How to research and verify trading ideas

If my hedge worked perfectly you would make no money (+ costs). Using SPY seemed pretty random to me but I see now there is some chance you would make money - to the extent the market does what the opposing stock would have done. If that's what works I stand corrected. I really like the idea of using opposing stocks and this seemed like the weakest point.

 

Great thread Talontrader.

I mentioned on the other thread where you first posted this system that I appreciated you doing so. I’m sure there are a number of other folks who appreciate it as well. No question for me, this is the kind of interesting material that occasionally makes ET shine.

Perhaps you could discuss the potential risks with this methodology and any methods you might use to reduce that risk.

I’m also hopeful that at a future date you’ll discuss (without giving away the farm) how you use OHLC data (especially as it relates to short and intermediate term).

Thanks again and please ignore the “Look at me look at me look at me” internet attention seekers. Nothing infuriates them more than a non-response.

 

let's do the background work on this system idea first and then we will look at some OHLC ideas so you can see how i evaluate those.

biggest risk is of course for some reason it just stops working one day... or there's always the chance that a short will explode 500% in your face. (due to the nature of the system and the what just happened to the companies it seems very unlikely... but if you did this for long enough you'd have bad surprises. i get up every day and remind myself, literally everyday, that ANYTHING can happen today.) best defense against both of those is to make sure you are trading small enough... maybe 20% or less of your trading capital allocated to an idea like this, but that's something you have to figure out.

you can play games with options too to reduce the risk somewhat... but i wouldnt advise it. just trade small enough you won't get blown up by a bad surprise.

actually, the biggest danger is "outsmarting" the system. if you do it, do every trade and do every trade the same size. you think you know which ones are the best and which ones to skip, but you don't. just take the trades, follow the rules, be small and sleep easy.

honestly, it doesn't get easier than this. one of the easiest trading ideas i use.

thank you for your kind words!

 

Talontrading,

I would also like to thank you for your contribution. I like to think outside of the box and this does help with that. Just so you know the other day I went through most of the SP500 adds/drops for 2009 and while there were a few losers the winners outweighed the losers.

It seems many traders who know of an edge or an inefficiency do not want to share because they feel that will speed the process of it not working anymore. Do you feel the same way? Are you worried that a system like this being shared on a public forum could cause it to not work? Obviously not if you were willing to share it but I am interested on your perspective.

Thanks

 

OK, so now that I've dealt with all the issues that would keep me from determining individual results for each trade, I guess my next step would be to take "some number" of past trades and do that computation. Advice on how many to do would be cool since it seems like going too far back is going into fundamentally different historical markets.

The first order of business would be to get the geometric mean of the returns adjusted for fees, bid/ask etc. That number will tell me if this is worth pursuing further.

Then I guess I take the histogram of the returns, fit some sort of distribution (normal?) to it to fill in hypothetical outliers that are missing from the small data set, and then use that result to compute the Kelly position size, and divide that by something between 2 and 4 (since I hate ulcers) and see where that puts me in terms of position sizing. See what kind of returns on my total account this will give me. Again, it may be time to stop there if it's just not worth it, or to decide I need to start playing option or stop games because Kelly says the position has to be tiny due to outliers.

From there, I would probably do a SMA on the returns, plot it, and look at it. I'd be curious to see if the point where the papers were first published is visible. I'd also be curious how much variance there is in the medium term results.

At that point, if the returns look good, the position sizing looks rational, and the SMA on returns looks flatish, I start trading tiny volume or a simulator to see what I missed. If I'm getting the prices I think I should be getting and nothing else bites me in the ass after say 20 trades, I start trading real money.

 

The other thing I forgot was sensitivity analysis on the data - I'd probably ask "how many of my best historic trades would I have to replace with my worst before this method tanks". If the number is 2, I need more data. If the number is 17, I'm probably good to go.

 

I don't trade stocks so this subject is really outside my expertise. However, I wonder how many times in a year you can trade this strategy. Doesn't S&P shuffle their index like every 3 or so years? Certainly not on a monthly basis as far as I can tell.

Therefore, my question is if you buy (or short) and hold your position and "exit the trade 25 days after the effective change date", what are you supposed to do for the next 35 months provided that the 3 year cycle is correct?

 

Patience is an edge.

 

Definition of a great thread:

Player 1 knows of which he speaks and has walked the talk

Player 1 exposes the detractors only by letting them talk and expose themselves.

Player 1 thread provides a great reference for this thread but also of who to ignore in all other threads.

============

 

I'm going to go out on a limb here and guess talontrading's going to be pissed at this one.

Why? Becauase had you actually bothered to look at the additions and deletions, you would know they happen on more or less a monthly basis - the average rate is probably greater than 1 pair per month. With roughly month long trades, this produces plenty of signals.

Now, here's a really good question: you say "as far as I can tell" but clearly that didn't actually involve looking to see how frequently additions/deletions happened. So how far did you go to tell? it couldn't have been very far...