7.07 (why a run of results says little about the method) → M7.09, the one-page plan where these three limits are written down.
Educational content only — not financial advice. Futures trading involves substantial risk of loss.
Saved in this browser (account sync comes at launch).