ForumSections

How many consecutive losing trades would make you pause trading? - page 2

New comment
Events of 2 and 3 losses are different. The event of 2 losses is a sign of warning and trading could continue. The event of 3 losses is a sign of error and it makes us thinking of pausing trading.
Minh Tue Le #:
Events of 2 and 3 losses are different. The event of 2 losses is a sign of warning and trading could continue. The event of 3 losses is a sign of error and it makes us thinking of pausing trading.
While that may be true for your particular trading strategy, that is not a case of one size fits all. As has been said repeatedly in this Forum, each trader must collect consecutive losses statistics from testing her/his particular strategy and then let that determine the count of average losses + 1.
Ryan L Johnson #:
... let that determine the count of average losses + 1

The problem lies in Average Losses + 1. This is the biggest weakness of the analysis. Let's say a strategy has 1,000 trades and the statistics show:

- Average consecutive losses = 3.

- Maximum consecutive losses = 9.

According to the formula above, the trader will pause the strategy when 4 consecutive losses occur. But this can lead to two problems:

First: Stopping the trade too early.

Four consecutive losses can be perfectly normal for that strategy. Pausing at this point could cause the trader to miss subsequent recovery trades.

Second: Average doesn't reflect the probability of an abnormal losing streak.

Average is an average value, not a probability threshold or a statistical anomaly threshold. A strategy with an average losing streak of 3 doesn't mean the 4th streak is abnormal.


Minh Tue Le #:
First: Stopping the trade too early.

Four consecutive losses can be perfectly normal for that strategy. Pausing at this point could cause the trader to miss subsequent recovery trades.

That is flawed logic. What part of the technical word, average, do you not understand?  Also, machine state monitoring can be used to reactivate trading when conditionally appropriate.

Minh Tue Le #:
Second: Average doesn't reflect the probability of an abnormal losing streak.

Average is an average value, not a probability threshold or a statistical anomaly threshold. A strategy with an average losing streak of 3 doesn't mean the 4th streak is abnormal.

That logic suffers from the same flaw.

Any trader who is proficient at using the Tester properly and accurately, and reading the Report, can obtain the average number of losses that, statically, the strategy typically generates.

If you genuinely detest an average (mean), feel free to custom calculate the mode or the median.

Frankly the vague words, normal and abnormal, have little to no meaning in terms of quantifying anything other than FOMO.

Hi, 
2 to 3 loss depending on my risk per trade. Some times more than 5 also. It's all depends on market volatility as well. Some time pausing the trade for some hours then resume also helps.
Ryan L Johnson #:

Assuming that a sufficient sample size is tested, the bigger loser(s) show up in the Maximal consecutive profot/loss stats:



Thanks Ryan,

agreed, with enough trades the max consecutive loss in money shows a streak that got expensive. I meant something a bit different. Let's say live the losses don't come in a row, but each one is twice the size it was in the backtest. The streak count stays normal and so does the max streak. Only the average loss shows it, so i compare that with the backtest too.

About average + 1 vs. @Minh Tue Le. I use two lines for that. At average + 1 i look closer at the trades, past the backtest max i pause the EA. For me thats where both points fit together.

Chris :)
Christian Kotschote #:
[A]greed, with enough trades the max consecutive loss in money shows a streak that got expensive. I meant something a bit different. Let's say live the losses don't come in a row, but each one is twice the size it was in the backtest. The streak count stays normal and so does the max streak. Only the average loss shows it, so i compare that with the backtest too.

Ah yes, that is an important point. My dynamic exit is basically a "hidden stop" at a fixed amount +/- spread and allowed slippage.

If your dynamic exit is literally dynamic in its distance from entry (unlike my own), you definitely need to analyze both the count and the amounts. In retrospect, perhaps Minh was alluding to that.

Depends on the winrate of your strategy, historical max drawdown to know if it's within ''normal'', also depends on your risk % to know if you can manage not so ''normal'' environment, if the risked %  exceeds the min lot in the market you're trading, and depends on the current balance. Just saying i stop after 2 or 66 is kinda senseless. Only data can tell you the correct answer. what i would do if i didn't have an ea and traded manually, first identify the regime your strategy works best at, if you feel like the market is in your favor (i'd take my worst losing 2 weeks, line up all the losing trades and make their total as my ''stop trading threshhold'', with this n of losing trades you can determine the risk %) if it feels like it's not the regime your strategy excels at, cut the stop trading threshhold in half. NOW what i just said is very variable and depends heavily on the n of trades per week, if you trade 2 times a week you shouldn't put 2 trades limit and risk 50% per trade.
I think this depends largely on whether you're a scalper or a swing trader. Scalpers generally execute more trades, while swing traders tend to take fewer trades and hold positions longer.

The poll asks the right question but measures the wrong thing. Consecutive losses are not evidence that anything changed - they are the expected texture of any edge with a win rate below 100 percent.

Before you use a losing streak as a signal to stop, ask what a streak of that length would look like if your system were working exactly as designed. With a 40 percent win rate the probability of a run of 5 or more losses inside 100 trades is high enough that stopping on it is not risk management, it is just abandoning the sample.

The number worth setting is not a loss count but a drawdown in currency that you decided in advance, while calm, was the point at which you would admit the premise was wrong. A streak is noise. Drawdown past your tested maximum is information.

123
New comment