Oct 19th, 1987 |
||||||
How does the system behaves in a crack? |
Our system tries to capture the
essential dynamics of the free evolution of the
market indices only in terms of market direction.
This kind of prediction is not very well suited
in times of great volatility because it cannot
distinguish between a great fall or a moderate
one. Anyway we have observed that when a crash
takes place, usually it begins in a stressed
moment when our system can effectively describe a
downturn. After this trend is fully developed, if
the fall was exaggerated, guarantee reposition
often induces agents to continue selling
generating the crash. These conditions make up an
scenario that is rare, and our method fails to
describe. In the following pictures is shown how our systems behaved the days around the fall of October 19th, 1987. The fist picture below shows in blue the DJIA daily closes and superimposed, in blue, the ups and downs of the same days. |
|||||
 |
||||||
| We will try to predict the
evolution of the ups and downs. In the following
pictures in red is shown the DJIA up/down time
series, in black the prediction. Other lines are
predictions with non optimal parameter set and
are depicted in order to show sensitivity to
parameter shift. All predictions are computed
after the close of the first day showed on the X
axis of each picture.
Beginning with Oct 05th we can appreciate that the predictor captures the free evolution well up to the 5th day showing the right market direction of 8 trading days.
Oct 06th: the prediction is still good in the near future showing that the trend is downward.
Oct 07th: This is the true beginning of the crash. The predictor shows a falling for 3 or 4 more days and a rebound afterward.
Oct 08th: the prediction still holds for one day but then the rebound that the system predicts does not show up.
Oct 09th: the prediction captures the last changes in the market allowing two more days of accuracy.
Oct 12th: the prediction is valid one day and then an unexpected fall begins.
Oct 13th: the predictor in unable to follow the changes and fails to predict the subsequent fall. From our point of view this day there was an event not related with the inner dynamics of the market, which forced the DJIA to continue falling.
Oct 14th: the predictor cannot assume that a negative trend is at work.
Oct 15th: the negative trend continues and the systems fails again to predict it.
Oct 16th: this is the last close before the big drop. The system begins to show erratic behavior.
Oct 19th: the big crash takes place and the rebound of the market arrives. The system was expecting this take over before, which means that the market, due to external circumstances, exaggerated the fall. The predictor shows now a good agreement with the trend that follows.
Oct 20th: the predictor shows erratic behavior with a little growing. The same as the market. No defined trend yet.
Oct 21st: No trend yet defined.
Oct 22nd: finally the trend shows up on the predictor.
Oct 23rd: the trend is now well defined. Buying is recommended.
Oct 26th: buying and holding positions is recommended.
Oct 27th: buying or holding is recommended.
Oct 28th: the trend continues. Buying or holding positions is recommended.
Oct 29th: the trend continues.
Oct30th: The predictor still shows trend.
From this analysis it can be inferred that this crash occurred as a continuation of a standard downturn of the market, that our predictor could capture very well. The anomalous period from our standpoint is from Oct 13th to Oct 15th. Those days, maybe due to guarantee reposition, the market exaggerated the fall beyond the usual limits imposed by its own dynamics. We are not aware that news affecting the market behavior arrived any of those three days, and we conclude that the guarantee reposition can be looked as the external shock causing this anomaly because it is a rare event that is not taken into account by our prediction system as a typical behavior. |
||||||
| ... back to Stock market forecast? ... 97.04.29 |
||||||
