Martingale debunked by Gary

[nanningbob]

I was shocked to see the level of hostility directed towards you in FF in the past few months. Makes me wonder who is directing this campaign. My guess is the criminals.

Bob, I was not aware of this as I was not reading your forum at FF - been a tad busy and never really 'got' 10.1 I always did put this down to my failings, not yours. I guess it is the crims at work as well; only an idiot or a crim would reject your work out of hand and try to destroy your desire to help the little guy.

Here Bob, not only do you have wall-to-wall appreciation, and a vibrant team of mods who keep threads clean of aggression and unpleasantness (even if one of them does occasionally ban himself inadvertently ) but you are a mod yourself. You can protect yourself from the cretins.

This forum attracts new members at the rate of just shy of 100 a week. If this continues, in a few years it will be the foremost Forex forum available to us little guys. FF has already discovered the roots of its own ultimate demise. This is a shame, but it is down to the likes of TIT to set the tone, and the tone they set is poor to dreadful; when the forum starts to lose members, it is their fault.

Here, we just go from strength to strength.

Thanks Steve and Astral, as you probably can see I am posting more and more here and less at FF. I dont have specific gripes or complaints against them but I am finding more trading positives here then there. Your group of coders here is impressive. I have been having a very hard time getting coders do some things for me there so time for me to move on. Anyway, you can expect me to continue me spending more time here in the future.

[Ray Singcar]

[Fx93]

Perhaps Martingale wasn't insane enough! Has anyone ever tested an exponential system: 1, 2, 4, 8, 16, etc? I did recently using the daily pivots, which are rather crude, and got a 7%/month return (pdf attached). This allows the retracement to be much smaller. The key to these is of course using very little risk, so that when the big DD comes one isn't washed out.

Where I think this may be best is actually on small timeframes, because there are always retracements at that level. I'm currently testing the 2 pip Euro range bar chart with a method, and will go back a few years to see how much DD is involved.

Edit: I should probably clarify myself. My view is that double or exponential averaging in is ok as long as one has a set TP in place, and one allocates enough risk margin in place to allow it be reached. I prefer a long period indicator for this. In case of averaged in buying, once the indicator has risen to a certain level, that is the TP, no matter if the averaging in method has worked: i.e. generated a profit or not, which limits losses. The method must have profit when this occurs more than loss of course to be viable.

[garyfritz]

Fx93, see the Martingale simulator attached to the first post in this thread. (It got omitted before but it's there now.)

1.2.4.8.16.etc progression is the default for that simulator. I think it's also the most common progression: "I lost 1, so I'll bet 2 and come out ahead one. Whoops, I lost again so I'm down 3, I'll bet 4 and come out ahead one. Whoops, I lost again..."

Play with that spreadsheet. See if you can find a Martingale progression that consistently makes money -- more, and more safely, than if you just used a larger fixed position size. I don't think you'll find one. Maybe my simulator is too simple-minded, and maybe with a more intelligent trading/TP/etc approach it might work better. Maybe your 1st S2 / 2nd S2 / etc approach actually works.

But until I see proof of that, my assumption is that the great 7%/mo returns you got just mean "you didn't hit the death trade... yet." Every Martingale I've looked at looked great -- until it blew up.

How did your approach do e.g. on USDJPY, which went almost straight up 700 pips in 6 weeks earlier this year? Or NZDJPY, which dropped 700 pips with only 1 up D1 candle?

[Fx93]

Hi Gary,

I found a period that killed my system rules I was working on and should be looked at for any low timeframe system: EUR/USD from 5/1/10-5/6/10 dropped 840 pips without any real retrace. Ouch, I bet that blew up many an account!

Edit: I am giving up on any averaging in strategy. I would say straddle methods like G+ are likely better.

Update: I designed it so it didn't blow up there, but one would need an $8k account to trade .01 lots, and the normal monthly return is paltry on this size of an account. An alternative is to only put $1k down with the knowledge that it will blow up sometime, and then withdraw one's winnings when it gets up 50-100%. To me though, the martingale average in method has proven unacceptable.

