anti-martingale

[charliechan]

this is where after every successful outcome, i increase size by x, and every unsuccessful outcome, i decrease size by x. the martingale for those who dont know is a gambling method where one doubles up after a loss.


has any one had an unfortunate 'comedown' from this? am i missing somethin?

so far its looking great.

 

[TheBramble]

This is to a large extent what we all do over the long run, otherwise our trading capital simply wouldn’t grow. Taking the axiom of ‘cutting losses short and letting profits run’ out of the picture for a moment and looking at your anti-Martingale in isolation, I don’t think it is as clear cut or formulaic as you suggest, though I appreciate you’re asking from a ‘what if’ perspective rather than an empirical viewpoint.

Three factors to be considered: One is probability of a consecutive winning trade, another is Risk size and the other is the nature and size of ‘X’ and specifically whether that is symmetrical.

Your odds of the next trade being successful diminish rapidly as the size of your string of successful consecutive trades increases. So with an increasing position/risk size, when you do lose, which you have increasing odds of doing as any run of consecutive winning trades lengthens, it’s going to take a relatively large (relative to your last trade’s profits) slice out of the sum total of your previous winnings – obviously the size will depend on the size of ‘X’ you’re using. For those that will counter that each trade has exactly the same probability of winning:losing – yes it does, but we’re not looking at single trades in isolation, we are looking at them in the context of consecutive winning trades.

If the size of the risk per trade is not a constant (is it ever?), that would need to be factored in and fed into the overall formula and particularly; the size of ‘X’. Unless position size always and absolutely directly correlates to risk you need to tweak the formula to look at risk rather than position size.

The size of ‘X;’ would need to be such that at no time did it take your position or more importantly risk size to a value greater than the sum total of your previously built profits.

The ‘X’ factor could be tailored to mirror current market symmetry in that in some market phases it will pay to decrease your size of ‘X’ by a different value to that which you use to increase size/risk. You’d need to backtest that as a possible variable in various market phases of course to see if it flies.

I have a suspicion that a formula based anti-Martingale would need to consider too many factors and take too much monitoring to tweak in current market dynamics to ever make it a viable turn-key low-maintenance operation. However, if you do manage it, I would not be the only one to be interested in your methods and results.

Good luck.

 

[new_trader]

this is where after every successful outcome, i increase size by x, and every unsuccessful outcome, i decrease size by x. the martingale for those who dont know is a gambling method where one doubles up after a loss.


has any one had an unfortunate 'comedown' from this? am i missing somethin?

so far its looking great.

The Martingale system can only be used when the fixed odds are ~50%, hence "doubling" after a loss. It works using laws of probability and odds which don't change and cannot be influenced. Anyone who uses the Martingale system to trade knows nothing about trading or probability.

 

[Hoggums]

this is where after every successful outcome, i increase size by x, and every unsuccessful outcome, i decrease size by x. the martingale for those who dont know is a gambling method where one doubles up after a loss.


has any one had an unfortunate 'comedown' from this? am i missing somethin?

so far its looking great.

You've not specified any close conditions for the trade(s). Anyway it wouldn't work - this is another system I tested extensively a couple of years ago. The system would only make profit when the market is trending. When the market ranges (which happens more often that it trends) the succession of losses caused by reversals in market direction eats away into any profit. I tested on EUR.USD on all types of sizes of x on over 5 years of tick data - it consistently lost money despite having weeks with spectacular gains.

 

[fat chance]

The Martingale system can only be used when the fixed odds are ~50%, hence "doubling" after a loss. It works using laws of probability and odds which don't change and cannot be influenced.

Before some poor sod goes and tries it... the Martingale system does not "work" when the probability is 50% and will, given long enough, bankrupt anyone with finite resources.

Anyone who uses the Martingale system to trade knows nothing about trading or probability.

Doubling up during a losing trade can work very well on mean-reverting spreads. You need to know something about probability to design the spread, and something about trading to know when it might not mean-revert.

The "Anti-Martingale" system that CharlieChan describes seems like a pretty good idea to me - should be a good way to manage your risk, trade trending markets and breakouts (probably anything with high kurtosis), and deal with markets that go through phases of behaviour that don't suit your trading style.

