Nothing Works!

[none4425]

I've run backtests on multiple systems for scalping the 5 min.
On 8 months of data simulating around 1000-5000 trades.

(Warning: next part is boring. For the tl;dr go to last paragraph)

Literally nothing works. The best edge I've found is around 1%. Which, to be frank, is more than likely just random. Based on volatility, breakouts, vwap. Here is my logarithmic equity curve on over fitted parameters with a risk management strategy that is optimised for drawdown no larger than 30% from peak with 1% kelly;

Here is the same curve with 10% kelly and risk management optimised for a max drawdown of 60%;

As you can see the curves are not so smooth...
There are 600 trades in each. The expectation of a 20% edge with 1% kelly would be (1 * 1.01 ^ (0.6 * 300)) * 0.99 ^ (0.4 * 300) = 1.8

Similar to the top curve. However because the curve is not smooth, when we apply 10% kelly sizing the profitability does not match the same expectation;
(1 * 1.1 ^ (0.6 * 300)) * 0.9 ^ (0.4 * 300) = 91.15
Which is different to the result of "10" in the second graph.

TL;DR
Where is the best place to go and look for an edge in scalping? The 4 hour+ time frames seem pretty easy to beat since you can just do a trend following system and it will have a 5-20% edge but the downside being you can only make 5-15 trades a month. Even on a 20% edge, it will hardly deliver consistent results.

Fundamentally, what drives price on a lower timeframe?

 

[krugman25]

I've run backtests on multiple systems for scalping the 5 min.
On 8 months of data simulating around 1000-5000 trades.

(Warning: next part is boring. For the tl;dr go to last paragraph)

Literally nothing works. The best edge I've found is around 1%. Which, to be frank, is more than likely just random. Based on volatility, breakouts, vwap. Here is my logarithmic equity curve on over fitted parameters with a risk management strategy that is optimised for drawdown no larger than 30% from peak with 1% kelly;

https://imgur.com/a/dhTtHo7

Here is the same curve with 10% kelly and risk management optimised for a max drawdown of 60%;

https://imgur.com/a/UG0tpDq

As you can see the curves are not so smooth...
There are 600 trades in each. The expectation of a 20% edge with 1% kelly would be (1 * 1.01 ^ (0.6 * 300)) * 0.99 ^ (0.4 * 300) = 1.8

Similar to the top curve. However because the curve is not smooth, when we apply 10% kelly sizing the profitability does not match the same expectation;
(1 * 1.1 ^ (0.6 * 300)) * 0.9 ^ (0.4 * 300) = 91.15
Which is different to the result of "10" in the second graph.

TL;DR
Where is the best place to go and look for an edge in scalping? The 4 hour+ time frames seem pretty easy to beat since you can just do a trend following system and it will have a 5-20% edge but the downside being you can only make 5-15 trades a month. Even on a 20% edge, it will hardly deliver consistent results.

Fundamentally, what drives price on a lower timeframe?

What you are doing doesn't work. Study how price/behavior works in the market. Study thousands, tens of thousands of charts which will take a number of years to do. You will start to see rhyming patterns/behaviors that you can assign probabilistic outcomes to which you can then trade. Most people never figure it out. It's the great revolving door of trading. Good luck.

 

[dozu888]

check my thread 'trading is easy'.... logical stuff there why something works.

'scalping' is a myth.. a cluster f of robots trying to screw each other.

 

[bone]

Where is the best place to go and look for an edge in scalping?

You’ve answered your own question quite frankly. Yes - there are successful scalpers. IMO it’s a more difficult path than swing trading for mere mortals. Made worse by retail cost structures.

If you can learn to swing trade consistently it also opens up a world of possibilities for leverage and asset management.

Just my own two cents - YMMV. I wish everyone good fortune!

 

[Scataphagos]

I don't know about the "tens of thousands" of charts... but you are definitely on the right track.

My favorite quote of all time is not from the markets but fictional. When Sherlock Holmes was asked, "just what is it that you do, Mr. Holmes?". He replied..."I observe and deduce". Recommend that for all traders.

 

[speedo]

What matters is not what "drives price" but the nature of price discovery itself. The only way I know to do this which is the way I learned is to spend many hours simply watching real time price development. In due time, patterns and tendencies will begin to appear. THEN you have something to test.

Experiment with different time frames, including transaction based charts, and observe price across different time frames. Play with various derivatives to see if they help you see what's happening. It took most of us years to develop into consistent traders. If there is a short cut, I'm not aware of it and I have been around awhile. If you have the will to do the work, both technically and personally you can succeed but expect it to be a more difficult task than anticipated.

