This certainly does appear to be true. Over the last three years, I have tested hundreds of robots, both of my own design and others, both in back tests and on demo accounts using live trading data, and have yet to place one for any length of time on a live trading account.
Some robots perform well for a period of time, and then fail miserably. Others have spectacular back test results and then perform completely horribly in forward tests or live accounts.
Currently I have several robots which have shown promising results over the last half a year, but I still do not trust them enough to run live.
So, it's really true then? No robots work?
In my experience, certainly none perform as hoped - that is generating huge amounts of income daily while the operator sits passively by or goes about his or her day leaving the robot to it. At best they are clumsy and crude, and you're a lot better off learning to trade manually. If you use 4 hour or higher time frame charts, it doesn't even take very long.
But here are a few examples of robots which "work" as long as your expectations aren't too high.
3 Pair Hedge Robot
I would feel comfortable running this robot on a live account - however it doesn't generate much income. Orders can be open for weeks, sometimes months, and it's designed only to return $20 profit from a $5,000 account. Even with a $50,000 account, that means $200 perhaps every week. However, it is reliable, and doesn't go into too much draw down.
It's best return was about $80 in one week. This happened only once during it's first 3 month test period. It's currently on it's 2nd test period, and since 2012-09-17 has returned only $62. 1.24% in 3 weeks.
It's designed to create a synthetic currency pair, which due to market fluctuations will eventually move the account into profit and it will then close the orders. Maximum observed draw-down was about $150.
This particular design hedges EURUSD, AUDUSD and EURAUD. I have guessed which lot sizes to use for the pairs. I tried another robot which hedged GBPCHF, GBPJPY and CHFJPY but draw down was slightly worse and it ran for over a month and never returned a profit. I think it was close, but I gave up on the design before it could. It's possible I got the lot sizes wrong, which made profit more difficult to come by.
So this is an example of a working robot. You can download and use it yourself. Just run the robot on a EURUSD chart with an account containing at least $5,000. In a few weeks you will be $20 richer.
This robot trades consolidations. True to form, in a consolidating market, of which recently we've had a lot, it works well. However eventually a trending market will come along and it gets killed.
I've added the ADX indicator to try to filter strong trends. The stop loss and other settings aren't particularly optimized. Usually once it entered an order I'd move the stop loss manually anyway.
I don't know how long I've run this robot for, but it's been a long time. Possibly 9 months or so. It's rarely ever exceeded $100 profit. Even now with the trend filter in place, $100 seems to be the best it can do.
Like many of the other robots I'm currently demo testing, it returns great results in back tests. I suspect every so often it catches a great trend or has orders open over the weekend and the market gaps heavily in it's favour and this may explain the great back test results. Whatever the reason, it's been running so long and has been so consistent I would be happy to run this on a live account. But I wouldn't expect much profit from it.
Additional manual intervention may possibly make this robot more successful. It's easy enough even without the ADX filter to see when the market is trending and when it's not, and it is good at entering trades. Just it's follow through needs work.
High-Low 1D
Here's the last example of a robot that "works". This is a shockingly bad example, so the code is not provided due to embarrassment. But nevertheless, we're showing robots that "work" even if their success is very mediocre.
This is a hedging robot. I have long forgotten how it even works. It's been running for so long and has had orders open all this time. Many only just closed in the last few days. I was beginning to think they never would.
So, from all the time it's been running ... weeks? Months? I really don't know. I long ago stopped paying attention to this robot ... but anyway, it's made $36.
Look at the AUDUSD chart in the middle of the screen shots. See the mass of orders which just closed? This is a silly, silly design. Yet, it works.