 

[new_trader]

Before some poor sod goes and tries it... the Martingale system does not "work" when the probability is 50% and will, given long enough, bankrupt anyone with finite resources.



Doubling up during a losing trade can work very well on mean-reverting spreads. You need to know something about probability to design the spread, and something about trading to know when it might not mean-revert.

The "Anti-Martingale" system that CharlieChan describes seems like a pretty good idea to me - should be a good way to manage your risk, trade trending markets and breakouts (probably anything with high kurtosis), and deal with markets that go through phases of behaviour that don't suit your trading style.

I should have said the Theory or Martingale only applies to fixed odds of ~50% such as tossing a fair coin. The reason being that statistically, Heads and Tails should come up an equal number of times. If you have enough money, you should, in theory at least never lose. This only "works" if you choose one side, either heads or tails and stick with it.

Unlike the markets, a coin has no memory and is not influenced by economic conditions, wars, analysts or market makers. If everyone in the world bets on heads the chances of it landing Tails is still 50%. Heads or Tails is not subjective, it is either one or the other.

Assuming the market has only 2 directions, either Up or Down, the odds of either is NOT 50%. Up or down is subjective and depends on where YOU placed your trade. One traders 'Up' is another traders 'Down'. The number of consecutive losses can be infinite and can't be calculated prior to placing a trade.

 

[fat chance]

I should have said the Theory or Martingale only applies to fixed odds of ~50% such as tossing a fair coin. The reason being that statistically, Heads and Tails should come up an equal number of times.

Statistically, it is extremely unlikely that after a large number of tosses you will have tossed equal numbers of heads and tails. Ok, that's a little pedantic, but the relevance to trading is significant so I thought it was worth mentioning.

If you have enough money, you should, in theory at least never lose.

(enough = infinite)

This only "works" if you choose one side, either heads or tails and stick with it.

Now you're just winding me up.

Assuming the market has only 2 directions, either Up or Down, the odds of either is NOT 50%. Up or down is subjective and depends on where YOU placed your trade. One traders 'Up' is another traders 'Down'. The number of consecutive losses can be infinite and can't be calculated prior to placing a trade.

I think that if you go long or short randomly the probability of you being in profit after a specified time will be 50% (neglecting the bid-offer spread, commissions, fees), so you can use the Martingale method if you like / dare.

Going back to CharlieChan's original post, I'm sure he's not entering trades randomly and using the "Anti-Martingale" as some sort of magic system. He's probably using whatever indicators he finds useful to get in to /out of a trade, but adding to his winners and cutting his losers.

 

[alexander]

Your odds of the next trade being successful diminish rapidly as the size of your string of successful consecutive trades increases.

True if one has a proven trading system with a solid winning bias otherwise this turns out to be just a "gambler's fallacy":

http://en.wikipedia.org/wiki/Gambler's_fallacy

"What is the probability of flipping 21 heads in a row, with a fair coin? (Answer: 1 in 2,097,152 = approximately 0.000000477.) What is the probability of doing it, given that you have already flipped 20 heads in a row? (Answer: 0.5.) "

If you toss a fair coin, the probability of heads coming up next is 0.50 regardless of what happened in the past.

The gambler's fallacy results from the false confusion between facts (what happened before) and future outcomes (the outcome of a probabilistic process).

Most people do not have a system with a significant winning bias (although they believe so) and the martingale or anti-martingale method results in total disaster.

Alex

 

[new_trader]

Statistically, it is extremely unlikely that after a large number of tosses you will have tossed equal numbers of heads and tails. Ok, that's a little pedantic, but the relevance to trading is significant so I thought it was worth mentioning.

In fact it's the exact opposite of what you say.

This is why casino's set a maximum on outside bets on a Roulette wheel. The more times you flip a fair coin the closer it gets to 50%. If you flipped the coin only once and it came up heads would it be correct to say the odds of getting heads is 100%?

Maybe you should read this:

http://en.wikipedia.org/wiki/Law_of_large_numbers

Now you're just winding me up.

