What would be a good filter for a Mean-Reverting system?

Oadmani said:
Now here is the issue. I just checked my notes, I have a system, with a win percentage of 61.79 with Average Gain/Average Loss ratio of 1.5. The winning percentage is lower, whereas the reward/risk is higher.

However, I don't know how much to risk per trade, given the constraints that I would probably have 15-17 trades at a time on average (this number also includes trades from a trend-following system that I would be employing at the same time as the mean-reverting one). They can go upto 20 in periods and be as low as 7-8 in periods.

This system has 18 consecutive losses when covid hit the markets. How do I filter out some of those trades? These 18 consecutive losses have an average loss which is more than the average loss of the system. It means that I could lose around 20-25 percent of my capital because of one bad period.
More...

If losing 'streaks' are the primary issue, then employ some methodology to stop taking positions after e.g. 3 losing trades in a row, and wait for a virtual winner then start to trade again. This will protect you from a long series of losing trades, but will also mean you miss out on each winning virtual trade, which may affect your overall w/l ratio. You'd need to back-test the approach to understand how it impacts you overall. Back-testing it will also allow you to understand the optimum 'losers in a row' cut-off. As always with back-testing though, it's no guarantee of future performance.

TDUK
 
Oadmani said:
Now here is the issue. I just checked my notes, I have a system, with a win percentage of 61.79 with Average Gain/Average Loss ratio of 1.5. The winning percentage is lower, whereas the reward/risk is higher.

However, I don't know how much to risk per trade, given the constraints that I would probably have 15-17 trades at a time on average (this number also includes trades from a trend-following system that I would be employing at the same time as the mean-reverting one). They can go upto 20 in periods and be as low as 7-8 in periods.

This system has 18 consecutive losses when covid hit the markets. How do I filter out some of those trades? These 18 consecutive losses have an average loss which is more than the average loss of the system. It means that I could lose around 20-25 percent of my capital because of one bad period.
More...

I also had to re-read this post. You are now stating a win ratio of 61.79 at a positive risk/reward of 1:1.5, i.e. you gain 1.5 for every 1 you risk?

That's VERY different to the aforementioned 67% win rate, at 1:1.

So what is it, or is this two different approaches?

Your 61.79% win rate at 1:1.5 is very close to the implied odds by that win rate of 60%, and I expect that spreads and commissions would negate any perceived edge.

TDUK
 
I thought about doing this. However, a lot of the losing streaks happen in clusters. For instance, in the covid crisis, the markets panicked, and 10 stocks together met the criteria for an entry of a mean-reversion trade on the long side. Now, the market went down further, and all of them turned out to be losses. There is simply no way to predict this, since the losing streak didn't happen one by one, but all the trades appeared together. I was thinking, if I get more than X number of trades, within 2 days, X being, say, 6...then I manage risk by exiting all the positions and don't take new positions. Of course, that could also mean that all of those trades become profitable, but it still mitigates the risk.

The other way of course is to follow the fundamentals, and if the market is about to crash, I avoid taking trades. The issue is that we cannot short stocks here, so where I could be trading in 40 stocks at a time and risk around 2.5 percent of the capital-- both long and short-- I can only take half of those trades, as shorting is not allowed on stocks here. Also, besides this limitation, I cannot also be profitable when the market collapses, and balance out my portfolio from the long side by shorting. So, the only option is to either follow the fundamentals or use some other trick.
 
TraderDaveUK said:
I also had to re-read this post. You are now stating a win ratio of 61.79 at a positive risk/reward of 1:1.5, i.e. you gain 1.5 for every 1 you risk?

That's VERY different to the aforementioned 67% win rate, at 1:1.

So what is it, or is this two different approaches?

Your 61.79% win rate at 1:1.5 is very close to the implied odds by that win rate of 60%, and I expect that spreads and commissions would negate any perceived edge.

TDUK
More...
I meant the average gain/average loss ratio is 1.5:1.
 
I don't think this is going anywhere and I don't want to take the thread too far off your original topic, so I'll butt out now. Best of luck to you.

TDUK
 
Oadmani said:
This system has 18 consecutive losses when covid hit the markets. How do I filter out some of those trades? These 18 consecutive losses have an average loss which is more than the average loss of the system. It means that I could lose around 20-25 percent of my capital because of one bad period.
More...
.

Oadmani,

Lots of good suggestions in this thread... I agree especially with the ideas put forth by Tomorton, Fonz and NazSherpa that moving averages can be used to improve mean reversion systems. Simply choosing not to take long trades when the slope of a medium-term moving average is down will eliminate lots of bad trades.

You mentioned the Covid correction last year... the attached chart shows how one could have avoided taking long trades during the retracement.


mean reversion w medium-term MA2_PNG500.png
 
For me most important is losing percentages, I can do very well with 15% winning percentages so long as breakeven percentages are 80% using hedges.

1:1.5 risk/reward and doing 62% is slow equity curve.
Examples of price movement to wait for better entries are H&S, or deep retracement within uptrend then higher low, risk exits would be recent lows and targeting highs.
Chart patterns work well and although some don't like Bollinger bands, combining trendlines with BB can help.

Just remember though, "volume" drives price, if you the only one getting Long, keep tight stops.
 
You must log in or register to reply here.
Back
Top