travis, imo - back testing keeps you in the past and tomorrow probably won't be the same as yesterday...look ahead instead and test tomorrow, the remainder of the week, then next week and you'll get a clearer picture...but don't hold the results as Gospel because it all changes over time, just recalibrate when it does.
i did some testing on this a while ago and i also found that the results were contrary to what i had expected... - i ended up not reading too much into the results because some of the drivers of volatility are period-dependent - for example, i have no idea when they launched the disclosure of Non-farm payrolls, but that must have had an impact of friday's volatility - also, when oil started going through the roof, the Wednesday oil number stirred up the equity indices quite a bit.
the mkt will always be under one or more news driving volatility (oil, payroll, home construction etc.) and in the future there will be others. i think the best thing to check is the +/- on the VIX that will give you a better understanding. i assume you can get it from yahoo and compare the days of the week. if you will find a reason why the volatility is going down on the long run i will be happy to hear good luck
I noticed that people are not commenting on bullish days of the week, but rather on volatility (it's my fault - I picked the wrong title for the thread). However, I found no correlation between days of the week and volatility. Instead I found a correlation between early week and bullishness, especially on mondays. Concerning your comments on volatility, of course I agree that it depends on the news, but the question is not whether it is influenced by news and events, like September 11th. The question is whether, besides these obvious influences, there is a bias due to the day of the week, like for example on mondays, when people have been waiting for 2 days to buy, and they should be restless. Yet, on my study of the Dow Jones, I found nothing. Or maybe I studied it the wrong way. For my study, volatility is not range, or ATR, which is the real definition of it. To make things simpler, I substracted yesterday's close from today's close, and made that value absolute (in order to be able to sum them, and prevent a +10% to null a -10% move). I think I will now go back and do that other calculation, and will let you know if I find any different results. ...10 minutes later... I went back and the results do differ. There is a difference in volatility between days of the week. If we count the 75 year period, the difference is as follows (counting each day's range, high-low, and summing the values for 1 year): Monday to Friday in last 75 years 83% 89% 90% 88% 86% That guy was right, when he said wednesday is the most volatile day. But not by very much. And, surprisingly, monday is the least volatile day. But it does go up, and maybe that is why it is not volatile. In an uptrend volatility is less. Now if we compute the average daily volatility in various days, we get: Monday to Friday in last 75 years: 1,72% 1,75% 1,77% 1,74% 1,71% Monday to Friday in last 5 years: 2,05% 2,07% 2,15% 2,11% 2,02% Whichever way we put it, wednesday is more volatily than other day, in any period of the Dow Jones history. But the difference is so small that it cannot be exploited to make money, I think. The big difference remains this - during any year on the last 20 years, on mondays we go up 21% and down 15%, whereas on thursday we go up 17% and down 19%. Also we can say that all (overnight) gains during any of the last 20 years were made on Mondays to Wednesdays, and the rest of the week adds up to (overnight) losses.
Another question I ask myself at this point is this. If my system is good at detecting the conditions for longs and shorts, knowing that mondays are a bit biased to the bullish non-volatile side, does it make sense to prevent it from trading short on mondays and from trading below the previous day's close? I don't think so. I think the bias we discovered, even the bullish yearly +6% bias, is too small, to adapt my system to it. -------- Another question is - what if within certain months these values are stronger? What if there are monthly biases that affect the daily biases making them strong enough to use them as a parameter in your system? This is what I will study next.
we recently did some stuff on long term dow jones data as well. i was especially interested in how the market used to react after dried out range as of recently. the 15 day range was extremely low and it turned out that this is actually bullish for the following 30 days. i find this kind of statistical work quite interesting and we will move forward to adding other time series into the picture. first one will be interest curve moves and their effect on dow jones returns. if anyone else is interested we could set up a thread dedicated to this kind of work. i was actually astonished by the weekday results. when i did something like that on sp futures a few years back i did not have such strong results. actually my thinking was more that weekday effects are not tradeable anymore, while year end stuff is still in place. so these results here are very interesting. thnx travis for bringing them to the board.
I've heard some pretty interesting stuff about the seasonal trends of the market. Something about a period in late sept-early october that has never been a negative period on the essnpee. There was a guy posting about it here somewhere, I'll see if I can find it. You may want to look at "seasons" or some longer time frames after you've done your month analysis. Of course, that may not fit your trading style at all, but it could be interesting. TNG
You get lower volatility in the markets in the direction of the long term trend. The same was true of 30 years treasury last time I did the analysis.