Hi,
My dad is a Forex newbie and I know nothing about Forex at all. A few days ago my father came up with an algorithm that he has been testing using play-money. Although all his other ideas failed, this algorithm has yielded 38 consecutive wins for him with 0 losses. Could this just be a coincidence, or does he really have something here? He is reluctant to believe that a newbie like him could create such an effective money-machine so quickly.

That would most surely be statistically significant. As long as his risk isn’t highly overshadowed by his reward, it could be a fairly significant algorithm.

I’d be curious to know what the drawdown is on his system (the max loss on any one trade before it turns out to be one of his winners). Also, what his risk/reward expressed as a fraction?

I once generated a neural net program to predict moves in the S&P 500 based on the price of gold. I was amazed with the good results. Turned out, it was using the London price, hours ahead of the NY market. It wouldn’t work once the time difference was accounted for. This is an example of postdiction.

If you take 1000 people and split them into two groups of 500 and have them flip coins against each other, one-on-one, by the time you get down to the last two flippers, they will have won hundreds of tosses in a row. Are they really good coin flippers?

I can design a Martingale system that will win one hundred trades in a row, as long as I’m willing to lose even more when a single loss comes around. The math of that system can’t be beat. People have tried for hundreds of years. Casinos love martingale players. In the end, the casino always takes all their money.

There are lots of possibilities like these. In my own humble opinion, it’s best to find an expert to look it over. Don’t risk any real money until it’s carefully checked, and even then, make it a very very small amount.

I am fairly sure its nothing like that… this is happening right now in real time with play money in metatrader 4.

If you take 1000 people and split them into two groups of 500 and have them flip coins against each other, one-on-one, by the time you get down to the last two flippers, they will have won hundreds of tosses in a row.

Actually, by the time you get down to the last two flippers, they will both have won only 8 times. To get 2 people who won 100 tosses in a row you need 1 nonillion people. Thats 10 to the power of 30.

I can design a Martingale system that will win one hundred trades in a row, as long as I’m willing to lose even more when a single loss comes around. The math of that system can’t be beat. People have tried for hundreds of years. Casinos love martingale players. In the end, the casino always takes all their money.

There are lots of possibilities like these. In my own humble opinion, it’s best to find an expert to look it over. Don’t risk any real money until it’s carefully checked, and even then, make it a very very small amount.

Yeah - well thing is there hasn’t been a loss yet. Thats really what makes this so hard to analyze. And there is no sales pitch because if this is as good as it seems then he is keeping this to himself ;).

Can anyone give me an example of one of these algorithms that can win 99% of the time but lose big when it loses?

It need not even be an algorithm. If you risk 1000 pips for a potential return of 20 pips, that’s 50:1 risk/reward. You could flip a coin and enter based on just that result. You could win 50 times out of 51, but that 1 loss will wipe out whatever profit you made the 1st 50 rolls.

If your results are genuine WITH good positive expectancy and sample size is large enough to provide statistical significance, then you might have a working system.

-Edit-

In other words, winning 38, 40, 50, 100 times in a row is meaningless without context. What is that context? Stop loss.

I dont think 38 trades is enough to evaluate any algorithm there could be a gaping hole in it somewhere that would not have shown itself in only 38 trades, to test it properly and get a realistic sample you should run it through several hundred trades

On a serious note, when I first started this, the demo account I was using, had all the attributes, and bells and whistles of the real deal.

It did, with one minor exception.

The “market buy”, or “market sell”, on the demo account, was based off of the “last”.

So, even though the ask was 3 pips higher than the bid, if someone on a live account sold, the bid number would light up as the “last”. If I hit buy then, I would buy at the bid, and be 3 pips up whenever someone sold at the ask price.

I made about 75,000 of funny money in about three days.

Of course I figured out that would never work, and then went back to the drawing board, but make sure it’s not some odd anomaly with the platform you are using.