[ballen]

Hi Guys,

Read this thread with interest. Like most, I subscribe to the concept that to blindly apply Martingale-type logic to trading is a guaranteed ticket on the road to ruin. What I find harder to shake is the idea that some form of grid in conjunction with a recovery system, could produce a comfortable long-term profit. To put theories to the test, I have been forward testing several automated trading strategies. Back-testing can be fun, but anyone who came down before the last shower knows the results can show whatever you want by carefully selecting the run parameters. Hence the forward testing -- to keep myself honest!

There are two accounts showing live results at the moment for anyone who might be interested...

IC Markets -- 69 days http://www.fxequilog.com/graphs.php?acc ... &pcode=wkb

Alpari UK -- 199 days http://www.fxequilog.com/graphs.php?acc ... &pcode=wkb

The numbers themselves are not that impressive. However, these accounts have been running non-stop and completely hands-free, just to see what happens [ even over the Christmas break for the Alpari account, which is an absolute no-no for real trading ]. In my early experiments with grid trading, the killer "Margin Call" would inevitably rear its head, and this later development work is all about avoiding that outcome while providing a modest profit growth.

[Iamshakey]

Fx93, see the Martingale simulator attached to the first post in this thread. (It got omitted before but it's there now.)

1.2.4.8.16.etc progression is the default for that simulator. I think it's also the most common progression: "I lost 1, so I'll bet 2 and come out ahead one. Whoops, I lost again so I'm down 3, I'll bet 4 and come out ahead one. Whoops, I lost again..."

Play with that spreadsheet. See if you can find a Martingale progression that consistently makes money -- more, and more safely, than if you just used a larger fixed position size. I don't think you'll find one. Maybe my simulator is too simple-minded, and maybe with a more intelligent trading/TP/etc approach it might work better. Maybe your 1st S2 / 2nd S2 / etc approach actually works.

But until I see proof of that, my assumption is that the great 7%/mo returns you got just mean "you didn't hit the death trade... yet." Every Martingale I've looked at looked great -- until it blew up.

How did your approach do e.g. on USDJPY, which went almost straight up 700 pips in 6 weeks earlier this year? Or NZDJPY, which dropped 700 pips with only 1 up D1 candle?

Here is the recovery method I outlined over in "recovery with a different twist" http://www.stevehopwoodforex.com/phpBB3 ... f=23&t=193. I have been using this for a couple years now. I also use this for Blackjack, and had some luck with it with roulette (where the odds are around 45% against you.

The advantage is that the drawdown is small, and you can take more consecutive losses without it damaging your account. It also helps offset the fact that often, as a trader, I simply suck.
The disadvantage is that it is a pretty slow recovery system. This is because you use this INDEPENDENTLY with each pair, and it takes a lot of book keeping. The recovery takes more trades to get back to break-even +>1 than with the Martingale (which recovers in only one shot. A good trading system with 60% wins, (still must honor some sort of risk/reward principles), and rare long losing streaks, is a pretty good match.

Concentrate on where you are in each cycle that you are running on Each Pair. Try to ignore your account balance as much as possible. If each pair wins with each cycle, then the account balance will take care of itself.
If you are trading This Cycle on several pairs, and you get one that the drawdown is too deep for comfort, the other winning pairs can make the loss less painful. Also, try keep your trades somewhat diversified. This can also keep losing streaks down because of betting on too many USD pairs, for example.

Here it is: Don't start the system until you have a LOSS.

1 Trade ONE UNIT you lost. CYCLE STARTS AT FIRST LOSS. YOUR TOTAL IS -1
2. REPEAT ONE UNIT TRADE you lost, so your total is -2
3. REPEAT ONE UNIT TRADE you won. Now your total is now -1
4. now TRADE 2 UNITS you won, your total is +1, so cycle is done.

Now a longer one:

1. LOSS you are at -1, this begins the cycle
(repeat base bet 1)
2. LOSS you are at -2
(repeat base bet 1)
3. WIN you are at -1
(next bet base x2)
4. LOSS you are at -3
(return to base unit 1)
5. LOSS you are at -4
(repeat base bet 1)
6. WIN you are at -3
(next bet base x 3)
7. LOSS you are at -6
(next bet 1)
8. WIN you are at -5
(next bet base x 4)
9. WIN you are at -1
(now bet just enough to get BE + one base unit) base unit x2
10. Win you are at +1 and DONE.

