Reverse Martingale position sizing

[Gringotts Bank]

Many of you would have heard of the Martingale system of betting (trading). If you are winning, keep betting the same amount (position size). At your first loss, you double up, and continue to do so til you win back everything you've lost, and start over. It's not actually doubling, (there's a simple formula), but you get what I'm saying. You might get lucky if you don't trade for too long, but the likelihood of not having the required capital to cover your nth losing bet is high. Not recommended.

Reverse Martingale can be used successfully with share trading, I believe. If your first bet wins you increase your bet size. The more you win the more you increase the increment (with some limits obviously). The idea behind this system is that it makes use of the 'winning streak' phenomenon. First loss will cause you to reduce your bet size. Lose again, reduce it again... and so on.

Probably best for high frequency trading systems.

 

[Garpal Gumnut]

Many of you would have heard of the Martingale system of betting (trading). If you are winning, keep betting the same amount (position size). At your first loss, you double up, and continue to do so til you win back everything you've lost, and start over. It's not actually doubling, (there's a simple formula), but you get what I'm saying. You might get lucky if you don't trade for too long, but the likelihood of not having the required capital to cover your nth losing bet is high. Not recommended.

Reverse Martingale can be used successfully with share trading, I believe. If your first bet wins you increase your bet size. The more you win the more you increase the increment (with some limits obviously). The idea behind this system is that it makes use of the 'winning streak' phenomenon. First loss will cause you to reduce your bet size. Lose again, reduce it again... and so on.

Probably best for high frequency trading systems.

I can see a number of problems with the system.

1. You may not get all your order for stocks or instruments filled at the price you wish
2. Brokerage
3. Taxes and Bookkeeping on the Trades.
4. A run of very bad luck.

gg

 

[Gringotts Bank]

Sigh.....

None of your objections is relevant. Read it again, it's a position sizing method.

 

[Garpal Gumnut]

Sigh.....

None of your objections is relevant. Read it again, it's a position sizing method.

Sorry sirG I didn't mean to make you sigh.

I still reckon you'll go broke using that system, its a bit like replacing Rudd with Gillard.

sigh.....

gg

 

[Gringotts Bank]

But it's not a system of trading, it's a system of deciding how much to put on each trade of your existing system.

It actually has safeguards against going broke through a run of bad luck. It aims to give you less drawdown in times of losing streaks. On the upside, it aims to capitalize on winning streaks (if you believe in such things).

Gillard has longer earlobes. Should count for something.

 

[tech/a]

A subject rarely touched on but a very important one.
It makes sense and is a good idea.
There will always be strings of winners and losers in any system
However anticipating through analysis when those streaks are more likely is a wise idea.
Equity curve stops seem to work well but havent tested fully yet.

Seen and heard of it before it basically increases your risk each winning trade.
If you move to B/E (Or minimum risk) as soon as possible the risk is no more than any other trade.

Penfold discusses and tests a number of really interesting position size models in "The Universal Principals of Successful Trading"
Worth buying the book if for nothing else.

Mind you there is plenty of "ELSE".

 

[skc]

The underlying reason why martingale systems don't work is because the next trade has the same probability of success as any other individual trade.

Of couse we will observe winning streak - that's the nature of statistics. Streak probabilities matter only when you look at them before the start of the streak, not after the streak has begun.

I am doubtful that the suggested reverse martingale actually improves the edge.

It improves the outcome of course on a positive expectancy system - you trade bigger so you will have a bigger payout over time. But you have not improved your edge. By that I think the improvement in total profit will be linear to the increase in position size.

Happy to be shown that this is wrong however.

 

[AlterEgo]

Fixed fractional position sizing already does this. As the account size increases, so do the position sizes (well technically it's the $ risked per trade that increases, but that's closely related to position size). As the account size shrinks, the position sizes reduce as well. But I guess you're talking about a much more aggressive approach than this?

 

[howardbandy]

Many of you would have heard of the Martingale system of betting (trading). If you are winning, keep betting the same amount (position size). At your first loss, you double up, and continue to do so til you win back everything you've lost, and start over. It's not actually doubling, (there's a simple formula), but you get what I'm saying. You might get lucky if you don't trade for too long, but the likelihood of not having the required capital to cover your nth losing bet is high. Not recommended.

Reverse Martingale can be used successfully with share trading, I believe. If your first bet wins you increase your bet size. The more you win the more you increase the increment (with some limits obviously). The idea behind this system is that it makes use of the 'winning streak' phenomenon. First loss will cause you to reduce your bet size. Lose again, reduce it again... and so on.

Probably best for high frequency trading systems.

Greetings --

As several other responders have commented, this is a position sizing method. It relies on having a trading system that has a positive expectancy, then increasing position size when the system is winning.

Thought of in gambling terminology:
1. Anti-martingale methods are based on "betting the run of the table". That is, bet that winning will be followed by winning.
2. Increase bet size by using "the house money".

There are several good books that discuss position sizing:
1. Any of Ralph Vince's books. His latest is:
http://www.amazon.com/Leverage-Spac...=sr_1_2?ie=UTF8&s=books&qid=1283092336&sr=8-2
2. Ryan Jones.
http://www.amazon.com/Trading-Game-...=sr_1_1?s=books&ie=UTF8&qid=1283092386&sr=1-1
3. Van Tharp.
http://www.amazon.com/Trade-Your-Wa...=sr_1_1?s=books&ie=UTF8&qid=1283092422&sr=1-1
4. Advanced AmiBroker, when it becomes available.

