This is sometimes called 'picking up pennies in front of a steamroller', and then there's the concern of slippage and commission/spread of course.

Have you tested if this strategy survives when incorporating these factors?

This is not a trading strategy to be honest. You need to add more features to find a better ticks. In some securities, using 75% winning rate can make money. I'm not claiming you can make money on this. I just wanted to show how can you predict and how can you use features to improve.

Lot's of people are trying to predict closing prices. That doesn't make sense.

For example, if you go long AMZN every day right at the opening and the target is Open Price x 1.0022; you will end up winning at least 75% of your trades. So you need to find out your own strategy to manage your risks. I'm using this model in my trading so often and now planning to automate it!

Good day Atilla

I see your two posts have been removed, is there any where i could find them as i had started reading them and now cant continue

kind regards

Hi. Sorry i have removed them. I just wanted to share how easily you can win but some people decided to not to believe or ignore and keeping for myself now!

Thanks

that suck, but i understand. please let me know if you decided to re-share for those of us that found it interesting

@warrick: the method is basically to calculate the probability distribution of historical (daily) ranges to set a profit target for the next day. Stationarity is assumed, of course ;).

@Ivory Ant, @Atilla,

Does Atilla use log normal distribution or pareto distribution or something else that is proprietary and unique for himself? I did not get to read the article, but was able to see The Curve in TradingView. It seems to me that it is similar in principle to BB, although Atilla says it is different. BB is a mean and standard deviation in the normal distribution sense using Closing Prices. So the difference should be on what price input he uses, log difference perhaps and how he calculates mean (i.e. Arithemic Average, Geometric Mean, Linear Trend Regression, etc.) and applying different levels of stantard deviation about the mean.

It just uses the empirical distribution of daily (or any other timeframe) ranges like screen traders have used forever. A couple of std. deviation lines are plotted and we pretend the ~67/95/99 rule applies because we assume normality.

The distribution allows us to pick our own probability of profit, and the profit that comes with it.

@Ivory Ant,

So, based on your description, it uses the conventional normal distribution where 1 standard deviation about the arithmetic average will give a 68% probability that the price (or price difference) will fall within a + and - price range. And 95% and 99% for 2 and 3 standard deviations respectively.

But how does "...distribution allows us to pick our own probability of profit, and the profit that comes with it."? Does it give directional probability? In other words does it tell you that there is a 95% chance that tomorrow's price will be up by 1% or vice versa?

You can calculate 95% probability with zscore in a normal dist. i have used same method in my today’s trade with eurusd. Check my twitter @atillayurtseven

BB is a totally different indicator. it is a great indicator but they are totally different

@James Villa: it doesn't give a direction, just the range. If you know that e.g. 75% of last year's days had a high that was $1.25 higher than the open, and 75% had a low that was $0.34 lower than the open, you can go long or short for a 75% chance of profit, based on your preference.

@Atilla,

How is BB totally different, in what sense? In your TradingView chart, I see you use mean calculation as Linear Regression with lookback of 252 . So is that the only difference between BB and The Curve? They both assume normal distribution, no? Looks like another support / resistance indicator to me.

Price is not normal distr. You need to bootstrap resample randomly!

Price is not normal distr. You need to bootstrap resample randomly!

And this is how you achieve normality and stationarity? How many resamples to make it statistically significant?

In one study of bootstrapping by Athreya KB here and to quote:

However, Athreya has shown that if one performs a naive bootstrap on the sample mean when the underlying population lacks a finite variance (for example, a power law distribution), then the bootstrap distribution will not converge to the same limit as the sample mean. As a result, confidence intervals on the basis of a Monte Carlo simulation of the bootstrap could be misleading. Athreya states that "Unless one is reasonably sure that the underlying distribution is not heavy tailed, one should hesitate to use the naive bootstrap".

Isn't it empirically studied that movement of stock prices are closely resemble more of a pareto distribution with fat tails? Just thinking out loud!

I have used more than 1000+ days to calculate (excluding last 100 days) and tested with last 100 days. It works for me! +with all security assets i have tried

Good for you! Keep us updated on your success. Perhaps one day you can share a notebook that articulates your discovery, spread the wealth, hehehe!

To be honest 1/1 so far in a real-trade! :)