Add training workflow, datasets, and runbook

training_data/curated/text/bc2bb2bbe3a55521695c01a7da02f066491cc25ee7053ac7f24ffb6e4bacbb18.txt

@@ -0,0 +1,37 @@
+832 Part VI: Measuring and Trading Volatility 
+moved two or three times that far with great frequency. Finally, there is a continu­
+ity to the points on the histogram: There are some y-axis data points at almost all 
+points on the x-axis (between the minimum and maximum x-axis points). That is 
+good, because it shows that there has not been a clustering of movements by XYZ 
+that might have dominated past activity. 
+As for what is not a "good" histogram, we would not be so enamored of a his­
+togram that showed a huge cluster of points near and between the "-1" and 'T' points 
+on the X-axis. We want the stock to have shown an ability to move farther than just the 
+break-even distance, if possible. As an example, see Figure 39-5, which shows a stock 
+whose movements rarely exceed the "-1" or "+l" points, and even when they do, they 
+don't exceed it by much. Most of these would be losing trades because, even though 
+the stock might have moved the required percentage, that was its maximum move 
+during the 10-month period, and there is no way that a trader would know to take 
+profits exactly at that time. The straddles described by the histogram in Figure 39-5 
+should not be bought, regardless of what the previous analyses might have shown. 
+Nor would it be desirable for the histogram to show a large number of move­
+ments above the "+3" level on the histogram, with virtually nothing below that. Such 
+a histogram would most likely be reflective of the spiky, Internet-type stock activity 
+that was referred to earlier as being unreasonable to expect that it might repeat itself. 
+In a general sense, one doesn't want to see too many open spaces on the histogram's 
+X-axis; continuity is desired. 
+If the histogram is a favorable one, then the volatility analysis is complete. One 
+would have found mispriced options, with a good theoretical probability of profit, 
+whose past stock movements verify that such movements are feasible in the future. 
+ANOTHER APPROACH? 
+After having considered the descriptions of all of these analyses, one other approach 
+comes to mind: Use the past movements exclusively and ignore the other analyses 
+altogether. This sounds somewhat radical, but it is certainly a valid approach. It's 
+more like giving some rigor to the person who "knows" IBM can move 18 points and 
+who therefore wants to buy the straddle. If the histogram (study of past movements) 
+tells us that IBM does, indeed, have a good chance of moving 18 points, what do we 
+really care about the relationship of implied and historical volatility, or about the cur­
+rent percentiles of either type of volatility, or what a theoretical probability calcula­
+tor might say? In some sense, this is like comparing implied volatility (the price of the 
+straddle) with historical volatility (the history of stock price movements) in a strictly 
+practical sense, not using statistics.