forex-scalping.com

A stop after three losses in a row: how often does it fire, and what does it do to expectancy?

A rule many people use: three losers in a row and you're done for the day. How often does it trigger? For independent trades, the chance of at least one run of three losses within n trades, for a system that loses 55% of the time:

5 trades: 32%
8 trades: 51%
10 trades: 61%
15 trades: 77%

For a system that loses 40% of the time: 14%, 25%, 31% and 45%.

I simulated 8 trades a day for a 45% winner at 1.5R, 200,000 days, with and without the rule. Without it, the average day is +1.00R over 8 trades. With the rule, the average is 6.4 trades and +0.80R. The rule fired on about half the days. The share of days losing 5R or more falls from 6.4% to zero, and days losing 3R or more go from 22% to 21%.

So the rule doesn't improve expectancy per trade, since each trade is independent. It cuts the number of trades, which removes about a fifth of the daily gain, and it cuts the very worst days. That's a real trade: some profit for a lower tail. Whether it's good depends on what the worst days do to you.

Interesting that the rule fires on about half of all days at 8 trades. The next question is whether trades after three losses actually perform worse in real journals, since the simulation assumes independence. If losses cluster, for example on choppy days or after tilt, the rule may help more than the simulation shows. If they don't, it simply cuts trading time. Checking one's own data for the outcome of the fourth trade after three losers is the way to find out.

I checked exactly that in my own journal, because I had used a three losses rule for over a year without ever testing it. Over roughly 900 trades on EURUSD and GBPUSD, the fourth trade after three losers won noticeably less often than my overall average, around eight percentage points lower. So for me the losses were not independent at all.

When I looked closer, most of the cluster came from two situations. One was choppy days when the daily range was well below normal, where my breakout entries kept failing one after another. The other was my own behaviour, taking entries a little earlier and with looser criteria once I was down for the day.

That made the rule worth keeping for me, though I changed it slightly. Instead of stopping completely, I now take a thirty minute break after three losses and then allow one more trade only if the setup scores fully on my checklist. If that one loses too, I am done for the day.

The case where the rule does help is when the trades aren't independent, when losing runs make you trade worse. You can test that in a journal. Compare the average result of trades taken after two or more losses with all other trades. If trades after losses average -0.3R and others +0.15R, the rule is removing the bad ones. If both average about the same, then the rule is only reducing volume. You need around 100 trades in each bucket to say much, so for most people a rule like this is a decision made on a feeling until the journal has a few hundred rows.

In dollars, the cost is easy to see. Using $100 per R, the simulated day gives +$100 without the rule and +$80 with it, a difference of about $20 a day. Over 21 trading days that's roughly $410 a month. The rule's benefit is the worst-day cap: no more days of 5R or worse, which is $500 on those days. Whether $410 a month is a fair price for removing 6.4% of days at -$500 or worse depends on what a $500 day does to your decisions. If it leads to a revenge trade or a skipped week, the price may be reasonable.

The trigger can be set differently, and the numbers change a lot. Using the same 55% loss rate in an 8 trade day, a rule of two losses in a row fires on about 85% of days, three on 51%, and four on 26%. So two in a row is a rule that applies almost every day, and four is one that rarely does. A stop after two losses is basically a smaller daily trade count, while four losses is a real emergency brake. Which one fits depends on how many trades you take per day, so pick the trigger from the table and not from habit.

To make it mechanical, don't leave it to judgment at the moment you're annoyed. Write it as a line on the plan: after three consecutive losses close the platform and write the date. In an EA, count consecutive losses from the deal history, with a reset at the start of the next trading day, and block new entries when it reaches three. On paper, three boxes on a sticky note, ticked on each loss and cleared on each win. The rule works when it's decided in advance and applied without exceptions. An exception on the first day it applies is the day the rule matters.