Martingale with "low" margin hit

[Alpenkorps]

Reducing bet-size is a martingale too... Some may call it anti - martingale.

Suppose next distribution: W L L L B L L L W
(W- win, B- Break Even, L - lose)

After each win we can reduce next bet to half. This will reduce DD and increase profits. (if any in first place)

Correct me if I am wrong but anti-martingale is not reducing bet size. Its about trading towards the trend, because original martingale theory works against the trend.

[garyfritz]

Nope -- from Wikipedia: In a classic martingale betting style, gamblers will increase their bets after each loss in hopes that an eventual win will recover all previous losses. The anti-martingale approach instead increases bets after wins, while reducing them after a loss.

[Gamma_gallus]

Sorry if it has been alreadx shared here, i didn't read the entire thread .. in case it didn't, what do you think about his approach ?

http://www.forex-central.net/sure-fire- ... rategy.php

especially the 3rd one at the bottom of the page ?

Olivier

[Alpenkorps]

@ Gary, Thanks for clearing up.
@ Gamma, they will surely set fire on your account!! I have seen this thing going upto level 10!!

[garyfritz]

Fact: I've never tested a Martingale that went any significant amount of time without blowing up the account.
Opinion: Martingales work great until they don't. And then they kill you.

[Alpenkorps]

Or, a 500+ trades sample over any length of time :

There may be a few examples of this kind; but they will be one of a kind things. Even, the performers unable to repeat the feat.

Khalid

Khalid,
I can show you backtest result of 10 years or more of many Martingale strategy that will survive and grow. But yearly growth will be very low like 50-100%. You can compound but its risky. So this is the twist about martingale. It works with a calculate RR and very conservative money management. Want to know how I have blown my account with Mr. M? I did backtested properly, set a x number of max levels that price may reach and selected my lot size accordingly. In one month, my DD never went below 10% ... so I thought ... lets increase the lot size a bit ... and boom !! See? Thats how a great plan goes wrong lol

Mr. M is so tempting that its very hard to resists. Lets say your max level is 10 with a 2x martingale system, where base lot is 1. If you can sacrifice one level (which very rarely get hit), you can double up your base lot and double up your output. And just when you will do that, Mr. M will blow you up!

There is another fact too .. lets say last 10 years max level for a system was 10, who says it wont go to level 11 just when you start it? No Guarantee. And even it doesn't, is it really worth to risk 90% of your account for one trade? Nope .. not for me.

Now, after many months spending time behind Mr. M, I gave up for the hope of a complete M based system. Yearly 100% growth is not I am looking for, and I cant risk 90% of my account for one trade. What I am trying to do is combine the theory of M with price action based system. As an example, lets say you trade pinbars on 4H chart .. how many pins on a 4H chart will go wrong in a sequence? 3?4?5? hard to believe, but even so, level 5 actually not that scary.

[garyfritz]

Agreed, Khalid. The only way to be sure your backtest has some hope of resembling actual trading reality is to do a full walk-forward analysis: optimize on period 1, simulate trading with those parameters, "out-of-sample," on period 2. Optimize on period 2, trade forward on period 3, etc. (It's actually a bit different than that but that's the concept.) You are running a backtest that simulates the way you would trade it in realtime: optimize on the recent past, trade forward with those parameters. You only look at the out-of-sample results, just as you only care about the real-money out-of-sample results when you trade.

If your out-of-sample walk-forward results are fairly consistent with your in-sample optimization period, and if you'd be happy with your OOS results in realtime, you have a good chance of the system continuing to work in the future.

Unfortunately good walk-forward analyzers are very hard to come by. I'm currently testing my systems with a WFA written by a friend of mine. There is a WFA for Empty4, but given how worthless the Empty4 backtester is, I doubt the WFA is any good either.

lets say you trade pinbars on 4H chart .. how many pins on a 4H chart will go wrong in a sequence? 3?4?5? hard to believe, but even so, level 5 actually not that scary.

I would not risk any money on a Martingale unless I was VERY VERY certain of the chances of a "death trade" sequence happening. E.g. I would write code to mechanically check for and count the occurrences of the failed pin bar in history. If the worst case in, say, the last 5 years was 5 failed pins in sequence, then I would want to be sure that I could survive at least 6 or 7 sequential failures. Because the trading gods will enjoy throwing exactly that kind of "test" at you.

It's kind of like skydiving. If the primary chute only fails once in 500 jumps, you may feel very safe with a secondary chute. But what if the primary AND secondary fail once every 5000 jumps? You may be comfortable with 1-in-5000 odds, but when the consequences are that terrible...