I think that if you go long or short randomly the probability of you being in profit after a specified time will be 50% (neglecting the bid-offer spread, commissions, fees), so you can use the Martingale method if you like / dare.

Now you're just winding me up!! You can't be serious at all.

 

[fat chance]

The more times you flip a fair coin the closer it gets to 50%.

Agreed of course. What I said was that the more times you flip a coin, the larger the magnitude of the difference in the number of heads and tails you have flipped is likely to become. This is why you can take a load of random traders and after some time find that some have done very badly and some have done very well; they don't all eventually break even.

I think there is really no point in us debating this... assuming that neither of us has any major problems with basic probability theory, all we will argue about is misinterpretation and misunderstanding of what the other has posted. I certainly won't resort to insulting you by posting links to shoolboy-level wikipedia pages.

 

[new_trader]

I certainly won't resort to insulting you by posting links to shoolboy-level wikipedia pages.

My apologies. I didn't think probability theory changed much from school-boy level to adult-level. You must know more than me.

 

[charliechan]

well, too put more meat on the bones for those who asked, heres a rough outline of what im doing to make it clearer...

i trade es with clip size of Y contracts. i basically trade ranges or congestions for a handful of ticks. im not trying to catch a trend, i always have a target in mind. the win rate is thus around 70%. so, i start with Y contracts. if i make a profit at the end of the session, i trade 2xY contracts the next day. if i win that day, i trade 3xY the next. if i lose that day, i reduce my size back to 2xY. if i win trading 2xY, i go back up to 3xY, 4xY etc.


i did take a bad knock on wednesday i admit - that wiped out several days of profit. ho hum! that is a bad side - you do take some pretty big hits when they come after a string of good sessions.i think the degrees eluded to this in his great post.

 

[alexander]

well, too put more meat on the bones for those who asked, heres a rough outline of what im doing to make it clearer...

i trade es with clip size of Y contracts. i basically trade ranges or congestions for a handful of ticks. im not trying to catch a trend, i always have a target in mind. the win rate is thus around 70%. so, i start with Y contracts. if i make a profit at the end of the session, i trade 2xY contracts the next day. if i win that day, i trade 3xY the next. if i lose that day, i reduce my size back to 2xY. if i win trading 2xY, i go back up to 3xY, 4xY etc.

Have you backtested this position management method to determine its risk? What is the maximum risk per position? How high is your leverage?

The problem with such methods is that the next losersmay come at the point when you have the maximum risk exposure and then you wipe out a good chunk of your previous gains.

IMO, the fixed fractional method is the best.

Alex

 

[TWI]

I have done a lot of work on this and in my opinion anti martingale is the way to go (Forget martingale, it is the route to potentially quick profits but ultimately ends in disaster).
The key is in exactly how much you increase or decrease your risk based on winning and losing trades. These certainly do not have to be linear in nature. It is best to look purely in terms of the amount of money risked on a trade rather than lot size and then work out from knowing your stop how much you can do based on that figure. In other words you may find lot size goes down after a winner if the stop is wider than the last trade. Also it is far more effective when applied across a portfolio of markets/systems rather than just one. It is important not to increase too quickly as a subsequent loss has to avoid wiping out all previous wins and also it is important to ensure that bet size decreases quickly enough that after a winning streak if you have a string of losers you are back to minimum bet before everything is blown. Over time it is the most effective and disciplined way I have found to increase bet size and hence returns.
The aim of the game after all when you are confident you have positive expectancy is to increase risk size as quickly as possible. Doing it in an undisciplined way i.e. before a system has taken itself there via a methodology is another quick way to blow up. Money management in terms of risk sizing is possibly the most critical aspect of trading to get right if you want to make proper money rather than just pocket change.

 

[charliechan]

fanx twi. your post is the sanity check i needed! you make an excellent point about using a fraction of capital rather than lot size due to variance of stop-out levels, volatility, etc. like all good ideas, it seems obvious when its pointed out, but truth is i didnt think of it.

looks like ive got some more research to do to find that level.

i agree 100% with your last coments. when youve got the expectancy, you gotta start using that leverage some more. ive never been an advocate of fixed fractional methods, and have always used a clip size based on my interpritation of probability that was quite basic and has enabled me to do pretty well, but with this anti-martingale method, its been like trading on crack! maybe thats the sign to reign it in a bit and use capital allocation as you suggest rather than another clip...