So, here is the progression for a rediculously long term with no accounting so you can see the pattern:

0.01 Loss
0.01 Loss
0.01 Loss
0.01 Win
0.02Loss
0.01Loss
0.01Win
0.03Win
0.04 loss
0.01Loss
0.01Loss
0.01 Win
0.05Win
0.06Win
0.07Loss
0.01Win
0.08Win
keep this up until you are back to breakeven +>1
Never Compound on a Loss, Always add ONE Base Lot AFTER A WIN to the PREVIOUS ESCALATON , NO MATTER HOW MANY TRADES BACK IT WAS.

Now, An important part is, when you get back to within range of the goal (starting balance +1 unit) and you don't need the entire next increase of another base bet, just bet enough to get ONE BASE UNIT PROFIT) This can keep you from prolonging the cycle with2 consecutive big losses sometimes. For instance, if you have just risked 5 base units, and you only need 2 base units to get your goal, DON'T increase to 6, just TRADE the 2 you need.

REMEMBER, Always return to base lot with a LOSS, Always add a base lot with a win.

Not a martingale, but it works pretty well if you have patience.

Shake

[Carioca]

MARTINGALE TAKEPROFT + LOT

Sorry for the english

Went through my head right now, I do not know if someone already thought of it

The idea is to multiply the TAKEPROFT up to level 4 level 5 and 6 multiply the lot

Alert: Not sure if my math is right

L1 - SL = 100 TP = 100 // Lot - 0.1 // 50% probability hit
L2 - SL = 100 TP = 200 // Lot - 0.1 // 33,33333333333333% probability hit
L3 - SL = 100 TP = 300 // Lot - 0.1 // 25% probability hit
L4 - SL = 100 TP = 400 // Lot - 0.1 // 20% probability hit
L5 - SL = 100 TP = 250 // Lot - 0.2 // 28,57142857142857% probability hit
L6 - SL = 100 TP = 150 // Lot - 0.4 // 40% probability hit

PROBABILITY

So if we buy / sell without using any indicator this would be our probability to hit

Lose - 66,66% // about

Win - 33,33% // about

I think this says that the 6 trades open 2 trades get it right / / and we need only one hit

Conclusion: Within my math may be wrong, let's get rich

An EA that tastes great and it would be great to test this idea would be that the white box Slope in using backtesting Tick date truefx modeling 99% four pairs he gave a hit rate between 60% -80%, then the probability this idea to work would be much higher

[garyfritz]

Here is the recovery method I outlined over in "recovery with a different twist" http://www.stevehopwoodforex.com/phpBB3 ... f=23&t=193. I have been using this for a couple years now. I also use this for Blackjack, and had some luck with it with roulette (where the odds are around 45% against you.

Shake, if I understand your rules, they are:

* If you're at an equity high, trade 1.
* If the last trade was a loss, trade 1.
* If you're in drawdown, and the last trade was a win, increment "win betsize" by 1. Reduce the "win betsize" to the minimum needed to pop you into a new equity high. When you hit a new equity high, reset the "win betsize" to 1.

I've attached a spreadsheet that implements those rules with random trades. You can specify the win%, the win size vs. loss size, the % of your account to risk on each trade, etc. It simulates 1000 trades and shows the equity curves of "trade with a fixed size of N times the base risk" and "trade with recovery." (Set N in cell B6.) Please let me know if I have your rules wrong, or if you notice any errors in the spreadsheet results, but it matches your example.

First thing to notice: if you don't limit the "win betsize" to some reasonable value, it can blow up REAL fast. So you'll want to limit that to 10 or so, which you can set in cell B7.

If you do that, it won't generally blow up. However if you hit an unlucky streak, you can have a really ugly drawdown.

Meanwhile if you just trade your base size, your results will be in the same ballpark as the recovery. Sometimes recovery will be better, sometimes worse. The recovery equity curve will usually be straighter, but the profits are roughly the same. But increase your fixed betsize, to say 2x your base size, and you'll almost always equal or significantly beat the results of the recovery system. And you do it without the risk of large bet sizes or nasty drawdowns.

