Thanks for that clarification. I'm still getting my head wrapped around Hurst but the descriptions you quoted from the article do seem quite familiar to how I interpret autocorrelation. The relationship is probably not directly map-able, just in ...

Just a bit of advice, for anyone, for free

Interesting discussion guys! This article describes a simple way to interpret Hurst. Where 0.5 is Brownian motion or a random walk and below indicates anti-persistence and above indicates persistence. It seems to me that a number very close to zero ...

Just a bit of advice, for anyone, for free

You profit from a retrace while your net position is zero? That's a neat trick. How exactly do you profit from price movement while your net exposure is zero?

Hedge a losing position?

True, the ratio is sound and the idea of scaling Sharpe's ratio by trades is in interesting innovation. Perhaps the scaling was inspired by Kelly's expected long term growth rato, which plots a single point on Markowitz's efficient frontier and is ...

High hit rate vs big R:R - Which one do you prefer and why?

I was playing around with SQN recently and after a while realized that if you assume your risk risk per trade is simply your average loss (and I don't see why you wouldn't do this in spite of what Tharp says) then the equation comes down to Sharpe ...

High hit rate vs big R:R - Which one do you prefer and why?

Yes I make no attempt to fit a given R:R or win rate for my strategies. Of course higher is better for both metrics, but I tend to measure other things rather than focus on R:R which tends to simply be a reflection of the underlying market dynamics ...

High hit rate vs big R:R - Which one do you prefer and why?

If you maximize return to drawdown ratio and use it as a selection method, you are very likely to have overfit the data. Why? Maximum drawdown consists of a single trade series that led to the largest adverse move. It is a sample size of 1. Small ...

High hit rate vs big R:R - Which one do you prefer and why?

That would be great, wouldn't it? But I think I've taken this idea as far as I care to. It was more to prove a point regarding the relative simplicity of engineering a system with a certain feature, such as a high win rate. But doing it in this way ...

High hit rate vs big R:R - Which one do you prefer and why?

For zero expectancy: randomly buy or sell at fixed time of day, stop 70 pips * multiple, target 30 pips * multiple, target win % is 70% but will vary more when trade count is low. (Where stop/target multiple is m ) For positive expectancy, buy / ...

High hit rate vs big R:R - Which one do you prefer and why?

Yes I agree that the adaptive nature of the market participants and thus the aggregate market causes the destruction of all systems at some times, but not all systems all the time. While some systems go down the tubes performance wise, other systems ...

What works, and what doesn't

I think the market's "character" is usually only noticed through the lens of our system/trading performance. When a system stops performing, has the market's character changed? Or has the system simply stopped working? What about other systems that ...

What works, and what doesn't

I've been meaning to read the Clenow book and hadn't heard of the Piard book so thanks for the references.

What works, and what doesn't

Can you explain why it isn't possible to evaluate the robustness of a trading strategy when a strategy changes its rules over time and how the machine learning concept that I alluded to is flawed?

What works, and what doesn't

Hi BigOldFella, Aside from the awkwardness of my sentence that you quoted , I think the idea I was trying to get across can be illustrated by this example: Let's say you wish to optimize a MA crossover system with two MAs with lengths from ...

What works, and what doesn't

As recently as 5 years ago I probably would have taken exception with this statement. I didn't realize it at the time but after creating and testing thousands of trading strategies and evaluating them for profitability and consistency with high ...

What works, and what doesn't

Good points katrooo. I was going to post something on this but decided not to bother, then I saw your post. HFT is more and more moving bid ask prices on micro timeframes without any order flow. This can be seen by studying the book on a futures ...

A jump to the beginning - chaos and market forces

I hear you nonlinear, I've got 8 machines cranking away right now and it just isn't enough processing power. Dealing with tick data really limits the speed at which this "factory" can produce trading strategies. They're still amping up - maybe a ...

Machine Learning with algoTraderJo

I used to hold the exact same view. Then I tested and found that my strategies with high trade counts on shorter term data didn't hold up very well on new data. So I went back and took a lot of time and did testing on longer periods. I found that ...

Machine Learning with algoTraderJo

Ok I'll try to make this simple, brief and explicit. We can compute the range of a bar (High - Low). We can also take the average of this range over the last X bars. If we do a bit more complex calculation we end up with ATR. If we forget about ...

Does taking opposite trades work?

I agree. But the ones who are really laughing are running server side MT4. As to the points made by others - you're right as traders we won't probably ever have access to the same kind of info (orders) that a broker / dealer has or the benefits / ...

Does taking opposite trades work?