For some traders, the drawdowns in the Martingale system are just too scary to live with. That “roller coaster” ride ends-up causing them sleepless nights and stomach ulcers.
If this sounds like you, there is an alternative. The anti Martingale system does what many traders think is more logical. “Martingale in reverse” hangs on to winning trades, and drops losers. If that sounds better, read on.
“Doubling-Up” On Winners
The standard Martingale system closes winners and doubles exposure on losing trades. If you’re not familiar with this strategy, I wrote about it last week here on Forexop. While it has some highly desirable properties, the downside with it is that it can cause losses to run up exponentially.
The reverse Martingale, as I’m going to describe now does the exact opposite. It closes losing trades, and doubles winners. The idea being to cut losses quickly and let profits run.
Anti Martingale is an effective trend following strategy. Unlike forward Martingale it doesn’t have “fat tail” risks.
Take the following example in Table 1. This shows how a “double up” sequence works in practice. I’ve set a virtual take profit, and stop loss target of 20 pips. The price starts at 1.3500. I start by placing a buy to open order. The price then moves up 20 pips to 1.3520. Following the strategy, I now double the size of my position. I add 1 lot at the new rate of 1.3520.
|Tick||Price||Change||Order||Lots||Total||Entry Price||Avg. Entry Price||Float||Realized|
Table 1: A “Double up” trade sequence.
This gives me an average entry price of 1.3510. I continue doubling-up the exposure for each 20 pip increment (my definition of a winning trade).
At tick 6, the price then drops by 20 pips. Following the reverse strategy, I now have to close the last position. Since it is a loser (according to my criteria). So I place a sell to close order at tick 6. The effect of this is to half my position size, or exposure. I now hold 8 lots instead of 16.
I’ve also realized a loss of -$16 on the losing trade. My net balance is now -$2. The table below shows how the overall balance is made up. At tick 6, the P&L of each of the positions is as follows:
|Trade||Lots||Entry Price||Current Price||Change Pips||Trade P&L|
Table 2: Snapshot of trade P&L on closing first position.
Notice that the last losing trade wiped out all of the profit on the existing open trades, and left me with a net loss of -$2. The net loss of the entire sequence is equal to my stop loss value. This relationship always holds.
The total profit of the group has been canceled out by the last move of just 20 pips. At this point in time there are still four open positions remaining. The first three are in profit and the last is at break even.
The question is then what to do with these left over positions.
Standard reversal approach At this point, some traders consider that the trend has reversed. So they cash in the profit on the remaining trades before further losses occur. This is what you’d do if you’re mirroring a pure Martingale strategy.
Hybrid approach Other traders prefer to hold the existing positions and wait. They do this especially if their indicators suggests the original trend might return. This is a hybrid strategy: In a pure reversal system, the trades in the entire sequence would all be closed at this point.
To see the whole process in action, you can download my Excel sheet which demonstrates the anti Martingale strategy:
The spreadsheet lets you to try-out various setups and market conditions. You can test the above variations in the algorithm by setting the “lose multiplier” parameter.
Like the original Martingale, the take profit and stop losses are “virtual”, in that they define winning and losing trades. And so when to increase or reduce exposure.
As with grid trading, you would typically also set a profit target for the “entire system”. Say for example after N “double up” legs or when the entire system of open positions reaches a certain profit amount.
Doubling Up Can Work Against You
As shown above, the lot-doubling, which marks the Martingale approach, can work against you too. With the reverse-Martingale, the averaging up rather than down means your profits can be turned very quickly into loses should the market turn against you.
On the plus side, your loss from a single sequence is limited to your stop loss on your starting lot amount. So say your stop loss is 40 pips, and your starting lot size is 1 micro lot, your biggest loss from a single sequence would be: 40 pips x 1 micro lot. That’s about $4 – depending on the currencies you’re trading.
Just as standard Martingale recovers losses on one winning trade. Anti-Martingale does the exact opposite. One losing trade in a double-up progression eliminates the profit of the entire sequence.