 

[alexander]

It is important not to increase too quickly as a subsequent loss has to avoid wiping out all previous wins and also it is important to ensure that bet size decreases quickly enough that after a winning streak if you have a string of losers you are back to minimum bet before everything is blown. Over time it is the most effective and disciplined way I have found to increase bet size and hence returns.

Sounds good in principle but how does one know a streak of winners or losers is about to end? This is the key question. It seems to me, unless one answers first this question, risk is increased rather than decreased and the method is subject to large expected drawdown.

I mean, anyone who proposes such method must also propose a way to determine whether a streak of winners or losers is about to end. But nobody has done that as far as I know. The end result is traders reducing risk too early or increasing risk to late with subsequent huge drawdowns.

Money management in terms of risk sizing is possibly the most critical aspect of trading to get right if you want to make proper money rather than just pocket change.

Yes, this is probably the most important statement made in these threads for a long time now. One must also add to it that at the same time money management can turn out to be a way of accelerating losses if not done properly. Implicitly, any position and risk management based on martingales either assumes knowledge of the future or affirmation of probability a posteriori , something that many doubt is possible.

Hence, in lieu of a a concrete and systematic way of implementing the method you proposed I better stick to fixed fractional risk and position sizing.


Alex

 

[Trader333]

In my view all you need to determine is the percent of your account that you wish to risk per trade and then position size relative to the volatility of the instrument being traded in the time frame you have chosen to trade.


Paul

 

[Crap Buddist]

well, too put more meat on the bones for those who asked, heres a rough outline of what im doing to make it clearer...

i trade es with clip size of Y contracts. i basically trade ranges or congestions for a handful of ticks. im not trying to catch a trend, i always have a target in mind. the win rate is thus around 70%. so, i start with Y contracts. if i make a profit at the end of the session, i trade 2xY contracts the next day. if i win that day, i trade 3xY the next. if i lose that day, i reduce my size back to 2xY. if i win trading 2xY, i go back up to 3xY, 4xY etc.


i did take a bad knock on wednesday i admit - that wiped out several days of profit. ho hum! that is a bad side - you do take some pretty big hits when they come after a string of good sessions.i think the degrees eluded to this in his great post.

If you like it what about this as a staking method.


1,2,4,5,6,7,8,9.just increase by one unit onwards until you hit a loss, then back to 1 and start over.

if you take a hit after 4 you still got 2 and not a wipeout . So lets you add on to take advantage of a roll , and indeed take advantage of any continued roll , but also protects against the inevitable loss from wiping out it all.

just an idea. Hmm how often do you get 3 on the trot right ? I suppose you could keep a record of runs of numbers of success if you often get 4,5,6, etc, then might be ok, to use that method. ?

I dunno.

 

[alexander]

In my view all you need to determine is the percent of your account that you wish to risk per trade and then position size relative to the volatility of the instrument being traded in the time frame you have chosen to trade.

Paul

Paul, this is excellent advice. Do you know of a formula to do that? Let's say the risk per trade is RT and the position size you are looking for is PS. How do you relate those to the volatility V of the time frame you are dealing with to get PS as in:

PS = function{RT, V, C) ?

probably you also need C, the trading capital value.

Alex

 

[Paul71]

Paul, this is excellent advice. Do you know of a formula to do that? Let's say the risk per trade is RT and the position size you are looking for is PS. How do you relate those to the volatility V of the time frame you are dealing with to get PS as in:

PS = function{RT, V, C) ?

probably you also need C, the trading capital value.

Alex

The subject of sizing is almost unarguable. Where the markets are concerned, money creates opportunities, the more money you have, the more opportunities you have, it's that simple.

This statement relies on all market participants being equal in knowledge.

Now do you see why the subject of sizing is almost defunct in itself.

There is no answer to it, no real answer.

Hope i don't seem too blunt, please, correct me if i am wrong.

Thanks.