A real edge hardly requires any testing

[smallfil]

Been trading profitably for the last few months, after putting in more than 5 years. The main thing I have noticed is that its all about finding the edge, and the edge is found by applying sound reasoning and logic, in the same way as is done to solve a very interesting but a very tough puzzle.

Testing has no part to play in it. Once you find the edge, or find your path through the puzzle, its almost always a proof in itself.

Its only when you don't have a clear edge, that you start testing with loads and loads of data, and I guess not much comes out of it..!!

You could get lucky and think you are a trading genius when you have just been lucky in your trades. I found what I thought was a trading system that was very good and had an edge. On paper (with a limited data of about 205 trades) seemed that would be close to a sure thing. Now, comes the actual test with real monies on the line and it fails terribly. Did not count on the huge slippage involved since, I trade options as a buyer. I took the lumps on the losing trades as I had proper risk management. Anybody can get lucky trading the stockmarket. You probably, need atleast, 500 actual real life trades to validate you even have something with an edge.

 

[notagain]

Some say tick charts and big candles are edge enough. The real edge is logic shouldn't be clouded by taking on too much risk.

 

[Doobs789]

Edge must be quantifiable, exploitable, and scalable.

How much edge exists? Can you capture that edge? Does that edge persist when increasing bet size?

 

[d08]

Been trading profitably for the last few months, after putting in more than 5 years.

The classic thread of illusory epiphanies™ after a few profitable months. We've all been there but your current state of clarity won't last.

 

[newwurldmn]

Been trading profitably for the last few months, after putting in more than 5 years. The main thing I have noticed is that its all about finding the edge, and the edge is found by applying sound reasoning and logic, in the same way as is done to solve a very interesting but a very tough puzzle.

Testing has no part to play in it. Once you find the edge, or find your path through the puzzle, its almost always a proof in itself.

Its only when you don't have a clear edge, that you start testing with loads and loads of data, and I guess not much comes out of it..!!

This post reeks of nativity. Over course you have been only trading for five months. Come back and read this post after you’ve been trading for 20 years.

 

[ma_trader]

What a load of crap. I hope your success holds.

If you don't understand something, it becomes a 'load of crap'..!! Thats great approach..!!

 

[tommcginnis]

maybe you can aggregate multiple smaller edges into something reasonable

^^THIS.^^

Don't be a purist. Be responsive to available profit. (Is this the 37th Rule Of Profit?)

 

[MoreLeverage]

Maybe someone could explain an outcome set (sans cheating or having inside info, latency edge etc) where the "outcomes are clear enough that there's no doubt."

To be fair, I said “not much doubt”, but let’s say you thought you had a high chance of -2% and a low chance of +2% depending on what happens across earnings or similar. On one hand, it’ll be hard to get enough data to be sure the odds really are say 4:1 in your favor or whatever, but on the other hand, if you keep putting on the bearish trades and more than 2 go against you before you’re making money, you know something’s probably wrong with your model.

 

[minmike]

If you don't understand something, it becomes a 'load of crap'..!! Thats great approach..!!

Right. I have been doing this for my sole source of income for over a decade. It is definitely my lack of understanding that is the problem.

 

[ma_trader]

A few profitable months doesn't mean much. Because even a negative edge systems can make money for 18+ months.

You could have a real edge. Just a few months is too short to tell. Thats why you need atleast a 3 year track record. And even that is really too short.

Thats the thought, which is not necessarily right, I feel. Let me rephrase myself, if and when you find a real edge, you will know it, thats what I meant.

If the edge is not good enough, you will have to go for testing it out.

But,

A few profitable months doesn't mean much. Because even a negative edge systems can make money for 18+ months.

You could have a real edge. Just a few months is too short to tell. Thats why you need atleast a 3 year track record. And even that is really too short.

An edge should stand by itself, testing on data sets give you an edge on those data sets only.