@Monobrow

Will do.

@GraHal

Yes I do (timeframe). It was the first and only backtest I’ve made. I was surprised though, because with stops and targets like 1.61 and 3.01, it certainly looks like it has been optimized.

Downloaded their latest free package, backtested 5 different algos and they all look bad. Their results on the home page dont fit the real thing

Optimization is not synonymous with over-optimization.
Concerning backtests, it is important to compare them with the same data: start and end date, spreads used, exact same instrument?
Aren’t the results presented by this company based on a portfolio of strategies rather than on individual backtests?
This applies to all strategy codes that can be downloaded here for free or for a fee on the internet. The assessment of a backtest according to one’s own analysis and one’s own view of what is good or bad is rather subjective.
I am well aware that you want to recoup your costs, but a strategy that does not ruin your account is already, in my opinion, not so bad, here as elsewhere, nobody will be able to tell you 100% that the future will be as pleasant as the past 😉

This being said, if you can test in demo for a good period of time, everyone will be able to make their own opinion and take note or not of whether these systems are worth it or not.

Nicolas wrote:

Optimization is not synonymous with over-optimization.

Concerning backtests, it is important to compare them with the same data: start and end date, spreads used, exact same instrument?

Aren’t the results presented by this company based on a portfolio of strategies rather than on individual backtests?

This applies to all strategy codes that can be downloaded here for free or for a fee on the internet. The assessment of a backtest according to one’s own analysis and one’s own view of what is good or bad is rather subjective.

I am well aware that you want to recoup your costs, but a strategy that does not ruin your account is already, in my opinion, not so bad, here as elsewhere, nobody will be able to tell you 100% that the future will be as pleasant as the past 😉

This being said, if you can test in demo for a good period of time, everyone will be able to make their own opinion and take note or not of whether these systems are worth it or not.

– Very true about optimization. But when I see moving averages like 63 and 89, or in this case a stop of 1.61%, then I get a bit suspicious regarding over-optimization. But that is most likely subjective as well.

– Not sure what you mean with the same start and end date. If I download a strategy from here or anywhere else, I would test it on my available 100k of data right? Because I can’t have anymore historical data that what I have available. But of course, correct spread and instrument is a must.

– “[] but a strategy that does not ruin your account is already, in my opinion, not so bad[]”. Totally agree about the future and the past. But I must say I expect more from a strategy than you described there. I’m a rookie in this though, only have 4 strategies on real money. But all 4 are winning and making good profits (one of them are very slow though). For me, a “not so bad” strategy is one that is break even (or better). A strategy that is losing in the long run is not a good strategy in my book, and should be replaced in my portfolio. But again, in a couple of years, I might think the same! Maybe it’s my new-beginners-luck that is talking here. 🙂

– I think that is a part of the problem. 3 weeks of testing (the trial lasts that long) is really not that long test period. For me it would be way too early to run on real money with that small testperiod.

Here’s a screenshot just taken. I started mine later than you, but why hasn’t your Japan taken a trade? Have I done something wrong there? I have my setting on UTC+1 timezone.

Edit: Thanks for replying.

Mattias wrote:

Here’s a screenshot just taken. I started mine later than you, but why hasn’t your Japan taken a trade? Have I done something wrong there? I have my setting on UTC+1 timezone.

On this specific account we run only 19 out of 20 algos due to the fact that the leverage of the Nikkei indice available is too high.

But yes, the PRA+ Nikkei M5 algo is in a long latent position. Entered at January 19th at 12:20(UTC+01)

Mattias wrote:

– Very true about optimization. But when I see moving averages like 63 and 89, or in this case a stop of 1.61%, then I get a bit suspicious regarding over-optimization. But that is most likely subjective as well.

Indeed, but MA are just “lines in the sand”, I think these periods are obviously optimized, but surely validated with a robustness test in OOS period.

Mattias wrote:

– Not sure what you mean with the same start and end date. If I download a strategy from here or anywhere else, I would test it on my available 100k of data right? Because I can’t have anymore historical data that what I have available. But of course, correct spread and instrument is a must.

No, because if a strategy is optimized during multiple IS periods, they are not validated in the past OOS ones but in the future ones. Since you can’t know how the strategy had been optimized (how many IS+OOS, since when, for how many times, etc..), you will never be able to replicate how the strategy would have behaved in the past and surely not with the current optimized settings!

I strongly recommend to read more about robustness testing and the walk forward analysis tool.

I’m only replying here since it seems to be an “hot topic” currently, and comments made are in strong relation with what we are trying here to educate people with (robustness testing), please don’t take it as a recommendation to use the systems you are talking about here, or not 😉