Guessing Daily Direction of ES/NQ

If the 10 am numbers are greater than open, what you are implying is a hit rate of 70%? That 70% is spurious since 70% is percentage change from <b>open to close.</b> Not 10 a.m. to Close, which is basically your profit zone, ie. the whole point of betting :)

The percentage change of Open to 10a.m. place bias into the number,i.e. its incorporated into your ex poste number.

To correct this:

Let criteria be
R1 = % change from open to 10 a.m.
R2 = % change from 10 A.M. to CLOSE (not open to CLOSE)

Run regression of R2 on R1. Set residual and predicted value. If t-ratio is above 1.4 ... then you have shot.


P.S. I ran data from Dec. 1995 - Sept 2001; and the average hit rate is 52%. I do confirm that the 70% comes later on. Overall hit rate is not very STABLE. T -ratio ranging from -.6 to 2.3





Quote from luxor:

Thanks for all the feedback.

By 10 am, I meant that if the ES price at exactly 10 am is greater than the opening of that session at 4:45 pm the previous day, there's a 70% chance that the day will be up. In other words, if ES opened at 1000, and by 10:00 am the price is 1002, chances are the day will close over 1000. Of course, the price may have had a 10 point draw down that day. This is just based on a year's worth of data, so it's possible with more data the probability becomes 50%, or worse.

I haven't traded this with real money, and I haven't thought of a way to trade this effectively using stops, etc. I'm still struggling just trying to trade 1 contract and trying not to lose my shirt.

I just thought I'd mention my observation since there are successful traders here who may perhaps build on it, or shoot my theory out of the water.

lux
More...
 
Here's the % of time the SP ended higher when the price was greater than the open from 1/1/96 - 5/31/03. The % is based on the move from the time check to the close (not from the open to the close...we all wish we had a way to turn back the clock).

10:00 53.16%
10:30 55.01
11:00 54.74
11:30 55.78
12:00 54.77
12:30 56.49
13:00 56.46
13:30 56.62
14:00 55.41
14:30 58.53
15:00 57.30
15:30 55.95
16:00 57.00

From 1/1/96 - 5/31/03 the overall % of up days has been 51.58%, so the hourly numbers do provide some basis for expecting above average directional bias.
 
Out of curiosity ... how stable is it year in and year out ..

I have the feeling that the number would fluctuate by a hefty margin.

Quote from acrary:

Here's the % of time the SP ended higher when the price was greater than the open from 1/1/96 - 5/31/03. The % is based on the move from the time check to the close (not from the open to the close...we all wish we had a way to turn back the clock).

10:00 53.16%
10:30 55.01
11:00 54.74
11:30 55.78
12:00 54.77
12:30 56.49
13:00 56.46
13:30 56.62
14:00 55.41
14:30 58.53
15:00 57.30
15:30 55.95
16:00 57.00

From 1/1/96 - 5/31/03 the overall % of up days has been 51.58%, so the hourly numbers do provide some basis for expecting above average directional bias.
More...
 
Quote from Satyrican:

Out of curiosity ... how stable is it year in and year out ..

I have the feeling that the number would fluctuate by a hefty margin.
More...

No kidding! :p
 
Here's a year by year table from 1/1/96 - 5/31/03.
The yearly % is the % of days in that year the day was higher (as a control). Interesting times were from 13:00 - 14:30. All 30 min. periods performed better than the win % for that year in each of the years checked. If there is a stable edge it's probably going to be between those time periods.
 

Attachments

On the downside if the market is trading less than the open at the time here's the % of time the market trades lower till the close. Again 1/1/96 - 5/31/03.


10:00 47.85
10:30 50.11
11:00 48.67
11:30 50.27
12:00 48.98
12:30 49.00
13:00 50.95
13:30 51.58
14:00 51.48
14:30 49.51
15:00 49.35
15:30 44.59
16:00 43.81

The % of down days overall during the period was 48.42% for comparison. Looks to me like 13:30 gives a 3% edge over random while after 15:30 the market like to rebound from a selloff.
 
Quote from acrary:
10:00 53.16%
10:30 55.01
11:00 54.74
11:30 55.78
12:00 54.77
12:30 56.49
13:00 56.46
13:30 56.62
14:00 55.41
14:30 58.53
15:00 57.30
15:30 55.95
16:00 57.00
More...
Do I understand you correctly, that these numbers represent the % of the time the close was higher than price at the given time of day when the price at that time of day was higher than the open?

If so, there are a couple things to consider:

1. The differential above flipping a coin is effectively within the random "noise" range.

2. It's not as important to know that prices just closed above the price @ a given time of day in your scenario as it is to know on average how MUCH it closed above (i.e., potential profit) vs. how much it closed below (the rest of the time - which is close to half the time).

THAT would provide you a more meaningful measure of the efficacy of what you're thinking about.

For example, if prices close above the 10:30 price 55% of the time when the 10:30 price is above the opening price but the average is only 1/2 a point while the other 45% of the time the average is that prices closed below the 10:30 price by an average of 2 points - you've clearly got a setup for overall losses with a net expectancy of an average -1 1/2 points per trade.

Of course, if the reverse were true, you'd have positive expectancy for the scenario. But since you're looking at overall averages, you could also be subject to some steep drawdowns.
 
Quote from ArchAngel:


Do I understand you correctly, that these numbers represent the % of the time the close was higher than price at the given time of day when the price at that time of day was higher than the open?

If so, there are a couple things to consider:

1. The differential above flipping a coin is effectively within the random "noise" range.

2. It's not as important to know that prices just closed above the price @ a given time of day in your scenario as it is to know on average how MUCH it closed above (i.e., potential profit) vs. how much it closed below (the rest of the time - which is close to half the time).

THAT would provide you a more meaningful measure of the efficacy of what you're thinking about.

For example, if prices close above the 10:30 price 55% of the time when the 10:30 price is above the opening price but the average is only 1/2 a point while the other 45% of the time the average is that prices closed below the 10:30 price by an average of 2 points - you've clearly got a setup for overall losses with a net expectancy of an average -1 1/2 points per trade.

Of course, if the reverse were true, you'd have positive expectancy for the scenario. But since you're looking at overall averages, you could also be subject to some steep drawdowns.
More...

Yes, you understood it correctly.

For this thread I think the orignal poster was interested in whether a time of day had predictive value. The 13:30 period on the upside averaged 57.27% correct vs 51.58 of the overall market. The std. dev. was 3.73 so it was one std. dev. better than random. Non highly significant, but interesting (for me anyway).

As far as the overall expectancy, here's the results using the 13:30 period to buy or sell 1 contract for the day in the SP market. No commissions or slippage were included. Nothing exciting, but something to do on a slow trading day.
 

Attachments

You must log in or register to reply here.
Back
Top