Thanks,
Howard

 

[Gringotts Bank]

Correct alterego, it's exactly the same as fixed fractional betting but more aggressive. Achieves the very same thing.

skc, I understand that the next coin toss will always be 50-50, mathematically speaking. However I'm looking at this from a psychological perspective, not mathematical. My assumption: that a win will tip the odds in your favour of picking correctly the next time because you're "on a roll". The more of a roll you're on, the more you should bet, and vice versa. Many of the hard stat/maths guys will look at this assumption and say "yeh right!". Well that's fine, I do happen to believe in streaks that have a psychological cause (ie. non random).

Getting back to the statistical side of things, reverse Martingale does have 2 advantages over Martingale, as mentioned in my earlier posts.

I favour this over Van Tharp's stuff purely because of it's simplicity and psychological bent. Van Tharp is the limit of my reading on this topic to date, but thanks for all those other great refs people have included here.

Looking forward to Advanced Amibroker, Howard. What are they planning to include?

 

[Gringotts Bank]

I thought Advanced Ami was a software upgrade, but I see it's howard's book.

http://www.advancedamibroker.com/advanced-amiBroker-contents.pdf

 

[waza1960]

I have a presentation from Dr Bruce Vanstone of Bond Uni who has done work on Anti-Martingale systems and they seem to have an edge.

 

[Gringotts Bank]

Can you post a link or original text waza?

 

[tech/a]

Can you post a link or original text waza?

Send me a private message with your email address and I'll send you his thesis.

 

[LimitBid100]

Gringotts,

In my trading system, I use risk management that increase trade size as my capital increase (ie, after a winning trade), as you suggest in this post.

However, more importantly, I use a leverage increasing risk management style within each trade as well. Ie, I open a trade with a tiny amount, and then add more as it goes in the money.....eventually ending up with a suitable size position.

I find this "micro reverse martingale" (for a lack of a better term) adds far more value to my trading than the reverse martingale risk management at a portfolio level. (That said, the portfolio level risk management is essential, and does add plenty of value on its own).

To put it in perspective for you, my initial position is only 1/6th of my final position. This means, that if a trade goes wrong quickly, i only lose on a tiny position. If it goes well....I trade up to a full position and let it run as best i can.

Hope this helps,

LimitBid.

 

[Naked shorts]

I have found that adjusting my position sizing relative to my confidence in the trade, to be optimal. Discretion in position sizing is an edge in itself, but is extreamly difficult to master and takes hundreds, if not thousands of hours to get right

"You got to know when to hold 'em, know when to fold 'em, / Know when to walk away and know when to run"

 

[Gringotts Bank]

Sounds good Limit. My current system won't permit scaling in/out (due to brokerage costs) but I can see how that could add quite a bit to many systems. Reminds me, Amibroker does allow coding for scaling positions and I should go and learn how it works.

Naked, there's a lot of truth in the lyrics of 'The Gambler' eh?

 

[LimitBid100]

I dont see how scaling in and out adds to your brokerage....

If you do 2 contract, then 4 more, then another 6......you pay for 12 contracts.........if you dont scale in and just do 12 straight up, you still pay for 12 contracts?

Am I missing something here?

 

[howardbandy]

Gringotts,

In my trading system, I use risk management that increase trade size as my capital increase (ie, after a winning trade), as you suggest in this post.

However, more importantly, I use a leverage increasing risk management style within each trade as well. Ie, I open a trade with a tiny amount, and then add more as it goes in the money.....eventually ending up with a suitable size position.

I find this "micro reverse martingale" (for a lack of a better term) adds far more value to my trading than the reverse martingale risk management at a portfolio level. (That said, the portfolio level risk management is essential, and does add plenty of value on its own).

To put it in perspective for you, my initial position is only 1/6th of my final position. This means, that if a trade goes wrong quickly, i only lose on a tiny position. If it goes well....I trade up to a full position and let it run as best i can.

Hope this helps,

LimitBid.

Hi LimitBid --

Using one of your examples, where the initial position is one sixth of the final position, I assume that the size of the final position is the size that limits your risk. (If it is the initial one-sixth that limits your risk, then every scale in amount is putting your account at greater risk than you probably intend.) If the account risk is one percent of equity, and the final position goes badly, then no more than one percent is lost. If the trade goes well, and it is expected to go well since every system being traded has a positive expectancy, then the amount participating in the trade was significantly less than allowed by the risk management calculation and you are leaving considerable profit "on the table".

In general, scaling in hurts overall system performance. And scaling out hurts overall system performance. Best results come from calculating both the account risk and the system risk, then taking the maximum position allowed at the entry signal and exiting the entire position at the exit signal.

A good way to test is to isolate each signal. That is, create a separate signal for each scale in, including the initial entry, and a separate signal for each scale out. Compare the results from following each of the signals with a fixed amount.

Thanks for listening,
Howard

 

[LimitBid100]

Howard,

I find that more than half of my trade lose money (actually closer to 70% lose money). Most of these go bad straight away, createing a loss on the 1/6th position.

The trades that go well.....go really well, far outweighing the losses on my many bad trades.

Essentially, the scaling saves me a lot of money by reducing most of my losses to a trivial amount (about 0.12% of capital), while only reducing my winning trades (which are far less in number) by a smaller proportion.

There is no doubt that my winning trades would make more money if i didnt scale in. But my losing trade would cost me a lot more as well........much more than the extra gains in the winning trades.

LimitBid.