Category | Prediction Rate | % Correct | Notes |
Direction | 12/16 | 75% | 14/16 disregarding minor closing range (losses were 0.30% and 0.65%) |
Range | 9/16 | 60% | 13/16 disregarding minor closing range |
Significant High/Low* | no instances | ||
Next Day Gap** | 2/4 | 50% | |
Gap Fill+ | 1/1 | 100% |
** The next day gap determines if the condition correctly predicted the gap.The next day gap test is measuring times the odds of a positive low or negative high were >= 14%.
+ When the odds of a negative high/pos low <= 3% and the market gaps in the low odds direction, the gap fills.
Tests:
- Since the odds of a positive low >= 14%, does the market gap up and never fill the gap? NO
- Since the odds of a positive low >= 14%, does the market gap up tomorrow? NO
Today was the ultimate failure for this prediction method, as not only did the above tests fail, but range and direction also failed. The amazing thing was how "confident" the odds were of an up day today, yet the market never even traded up all session. Losses weren't horrible, but there's no doubt today was a bad day for this method. Can I excuse it because the Cisco announcement was a "surprise"? No. Every trading model has to survive the occasional headline in order to be useful.
When the 20 day sample has concluded (will be done next week), I'll be doing a full write up on the results and present some options about how this information can be used in a trading system. I can see several ways it could be useful. This assumes, however, that it stays around 70% accurate or higher -- and I'll be paying attention to the figure in the "Notes", not the literal figure, since the days you would have lost on some of those failures were nominal. I was concerned that the results were so good because the market was in a strong trend. Now we'll get to see how it performs in a sideways or down-trending market.
The trajectory has been thrown out because it just hasn't been particularly useful. I'm going to retain the additional rows in the odds table for my own information, but I don't think the trajectory is of any use. The next day return and 2 day return predictions, however, seem to be of use, so I'd pay attention to those most of all. It's possible that this method can only look a few days out with confidence, since the view point is only a 5 day span of information. It can't speak to the trend at large.
Tomorrow's tests:
- Since the odds of a negative high >= 14%, does the market gap down tomorrow?
No comments:
Post a Comment