 

[none4425]

What you are doing doesn't work. Study how price/behavior works in the market. Study thousands, tens of thousands of charts which will take a number of years to do. You will start to see rhyming patterns/behaviors that you can assign probabilistic outcomes to which you can then trade. Most people never figure it out. It's the great revolving door of trading. Good luck.

Looking for patterns seems like something that can be easily done with machine learning... But of course that doesnt work, and the way stocks move aren't random or statistical. They move based on people making real trades, buying and selling at specific levels to points of liquidity in the market. On a higher time frame, they tend to move in a reflexive way (based on George Soros' theory) which explains maximum volatile crashes preceded by parabolic rises - This is why trend trading works. (forgive the way i write, i know it sounds arrogant and dismissive but im just trying to get to the point).

Clearly, on the lower timeframes, reflexivity does not play a key role... However I have still found that trend following, buying/selling breakouts on the 5 min is the only thing that has given an edge (even if it is tiny) - obviously another force is at play though. why does the 5 min chart behave in a near completely different way to the daily chart. Not asking for a holy grail but maybe a hint in the right direction?

 

[krugman25]

Looking for patterns seems like something that can be easily done with machine learning... But of course that doesnt work, and the way stocks move aren't random or statistical. They move based on people making real trades, buying and selling at specific levels to points of liquidity in the market. On a higher time frame, they tend to move in a reflexive way (based on George Soros' theory) which explains maximum volatile crashes preceded by parabolic rises - This is why trend trading works. (forgive the way i write, i know it sounds arrogant and dismissive but im just trying to get to the point).

Clearly, on the lower timeframes, reflexivity does not play a key role... However I have still found that trend following, buying/selling breakouts on the 5 min is the only thing that has given an edge (even if it is tiny) - obviously another force is at play though. why does the 5 min chart behave in a near completely different way to the daily chart. Not asking for a holy grail but maybe a hint in the right direction?

I'm an experienced software engineer so if there was a way to write an algorithm to do it I would have by now. It's a lot less about finding an a-b-c-1-2-3 pattern and more about storing thousands of historical charting moments in your memory bank and then looking at a live chart and getting that gut feeling. The gut feeling being what you are viewing live on the charts and making connections with all of that knowledge and experience you have built up to that point. There are of course patterns or rhymes to highly probabilistic setups, but way more complex than a simple rule based pattern.

This is of course only part of the equation. Combine that with solid trade management and capital preservation rules and you may have something.

 

[ET180]

It took me longer than I would like to admit to realize that if there was an easy, straight-forward algorithm for extracting money from the market, then everyone would already be doing it.

I think the only edge is in finding something that is not correct all the time, but rather finding trade setups that work sometimes, but where the potential reward is much greater than the amount of capital lost if the trade does not work out. If I have a system that on average returns 5x the amount of capital risked, but only works 33% of the time, that's not a bad system.

 

[murray t turtle]

I've run backtests on multiple systems for scalping the 5 min.
On 8 months of data simulating around 1000-5000 trades.

(Warning: next part is boring. For the tl;dr go to last paragraph)

Literally nothing works. The best edge I've found is around 1%. Which, to be frank, is more than likely just random. Based on volatility, breakouts, vwap. Here is my logarithmic equity curve on over fitted parameters with a risk management strategy that is optimised for drawdown no larger than 30% from peak with 1% kelly;

Here is the same curve with 10% kelly and risk management optimised for a max drawdown of 60%;

As you can see the curves are not so smooth...
There are 600 trades in each. The expectation of a 20% edge with 1% kelly would be (1 * 1.01 ^ (0.6 * 300)) * 0.99 ^ (0.4 * 300) = 1.8

Similar to the top curve. However because the curve is not smooth, when we apply 10% kelly sizing the profitability does not match the same expectation;
(1 * 1.1 ^ (0.6 * 300)) * 0.9 ^ (0.4 * 300) = 91.15
Which is different to the result of "10" in the second graph.

TL;DR
Where is the best place to go and look for an edge in scalping? The 4 hour+ time frames seem pretty easy to beat since you can just do a trend following system and it will have a 5-20% edge but the downside being you can only make 5-15 trades a month. Even on a 20% edge, it will hardly deliver consistent results.

Fundamentally, what drives price on a lower timeframe?

%%
What dives price ?? Well Don Bright Daytrading Co did not like 5 minute candles =too long a time period LOL/true. Profitable Swing / position Traders/Investors like 44 days/44 weeks......... Wisdom is profitable to direct.,