How to research and verify trading ideas

Status
Not open for further replies.
Talon,

After reading that original Larry Connors book, do you have any other recommended readings? Being that this kind of trading is very new to what I have learned previously, I found that Larrys material helped me really catch up with a good portion of understanding this thread. Anything beyond UNDERSTANDING THE MARKETS?
 
Talon thanks for the input on the position sizing question. I put together a quick analysis of how both approaches fare in various vol regimes (using the system in this thread before filters), and the dynamic position sizing performed worse in nearly all scenarios. Now I should note that its far from conclusive at this point because it was only on the 5 symbol portfolio and back to 1/1/2004 (and likely a weak testing methodology). I did not include transaction costs - so as Mike mentions, I believe that this is due to stronger tendency to mean revert with larger absolute distances from the mean, irrespective of the current vol regime - equating to more quality opportunities during high volatility periods. (As in the earlier example, what i'm trying to say is that being down 6% when avg daily vol = 4% is a higher quality signal than being down 3% when avg daily vol = 2%...this is despite both being 1.5x the avg daily range and equally likely to occur within their respective regimes).

I plan on doing a more complete analysis of dynamic position sizing, with the hopes of demonstrating this stronger tendency to mean revert. Admittedly it isn't the most straightforward way of testing this (could do something like compare levels of serial corr in different vol scenarios), but it will be good to get a feel for variable position size. In addition to that potential issue, there is the big one you mention below and the increased transaction costs that Mike also highlighted. Basically it looks like the idea stinks.

Quote from talontrading:


-The behavior of volatility is complex... extremely low volatility conditions can actually lead to explosions in volatility. (this is probably a result of mean reversion.) So... think about this... when volatility is lowest it may be ready to explode... at that time you are putting on your largest position because you size that way.
More...

It does seem to be an improvement, but acknowledging that raises the queston of how does one make money in a prolonged vol crunch?

Also, why is it the ratio to long term vol that matters rather than just the recent realized vol? I would think that the duration of our trades is short enough that vol reverting to its long term mean isn't a consideration. If it is just the recent realized vol that's important, then I guess that would answer the first question - only trade high vol products during a crunch. You're pnl will be reduced because of trading a smaller # of stocks but everything else should remain the same...?

Quote from talontrading:

Question: does this look like a worthwhile addition to the system? Thoughts??
More...

This comment looks like it was intentionally placed, but I'm not sure that I fully understand its consequence. Does this imply that one should be wary of a filter which removes your big losers (you most likely curve fit)? If the big losers were removed after you applied the filter in this instance, would the same logic apply even though the filter was not very selective (number of trades was reduced by 60%)?

Quote from talontrading:

Biggest and smallest trades are no affected.
More...
 
If we would have followed this strategy as-is, we would have almost doubled our money across the backtest period.

But if would have just bought and held those same five stocks you noted for the backtest, we would have more than tripled our money.

(And if we put $1000 at risk on each trade, and charge ourselves $2.50 to buy, and $2.50 to sell, this strategy comes out at a net loss.)

I guess I'm a little underwhelmed with the results. :(
 
Also, what happens when you run this strategy on a company that cratered like AIG, GM or C? Does our short side provision make us money? Or do we chase these stocks all the way down and bleed?
 
Quote from talontrading:

Use something like this:

= (today's close - lowest low of past 10 days) / (highest high of past 10 days - lowest low of past 10 days)
More...

[
Sorry Talon (or if anyone else could help out), Jut want to check Im on the right track here....
But from what I see here, is it just adding 25% of 10 day range to the the lowest low of the previous 10 days.

So an entry rule would be something like
if (today's close - lowest low of past 10 days) / (highest high of past 10 days - lowest low of past 10 days) < 0.25 then Buy
 
Cool thread! I am not surprised this system seems to work, but I am surprised it works so well on blue chips, I think a lot of stuff like this works on small illiquid stocks but not msft.

So my first concern with this system would be what is your market exposure? We are just looking at it one trade at a time so far, but if you really ran it (on a larger universe) do your longs and shorts tend to balance out naturally? I would guess not - raw returns give you the entry so I think it would be very short after the market is up and vice versa. Assuming you naturally get some market exposure and then hedge it out, do you still get all the returns or does some part of it come from the market as a whole mean reverting?

I threw the trades from BS's spreadsheet into an event study program and looked at the return less the S&P 500. No exits, just day by day returns for 30 trading days after the entry, and data only through 2008. I'll post the results separately, but the short answer is that it works well for the long side in both raw and market-adjusted returns, and doesn't seem to work at all on the short side. Based on this I would say if there is an edge on the short side it comes from the exit, and the long side works fine w/ just a time stop.

So unless the short exits help alot, the way I would trade this system is take the long trades and always stay flat the market - hedge with S&P 500 futures or SPY.

Some other ideas:
1) look at some type of market adjusted entry - do you really want to enter these trades when the whole market has an extreme move, or just when the stocks return in excess of the market is extreme?

2) use a normalized volume filter. reversals are stronger when the original move is on low volume.

3) since the event study results show the returns are still somewhat positive for quite a while after the open signal, and these are large caps w/ relatively liquid options maybe a good exit/risk management strategy would be collar it with the nearest expiration options and exit when they expire. or you could equivalently just do the trade in option spreads. I have never tried this, does it sound reasonable?
 
I ran the event study on the closes too, here are the results for the long side. It looks like the exit signal adds some value, but waiting 2 more days would get another 55 bps or so.

Day 0 is the day the close is given. I wouldn't pay too much attention to the negative returns on day -3 and earlier, these will include some datapoints from before the trade was entered for the shorter trades, and we know the pre-entries are down days based on the rule.

So do you guys think this is a valid way to look at it or curve fitting?
 

Attachments

And here are the short close results, it looks like they add value too. I'm not sure how to think about a trading rule where the entry doesn't seem to do anything but the exit seems give you an edge?

Sorry for the string of posts, I can only upload on file at a time.
 

Attachments

Status
Not open for further replies.
Back
Top