This can be seen in action in Tables 1 and 2. The “hitting of stops” can be a significant problem when the price action is especially volatile.
The Strategy is Risk Balanced
When applied systematically, both Martingale and anti Martingale have equal risk verses reward. That is, they are risk-reward balanced. So say your success in picking trades is no better than chance. This means the system has a 1:1 risk-reward ratio and a net expected return of zero.
The analysis is just the reverse of the Martingale. Every losing trade is closed at it’s stop loss. And there’s an equal probability of picking winning verses losing trades. So the expected loss from the losers is:
E ≈ -½ N x B
Where N is the total number of trades, and B is the fixed amount of loss on each trade. In the anti Martingale, let’s say I close the system and take profit after 8 winners in a row. That is, after doubling-up 7 times.
The probability of 8 winners in a row is (1/2)8. That means I’d expect this to happen after 28, or after 256 trades.
So after 256 trades:
- My expected loss from the losers: (½) x 256 x 1 = -128
- My expected profit from the one winning sequence: 27 x 1 = 128
- My expected net return: 0
This is because in this setup all other combinations, other than 8 winners in a row results in a loss. The same is true whichever number you choose.
In practice of course, your expected net return, and risk-reward will actually be slightly less than zero. This is because of the spreads.
Avoid Those Scary Drawdowns
The standard Martingale system blindly doubles down on consecutive losing trades. Sometimes this takes the trader into the abyss with disastrous results.
The profit-loss pattern of anti Martingale is the opposite of this. Typically, in this strategy you see frequent small losses, and a few one off big wins. You rarely see the scary drawdowns that you get with Martingale. This can be seen by comparing the two return charts. See how much smoother the returns in the reverse strategy are.
The reason being that loss exposure is cut rapidly, and doesn’t escalate at an exponential rate. The “saw tooth” appearance of Figure 5 shows these “rare but catastrophic” losses.
You will still see losses in the reverse system, but these are more contained.
Choosing An Entry Signal
In my spreadsheet I’ve used the first derivative of the 15 day moving average line. This is a very basic indicator and simply tells me if there’s a trend, and in which direction.
With anti-martingale, the indicator should “say” when there’s a high probability of an existing trend, or the start of a new trend.
The following technical indicators can be useful in deciding your entry signal:
- Moving average lines, and convergences (more)
- Volatility “squeezes”
- Break through of key support/resistance levels
- Structural chart patterns
- Momentum indicators
- ADX/Elliott Waves
Some of these are more subjective in interpretation and are difficult to automate. However the stronger the combined trend signal you have, the higher the chances of a profitable trade sequence.
Warning Beware of relying on technical indicators on their own. Ideally, if you’re trading manually or creating an expert advisor, you should incorporate a fundamental viewpoint as well.
This is harder to do if you’re using automation. But then when did anything easy ever make a profit?
To see the potential for false signals, download the spreadsheet and take a look at the chart. Press F9 a few times to run the calculations. You’ll notice common technical patterns appear in there time and again. Namely trends, tops, bottoms, head and shoulder patterns. Even Fibonacci levels and supports and resistances appear to be there.
But this shouldn’t happen! This data is entirely random and simulated. It goes to show, as humans we’re pre-programmed to see patterns and relationships. Even when they don’t exist.
Short term performance can be misleading with any trading system. To analyze the long-term behavior, I ran the simulation 1,000 times in each set using a random pricing model. Each set simulates different market characteristics using the trend “drift” parameter in the spreadsheet:
- Flat market (trend parameter=0)
- Bullish market (trend parameter=0.1)
- Bearish market (trend parameter=-0.1)
An average spread of 4 pips was also used. The results are summarized in Table 3.
|Market||Mean (pips/lot)||Standard Deviation||Mode (pips/lot)||Kurtosis||Skew|
Table 3: Anti-Martingale – Summary of long-term performance.
The important thing to take away from this is the marked performance differences. Anti Martingale doesn’t do well in flat markets. See the huge difference in the mean returns. In fact, it does far worse than random. This highlights the importance of choosing the right strategy for the right market.
