from Tom Ryan re VIX; posted manually [Tom: possibly you sent to <tech-spec-bounce@xxxxxxxxxxxxx> rather than <tech-spec-@xxxxxxxxxxxxx>?] Subject: VIX if up Friday-Friday cls Dow 01/01/1990 - 09/17/2004 VIX 01/01/1990 - 09/17/2004 Calc dVIX Friday-Friday cls as % (VIX,2 - VIX,1)/VIX,1 Calc dDOW Friday-Friday cls as % (DOW,2 - DOW,1)/DOW,1 If dVIX>0 dDOW next week; 353 mean=0.43% stdev 2.3% If dVIX<0 dDOW next week; 387 mean=0.05% stdev 2.1% Ok If dVIX>0 create a table of values with time stamp Calculate 3 day sum add to table Now If last 3_day_sum is above or below a value, calculate mean expectation stdev, number trades set number mean sdev all 353 0.0043 0.023 >.05 46 0.0087 0.021 >.04 78 0.0063 0.021 >.03 115 0.0058 0.019 >.02 158 0.0046 0.019 >.01 200 0.0044 0.020 >0.00 247 0.0045 0.020 <0.00 106 0.0037 0.031 <-.01 78 0.0022 0.034 <-.02 58 0.0021 0.037 <-.03 42 0.0000 0.042 <-.04 30 0.0075 0.037 <-.05 24 0.0140 0.037 all weeks 741 0.0016 0.022 1990- Q:What the hell am I doing. A:For years have tortured VIX data. Have noticed that expectations for (changes in VIX) have cycles. Trying to understand the cycles. Hence, take all weeks where VIX was up. The Dow in general has performed better the next week than random and better than when VIX was down. Of those weeks when VIX was up one could adjust trade size based on most recent performance (summed, total) of the past three trade signals (when VIX was up). Notice that if VIX is up this week, and the last three times VIX was up yielded big losses the following week, it gets very bullish. Similarly, if VIX is up this week, and the last three times VIX was up yielded very good profits, the trade continued to be more profitable than the average. When the recent three trades were not terrible, but below the population mean, the trades were probably not worth taking on this VIX-up signal alone. In brooklynese, if it is working work it, if its broke go for it, otherwise leave it alone. Does this make any sense? tom