For example, here's a pretty typical case. At first the recovery system kept up with the basic system (with a fixed betsize of 2). Then the system went into a drawdown. The recovery took a nasty drawdown but eventually recovered. But later it got into a drawdown and never got out of it. (In this case the basic system is also still in drawdown, but it's still way ahead of the recovery. Frequently the recovery drawdowns are much worse than the basic drawdowns.)

recovery.gif

The BEST case with this recovery system is that it just about matches the 2x fixed betsize. I've seen very few cases where recovery beat the 2x betsize by a noticeable amount. The worst case is a WHOLE lot worse.

The results depend some on the characteristics of your system. You can get lucky and have long streaks that look fabulous. But you can also get into streaks that are really, really ugly -- while the basic system is just taking a modest drawdown. IMHO you're better off trading a fixed 2-unit size instead of using this recovery system.

[Iamshakey]

Here is the recovery method I outlined over in "recovery with a different twist" http://www.stevehopwoodforex.com/phpBB3 ... f=23&t=193. I have been using this for a couple years now. I also use this for Blackjack, and had some luck with it with roulette (where the odds are around 45% against you.

Shake, if I understand your rules, they are:

* If you're at an equity high, trade 1.
* If the last trade was a loss, trade 1.
* If you're in drawdown, and the last trade was a win, increment "win betsize" by 1. Reduce the "win betsize" to the minimum needed to pop you into a new equity high. When you hit a new equity high, reset the "win betsize" to 1.

I've attached a spreadsheet that implements those rules with random trades. You can specify the win%, the win size vs. loss size, the % of your account to risk on each trade, etc. It simulates 1000 trades and shows the equity curves of "trade with a fixed size of N times the base risk" and "trade with recovery." (Set N in cell B6.) Please let me know if I have your rules wrong, or if you notice any errors in the spreadsheet results, but it matches your example.

First thing to notice: if you don't limit the "win betsize" to some reasonable value, it can blow up REAL fast. So you'll want to limit that to 10 or so, which you can set in cell B7.

If you do that, it won't generally blow up. However if you hit an unlucky streak, you can have a really ugly drawdown.

Meanwhile if you just trade your base size, your results will be in the same ballpark as the recovery. Sometimes recovery will be better, sometimes worse. The recovery equity curve will usually be straighter, but the profits are roughly the same. . . . The BEST case with this recovery system is that it just about matches the 2x fixed betsize. I've seen very few cases where recovery beat the 2x betsize by a noticeable amount. The worst case is a WHOLE lot worse.

The results depend some on the characteristics of your system. You can get lucky and have long streaks that look fabulous. But you can also get into streaks that are really, really ugly -- while the basic system is just taking a modest drawdown. IMHO you're better off trading a fixed 2-unit size instead of using this recovery system.

Hey Gary, Thanks for your response.

Pretty horrendous drawdown isn't it? I read once about the BETTING system that some mathematition calculated that only 1 out of 1500 cycles would blow you out. BUT, this is with bottomless pockets and horrendous DD. This would be un-doable in Trading. This system, or any kind of martingale strat will probably never profitable in the end run for any kind of EA or Robot.
I genuinely have used this method for a couple of years or so, BUT I am a manual trader, so I intervene often . . . ending the cycle when common sense prevails, taking smaller profits when price action goes against, and sometimes closing all open trades as a basket when good profits are made, and resetting the cycle even on all the losers. Also not allowing good trades to go bad.

If i use this system to play something like blackjack (which is where I first came on to this sequence) I can always get up from the table and play another night.
I am only a modestly successful trader (at best, and probably always will be) which means that my account balance tends to creep up instead of down, but not by leaps and bounds (well, sometimes). I have always had trouble accepting reasonable DD and giving up. That's why I have to use a crutch like this to keep things me discplined.

I have often wondered how this would work with a good EA. Your Spreadsheet pretty well answers that question.

By the way, you have the sequencing exactly right, so far as I can see, and that's a great Spread Sheet. Thank you.

I think, (and I am not a strong mathematically, nor am I educated in odds statistics) logically, the reason these systems don't work with EAs is that there is no HISTORY. With Blackjack when used with card counting, you can safely say that if there is so many of specific cards have been shown, the odds tip steeply to you favor if you bet a certain way. In forex, you can't say "well, since I have seen THIS PRICE, or THAT PRICE ACTION, or THAT CHART PATTERN, the odds are now TOTALLY WITH ME."

IMHO, as an amateur, this is why martingales can't work with automated systems.

See ya,

Shake