LOVINIDA - CL automated strategy

[jelite]

some ES trades from last week bad and good to me looks very realistic

I am not sure you are really asking the question. I wish nothing more for you than for this to be a true strategy that will make you rich. But think about it. Assuming you believe your backtest on ES for example is to be expected to continue to perform as shown above, then you should be making about 30ticks per contract per day on that instrument. That's close to 8 full points/day on ES. That is an awful lot compared to its average daily range (way above 50% of daily range).
There is just not that much alpha in the market 'available' for you. It is so far from anything that you could realistically expect that the only conclusion is that there must be an error somewhere (or it's a curve fit). Or you are running some sophisticated market making strategy with queue priority on ES, trading hundreds of times a day. All I am saying here is that a simple sanity check is quite useful when determining if it's even possible that those results could be real.

 

[vit007]

I am not sure you are really asking the question. I wish nothing more for you than for this to be a true strategy that will make you rich. But think about it. Assuming you believe your backtest on ES for example is to be expected to continue to perform as shown above, then you should be making about 30ticks per contract per day on that instrument. That's close to 8 full points/day on ES. That is an awful lot compared to its average daily range (way above 50% of daily range).
There is just not that much alpha in the market 'available' for you. It is so far from anything that you could realistically expect that the only conclusion is that there must be an error somewhere (or it's a curve fit). Or you are running some sophisticated market making strategy with queue priority on ES, trading hundreds of times a day. All I am saying here is that a simple sanity check is quite useful when determining if it's even possible that those results could be real.

In my understanding if you have a curve fit or looking foreword strategy the # of winners should be much higher compare to looser. But if you look at my strategy its almost 50/50 .

 

[vit007]

I am not sure you are really asking the question. I wish nothing more for you than for this to be a true strategy that will make you rich. But think about it. Assuming you believe your backtest on ES for example is to be expected to continue to perform as shown above, then you should be making about 30ticks per contract per day on that instrument. That's close to 8 full points/day on ES. That is an awful lot compared to its average daily range (way above 50% of daily range).
There is just not that much alpha in the market 'available' for you. It is so far from anything that you could realistically expect that the only conclusion is that there must be an error somewhere (or it's a curve fit). Or you are running some sophisticated market making strategy with queue priority on ES, trading hundreds of times a day. All I am saying here is that a simple sanity check is quite useful when determining if it's even possible that those results could be real.

I dont trade ES . I build strategy to work on CL so happens its can be used on other markets . Re. ES I did run just ruff test on ES without optimization . If I spend some time on it the results should be better.

 

[globalarbtrader]

I dont trade ES . I build strategy to work on CL so happens its can be used on other markets . Re. ES I did run just ruff test on ES without optimization . If I spend some time on it the results should be better.

I'm not sure what your motivation is in posting your strategy. You wanted advice (all of which I agree with), and you've got it. Fitting on one market isn't a great idea for a model you hope to use across multiple markets (even if you then extend it to other markets as an out of sample test, I still think its a bad idea. Do your out of sample in the time domain). Fitting on a few months or even a year of data is a bad idea.

But you don't seem to want to listen to it. I know its difficult when you've produced something that you're proud of, and you get criticised for it.

Just looking at raw performance numbers you're not going to get any more feedback than you've already had. If you were willing to give more detail about how it works then we could help you some more. But I guess you think you've found the holy grail and don't want to share.

If that is the case then I suggest you forgot about posting further, trade this thing, ideally on paper (being honest and realistic about execution, eg crossing the spread), or if that isn't possible using the smallest possible amount of money. If your sharpe ratio is as high as you think then you should see some stellar results quite quickly since the distribution of your actual returns will almost always be above zero.

So if we're all wrong then you'll be a gazillionaire by next year.

 

[mosheobrien]

I dont trade ES . I build strategy to work on CL so happens its can be used on other markets . Re. ES I did run just ruff test on ES without optimization . If I spend some time on it the results should be better.

Wake up, vit007, you were using the "optimization" feature of NT7, of course the results look so good. CL and FDAX are both extremely difficult to daytrade, to have a high Sharpe like yours is next to impossible. Can you show us some "walk forward optimiztion" using NT7?

 

[vit007]

Wake up, vit007, you were using the "optimization" feature of NT7, of course the results look so good. CL and FDAX are both extremely difficult to daytrade, to have a high Sharpe like yours is next to impossible. Can you show us some "walk forward optimiztion" using NT7?

 

[mosheobrien]

Very impressive! Let's see some real-time results! Good luck.

 

[softdown]

vit007,

Before trading your strategy with real money, you should trade in a simulated account in real time.
You can trade it for a few days and at the end of each day you compare the real time results with the back-tested results for that day.
If there is a gross error like looking in the future, you should find it quickly.
If the results match reasonably well, then you can start trading it with real money.