Hey Talon, Don't let the crowd get you down. I've seen similar results a long time back, but can't chime in on how well it would work today. The strategy (and variants) seem so well known by now, that I wouldn't expect it to; but then again, I haven't run any tests lately. I would ask you to post statistical results, but since your objective is to teach fishing, then I can understand your reasoning. As for the nay-sayers, the onus is upon you to prove him wrong. He gave you explicit and objective rules that you could go back and test, then come forward to debate the hard statistical results. If he started to add funny qualifications after your results proved him wrong, like, well it depends upon certain unquantifiable things, etc.. then you can get out your pitchforks. Until then, however, quit complaining and let the man speak. If there is anything I can tell you, Talon, based upon my experience here, is that, don't expect more than about .05% of any posters here to actually do any type of work and present it. As you astutely observed, people want the fish, not the fishing mechanics. And there's no shortage of hecklers in the crowd (which is fine as long as you have sound reason to heckle).
I got some very supportive notes and posts here. Thank you for those. I don't feel we made great headway in the discussion re why it should work. We won't ever know the real reason, but to me there is a clear explanation. There are MANY funds that must hold the S&P 500 or are only allowed by charter to hold the S&P 500 stocks. So, imagine a stock that is deleted. All of a sudden, many large institutions are not allowed to hold this stock so what must they do? Sell and sell before the drop date? What will that selling do? Make the stock go down. What happens when artificial selling pressure is removed? The stock goes back up. Reverse for additions to the index. Does that make sense?
it is me belief that you can't do this on a shoestring, contrary to what a lot of internet posters would have you believe. I think anyone who thinks they are going to learn to trade and has less than $10K available.. first go to vegas and put it on red a few times... your odds are better. Having said that, most analysis can be done rather simply in excel. tradestation is good for some things but you have to really learn how to program around easy language. i am a big fan of R and visual basic for easy things. I think you would be best served getting TS if you don't already have anything... and if that takes a big chunk out of your capital you don't have enough to trade with. I would advise that person to spend money learning all they can and trade on a sim until they can show consistent results... then see if someone will stake you to cover your first losses and split the profits. You can't learn this without having some losses. Having a bankroll to sustain those first losses is critical.
we don't put a lot of weight on monte carlo... it's overused by math geeks and engineers in my opinion. in this case, if i'm sizing a system with each trade risking a % of account size, the equity curve is extremely path dependent. meaning if i have large losses at the beginning vs large wins clustered at the beginning it's a completely different thing. so it's just to get a feel for the variability and interaction of position sizing with the system. NOT a critical part of the test imo.
A completely different question is monitoring an existing system to make sure it still works. How do you tell a normal drawdown from something that needs your attention? things do change and systems do stop working. We maintain a lot of stats on our systems... a slight oversimplification would be assume i have 1000 backtest datapoints... i know mean and standard deviation of those trade returns. if i look at the results i'm getting and start seeing a lot of > +/- 3 standard deviation returns (allowing for distribution of the trades... some systems are kind of screwey) then i have to pay attention. there's math involved but more intuition. keeping a running tally of recent stats vs. longer period vs. backtest lets us know when a system is performing within expecations... kind of like manufacturing processes keep control charts. make sense? btw... this system does still work.
Just curious, talon, why do you trade all of these things? If a person could, for example, trade the ES profitably on a daytrade basis, why would they 'over-extend' (if that's the right word), themselves by scattering their attention with other instruments? Hope your thread accomplishes something for somybody before getting hijacked.
Perfectly plausible hypothesis. Although, years ago, the buying and selling activity were much more transparent, IMO. You could see gigantic block activity going off pre and post announcement, and the post buying lasted a while. These days, with the advent of dark pools, algo trading, 3rd party exchanges, etc... I think it is much easier to disguise that type of activity and have less of a pronounced effect on the observable stock price pre/post addition. Which is one reason, I intuitively would expect the phenomena to be less pronounced now. There are some studies around that show that modern post announcement activity has shorter decay time as a result of such efficiencies, than was evident in prior behavior over the years.
If there is an artifical selling pressure, won't funds sell before the deletion but after the announcement? So the effect would diminish because funds try to get out fast after an announcement?
And one last post at the end of a long day... Here's why mr bwolinksy irritated me... i suggested googling "s&p 500 index additions deletions" without really knowing what it would pull up and this pompous ass replied as you can see below. However... today I googled that exact phrase and noticed: the very first link on the search results http://www2.standardandpoors.com/po...ices_500/2,3,2,2,0,0,0,0,0,4,1,0,0,0,0,0.html gives a list of all index changes. gee mr bwolinksy... there have been a few more than GM recently. Sorry you "don't recall" them but the worst part is you didn't have any idea how to find out. the second link http://www2.standardandpoors.com/spf/pdf/index/070902pricechanges.pdf is a study from the S&P website itself acknowledging the effect I mention in the system. the remainder of the first search results page is a list of academic papers (that i would hope bwolinksy would find acceptable) on the phenomenon. Point being that anyone with a little bit of interest could have spent 5 minutes googling and realized they might have something interesting in this system... and then you have the dates of the changes and links to press releases that would give you the announcement date... and then you could think about backtesting, couldn't you? that is what i hoped would happen when i threw the idea out there.