Time Segment Bias
Looking for a bias
Objective of Study: To discover if there is a bias to the market moving in a particular direction due to the time of day.
Data used
One minute OHLC bars for the E-mini S&P 500, E-mini NASDAQ-100, E-mini Russell 2000 and mini-sized Dow ($5) Futures. Symbols: ES, NQ, ER2, YM.
Dates/time used
All days during 2004 but only data from between 09:30 and 16:15 EST.
Results for ES 2004
E-mini S&P 500 (2004)
| Time | %Ups | %Downs | %Flats |
|---|---|---|---|
| 9:30 | 50.8% | 45.3% | 3.9% |
| 10:00 | 45.7% | 50.8% | 3.5% |
| 10:30 | 44.6% | 48.4% | 7.0% |
| 11:00 | 47.3% | 44.6% | 8.1% |
| 11:30 | 44.0% | 48.4% | 7.5% |
| 12:00 | 52.0% | 34.9% | 13.1% |
| 12:30 | 46.0% | 44.4% | 9.5% |
| 13:00 | 47.6% | 45.6% | 6.7% |
| 13:30 | 49.0% | 43.0% | 8.0% |
| 14:00 | 49.8% | 46.6% | 3.6% |
| 14:30 | 45.8% | 48.2% | 6.0% |
| 15:00 | 49.8% | 43.0% | 7.2% |
| 15:30 | 49.4% | 44.2% | 6.4% |
| 16:00 | 45.0% | 46.2% | 8.8% |
Results for NQ 2004
E-mini NASDAQ-100 (2004)
| Time | %Ups | %Downs | %Flats |
|---|---|---|---|
| 9:30 | 50.4% | 45.0% | 4.7% |
| 10:00 | 47.3% | 48.1% | 4.7% |
| 10:30 | 47.7% | 46.1% | 6.2% |
| 11:00 | 48.1% | 47.3% | 4.7% |
| 11:30 | 47.2% | 44.0% | 8.7% |
| 12:00 | 50.8% | 43.3% | 6.0% |
| 12:30 | 47.2% | 44.8% | 7.9% |
| 13:00 | 47.6% | 44.0% | 8.3% |
| 13:30 | 47.4% | 45.8% | 6.8% |
| 14:00 | 48.2% | 47.4% | 4.4% |
| 14:30 | 47.0% | 45.4% | 7.6% |
| 15:00 | 50.2% | 45.8% | 4.0% |
| 15:30 | 45.8% | 47.0% | 7.2% |
| 16:00 | 48.6% | 43.0% | 8.4% |
Results for ER2 2004
E-mini Russell 2000 (2004)
| Time | %Ups | %Downs | %Flats |
|---|---|---|---|
| 9:30 | 52.5% | 45.5% | 1.9% |
| 10:00 | 46.1% | 51.6% | 2.3% |
| 10:30 | 47.7% | 50.0% | 2.3% |
| 11:00 | 48.2% | 49.4% | 2.4% |
| 11:30 | 46.4% | 48.8% | 4.8% |
| 12:00 | 53.6% | 40.9% | 5.6% |
| 12:30 | 48.0% | 47.6% | 4.4% |
| 13:00 | 52.4% | 43.7% | 4.0% |
| 13:30 | 49.4% | 46.6% | 4.0% |
| 14:00 | 50.6% | 46.6% | 2.8% |
| 14:30 | 45.8% | 50.2% | 4.0% |
| 15:00 | 47.8% | 47.8% | 4.4% |
| 15:30 | 45.8% | 51.4% | 2.8% |
| 16:00 | 49.8% | 43.8% | 6.4% |
Results for YM 2004
mini-sized Dow ($5) Futures (2004)
| Time | %Ups | %Downs | %Flats |
|---|---|---|---|
| 9:30 | 51.7% | 46.7% | 1.7% |
| 10:00 | 45.5% | 52.1% | 2.5% |
| 10:30 | 47.1% | 48.8% | 4.1% |
| 11:00 | 47.9% | 47.9% | 4.1% |
| 11:30 | 47.5% | 49.6% | 2.9% |
| 12:00 | 51.7% | 45.0% | 3.3% |
| 12:30 | 45.5% | 49.2% | 5.4% |
| 13:00 | 46.7% | 49.2% | 4.1% |
| 13:30 | 49.2% | 48.3% | 2.5% |
| 14:00 | 49.8% | 46.9% | 3.3% |
| 14:30 | 44.4% | 50.6% | 5.0% |
| 15:00 | 51.5% | 44.4% | 4.1% |
| 15:30 | 53.9% | 43.6% | 2.5% |
| 16:00 | 47.3% | 48.1% | 4.6% |
Interpretation and Application
Interpretation and Application
The tables above show us the percentage of times that the market moved up/down/sideways from the opening minute of the relevant time segment to the closing minute at the end of the time segment. All the segments are 30 minutes in length except the last one which is 15 minutes in length.
The tables do not show us how much the moves were but just if the move was positive, negative or zero.
The objective of this study was to ascertain a bias in the market due to the time of day. Have you ever heard something like: "the market almost always rallies at 11am" or other such axioms of market behavior?
The only anomaly that appears in any of the tables (in my opinion) is the 12:00 to 12:30 EST segment in the E-mini S&P 500. This shows an unusually high number of flat closings and low number of down closings.
In order to see if this anomaly extended further back in time I re-ran the test using 6.5 years of E-mini S&P 500 data instead of just the 1 year that goes to make up the tables above and came up with the following table:
Results for ES 1996 to 2004
E-mini S&P 500 (1996 to 2004)
| Time | %Ups | %Downs | %Flats |
|---|---|---|---|
| 9:30 | 49.0% | 48.0% | 3.0% |
| 10:00 | 47.3% | 49.8% | 2.9% |
| 10:30 | 48.8% | 47.2% | 3.9% |
| 11:00 | 47.5% | 47.1% | 5.4% |
| 11:30 | 47.0% | 48.4% | 4.6% |
| 12:00 | 49.7% | 44.1% | 6.2% |
| 12:30 | 47.5% | 46.9% | 5.6% |
| 13:00 | 50.6% | 44.9% | 4.6% |
| 13:30 | 47.7% | 48.2% | 4.1% |
| 14:00 | 46.4% | 49.6% | 4.0% |
| 14:30 | 46.7% | 49.6% | 3.7% |
| 15:00 | 51.5% | 45.3% | 3.2% |
| 15:30 | 51.6% | 45.5% | 2.9% |
| 16:00 | 53.1% | 42.3% | 4.6% |
This table with more data shows that the anomaly disappears when the test is run over a longer period of time. What this demonstrates is that anomalies often show up when insufficient data is used and caution should be taken in these instances.
The running of a test or study over a large number of periods is often referred to as a Monte Carlo simulation and if sufficient data is used then we can be more certain of what will happen over a similar number of periods in the future.
How can we use this to make money?
This study demonstrates that there is no advantage to trading any of these futures based on time of day. We can make money with this information by ignoring suggestions that a time of day has a particular up or down bias and rather focusing on strategies that give us a true edge.
I have seen strategies that combine time of day with other factors but they have not been studied or reported on here.
Copyright © 2004-2023, MyPivots. All rights reserved.