Figure 1 shows a typical profit pattern from a single run. Compare this to Martingale, in which the drawdowns are frequent and severe. Click here to open image in new window.
The figure above shows the frequency distribution of each of the different market conditions. Notice how the distribution of returns for “trending” is shifted to the right. This represents the higher returns.
The following Excel spreadsheet will allow you to test the strategy yourself and try out different scenarios.
You can configure different volatility and trending conditions, then see first hand how the algorithm behaves.
Anti Martingale is Better In Trending Markets
There’s no point running both Martingale and Anti-Martingale at the same time, in the same market and with the same setup. The strategies are opposites, and suited to different situations.
While standard Martingale works better in flat, range bound markets, anti Martingale is better suited to volatile, trending markets. That’s not so say it won’t work sometimes in flat or trendless conditions. It’s just the idea behind it is to escalate the exposure on a rising or falling market. This is where most of the big profits are made.
The “preference” for trends makes the reverse algorithm better suited to trading volatile pairs or for positive “carry” opportunities.
Table 4 below shows the long term performance characteristics of both algorithms. The data is based on 1,000 runs of both algorithms under different market conditions: flat, bullish, and bearish. Each run can execute up to 200 trades. This sample data therefore consists of 1.2 million data points (1000 x 6 x 200).
Average Return (Pips/Lot)
Distributions (All Markets)
Table 4: Performance comparison Anti-Martingale vs. Martingale.
Figure 3 below shows the return distributions of both strategies. As can be seen, the distribution of Martingale is highly peaked with a double “fat tail”. The most negative of which is well “off the chart”. The worst case run returned a massive loss of -772 pips/lot – shown in Table 3 above.
The long term averages, as shown in Table 4, highlight the variability of performance, depending on market conditions. Anti Martingale gives a much stronger mean return in a rising/falling market.
However, this is exactly where the conventional strategy suffers. Mostly due to “doubling down” against prevailing trends. Whether the trend is bullish or bearish doesn’t have a significant impact in the outcome. The heavy tail results in a very large kurtosis.
The figure above shows the long-term cumulative gains in pips for Anti Martingale. Note that the return graph is significantly smoother than the standard Martingale returns below.
In Figure 5, you can see the returns from Martingale show a characteristic “saw tooth”. These demonstrate the big “one off” losses that happen in the “classic algorithm”.
The “Bottom Line”
- The reverse strategy is much better suited to volatile, trending markets. However, Martingale gives better results in flat, predominantly trend-less markets.
- Both strategies yield an expected long term return of zero. Therefore, choice of suitable currency pair, market conditions and entry signal are critical to success with either strategy (see return charts).
- Standard Martingale is characterized by steady, positive returns, and “one off” big losses. These appear as “fat tails“ in the return distribution (see comparison return graphs). These lead to highly variable outcomes in the long run.
- The return distribution of Anti Martingale is significantly flatter, with lower variance as overall returns are more “clustered around the mean”.
- Optimal long term returns can be achieved by “flipping” between the two strategies according to the market conditions. This can be automated if you’re using and Expert Advisor.
Why Use It
- It decreases exposure on losses, and increases it on profits. Most traders believe that makes more sense than doing the opposite.
- Like Martingale, it has an outcome and risk-reward that can be statistically estimated.
- It’s well suited to algorithmic trading.
- It doesn’t encounter exponentially increasing losses – provided stops and take profits are correctly executed.
- Using the forward system it’s difficult to profit from trends. Even when a strong trend is detected, the upside is limited to a linear progression.
Why Avoid It
- The doubling up of position sizes can work against you. If your biggest trade loses, it wipes out the gains for the entire sequence.
- The big lot size multiples means there’s a risk of heavy losses if your stops are overrun. This can happen if the market gaps and falls through your stop levels.
- Execution problems with your broker can also cause stops to fail or to execute at levels that make the outcome unprofitable. However this is a problem most algorithmic strategies are prone to.