Neural Networks Revisited

about NADEX? yeah I have referenced them several times before, probably with the same syntax. I don't think I've ever alluded to my near-perfect trade though.

I'm in bed half asleep still but will PM you later today; I have some Qs that I want to ask if that's OK?

What I've discovered in this >2yr quest with various statisticians is that they know absolutely dick about real world applications with regards to trading. The team I have now, I am pretty sure they get it and I'll find this out in a few days or not. So your point on the indicator & correlation was particularly appreciated on my end.
 
lucysparabola said:
about NADEX? yeah I have referenced them several times before, probably with the same syntax. I don't think I've ever alluded to my near-perfect trade though.

I'm in bed half asleep still but will PM you later today; I have some Qs that I want to ask if that's OK?

What I've discovered in this >2yr quest with various statisticians is that they know absolutely dick about real world applications with regards to trading. The team I have now, I am pretty sure they get it and I'll find this out in a few days or not. So your point on the indicator & correlation was particularly appreciated on my end.
More...

Sure, that's fine.

It all reminds me of the theme song from the old "Weeds" show:

 
@Simples A question for you. You've seen the chart. Give an estimate: How long (additionally) would this particular system have to 'work' before you would be convinced it's not possibly 'fools gold?'
 
userque said:
@Simples A question for you. You've seen the chart. Give an estimate: How long (additionally) would this particular system have to 'work' before you would be convinced it's not possibly 'fools gold?'
More...

Intriguing, however, not sure which chart you are referring to? Your PNL chart, the excel spreadsheet or something else?

When you say 100% accuracy, do you mean that the indicator should predict every direction of next bar's close, or can you skip bars?

How long depends on what method used and on what foundation. If the indicator is pure random values, then my trust in it would be minimal :)
Since I'm not sure how simple/complex your method of reading is, there's many possibilities and hard to read from just an image-file.
 
Simples said:
Intriguing, however, not sure which chart you are referring to? Your PNL chart, the excel spreadsheet or something else?

When you say 100% accuracy, do you mean that the indicator should predict every direction of next bar's close, or can you skip bars?

How long depends on what method used and on what foundation. If the indicator is pure random values, then my trust in it would be minimal :)
Since I'm not sure how simple/complex your method of reading is, there's many possibilities and hard to read from just an image-file.
More...

You can skip bars.

I was referring the the chart I posted from C2, also showing a few statistics for the system.

My question was simply about looking only at results; how long does one have to see those type of results before they would be convinced that the system is not 'fool's gold,' even without knowing anything else. 100 years? 1 year? Never-in-Life? Etc.

It's an open question to anyone.
 
userque said:
Sure, that's fine.

It all reminds me of the theme song from the old "Weeds" show:

More...

lol i just got around to listening to the song.

I should also state that my (near) 100% setup isn't (of course) exclusive to all bars. It's a fraction of a %. I get 2-3 signals per instrument on 5-minute bars per 24hr, on average. I'm having a screener built for me that will mine all of the available NADEX products and their respective expiries so that I can solve for those and hopefully find more opportunities.

my ML project, according to the team, has about 20% coverage across the data they've tested. I can still work with that as long as it proves universal (as it should) to all instruments.
 
to answer the durability Q - I am going for 100 trade sample size on my ML stuff before determining validity. It seems low, but I am obviously very familiar with how it SHOULD work, so that's a comfy # for me.
 
lucysparabola said:
lol i just got around to listening to the song.

I should also state that my (near) 100% setup isn't (of course) exclusive to all bars. It's a fraction of a %. I get 2-3 signals per instrument on 5-minute bars per 24hr, on average. I'm having a screener built for me that will mine all of the available NADEX products and their respective expiries so that I can solve for those and hopefully find more opportunities.

my ML project, according to the team, has about 20% coverage across the data they've tested. I can still work with that as long as it proves universal (as it should) to all instruments.
More...

Yeah... :) You should check out the show too if you can.

Wow...sounds like you've got some heavy stuff going on over there. Sounds expensive too.
 
yeah I will. with Billions being on hiatus until S2, I need some new entertainment.

It's less than you'd probably imagine. I also have a good-faith promise to use the system for the team that is developing it for me which helped reduce the rate.
 
lucysparabola said:
to answer the durability Q - I am going for 100 trade sample size on my ML stuff before determining validity. It seems low, but I am obviously very familiar with how it SHOULD work, so that's a comfy # for me.
More...
Right. I can understand that. But what about a system someone else built; that you know nothing/little about?

At what point is 'the proof in the pudding?'
 
You must log in or register to reply here.
Back
Top