Add training workflow, datasets, and runbook

training_data/curated/text/7901ea09bc4602338e29ce64cb8a1e11735fa1f44721b29de071e409e01565af.txt

@@ -0,0 +1,37 @@
+802 Part VI: Measuring and Trading Volatility 
+Using 1,000 days of data: 
+Median 100-day historical volatility: 48% 
+Median 50-day historical volatility: 49% 
+Median 20-day historical volatility: 52% 
+Median 10-day historical volatility: 49% 
+If these were all the data that one had, then he would probably use a volatility esti­
+mate of 48% or so in his option models or probability calculators. Of course, this is 
+starkly different from the current levels of historical volatility (shown at the begin­
+ning of this example). So, one must be careful in assessing whether he expects the 
+stock to perform more in line with its longer-term (1,000 trading days) characteristics 
+or if there is some reason to think that the stock's behavior patterns have changed and 
+the higher, more recent volatilities should be used. 
+The pertinent volatilities to consider, then, in a strategy analysis are the medi­
+ans as well as the current figures. If the trader were going to be buying options in his 
+strategy, should he use the minimum of the volatilities shown, 48%? Probably. 
+However, if he's a seller of options, should he use the maximum, 130%? That might 
+be a little too much of a penalty, but at least he would feel safe that if his volatility 
+selling position had a positive expected return with that high a volatility projection, 
+then it must truly be an attractive position. 
+In an analysis like that shown in this example, there is nothing magical about 
+using 1,000 trading days. Perhaps something like 600 trading days would be better. 
+The idea is to use enough trading days to bring in some historic data to counterbal­
+ance the recent, erratic behavior of the stock. 
+Among other things, this example also shows that volatilities are unstable, no 
+matter how much work and mathematics one puts into calculating them. Therefore, 
+they are at best a fragile estimate of what might happen in the future. Still, it's the 
+best guess that one can make. The trader should realize, though, that when volatili­
+ties are this disparate when comparing recent and more distant activity, the results of 
+any mathematical projections using those volatilities should not be relied upon too 
+heavily. Those results will be just as tenuous as the volatility projections themselves. 
+Of course, in any case, the actual volatility that occurs while the position is in place 
+may be even more unfavorable than the one the trader used in his initial analysis. There 
+is nothing that one can do about that. But if you choose what appears to be a somewhat 
+unfavorable volatility, and the position still looks good under those assumptions, then it 
+is likely that the trader will be pleasantly surprised more often than not - that actual 
+volatility during the life of the position will tend to be more in his favor than not.