How to research and verify trading ideas

[u21c3f6]

Back to the subject matter.

As for "surprise" earnings, this has worked well for many years. I first came across this in a book titled "The Science of Winning" by Burton Fabricand (a physicist) published in 1979 (was he the first?). The book had two parts, one on horse race wagering and one on the stock market. I recommend the book not so much because I think it will make you rich (I do think it will help you to be profitable) but because (at least in my case) it gets you to think about things differently. I actually bought the book because at the time I was more interested in horse race wagering than in the stock market! (and frankly I still am more interested in sports wagering because it is more exciting)! :eek:

The way the author looked at this was he compared the earnings estimates with the actual earnings. He broke the earnings into 20% higher, 10% higher, flat, 10% lower and 20% lower. He showed over his testing period that overall the higher earnings outperformed the lower earnings when compared to the DJIA. If the DJIA was up, the 20% and 10% stocks were up higher than the DJIA and the other stocks were not up as much or were lower. When the DJIA was down for the period, the higher % stocks were not down as much and the lower % stocks performed worse than the DJIA. He basically bought the "surprise" earnings stocks and continued to hold if earnings continued to be reported above estimate. If future earnings were revised down or reported down, he sold.

He briefly touched on shorting and options (which were not anywhere near as available then as today). The obvious approach was to short stocks that had "surprise" earnings lower and to buy the appropriate option for the "surprise" earnings higher or lower stocks. Bottom line is that he used this method to significantly outperform the DJIA from that time to at least his last edition of his book in 2002.

Joe.

 

[dtrader98]

Quote from maler:

Here is a little event study I did looking at the tail part of
talon's index rebalancing idea.
I looked at all additions to the SP500 from March 2003 up until
last August and retained all those that intersected the universe
of securities that I monitor. I monitor roughly the top 500 liquid
names listed on NYSE or NASDAQ and refresh the universe every
6 months. Also there is no suvivor bias in the way I look at things
because I consider all securities available at the time when I sort them by liquidity (so for example the universe in 2003 contains the most liquid stocks as determined on the first of Jan 2003 even though many stocks may have been aquired or were delisted meanwhile).
The spreadsheet is self explanatory. On the effective date,
short 100K worth of stock and hedge it with a long 100k worth of
SPY, executing everything on the close. Exit on the close 25 business days later.
I did not account for dividends or transactions costs
but the average profit will easily cover even very unfriendly execution and financing cost terms.
The results speak for themselves (the tStat of 4.9 is way beyond
the realm of just randomness even without accounting for heteroskedasticity).
A slight improvement could be had by being beta rather than dollar neutral, and scaling the size of the bet with the inverse of the specific residual stock volatility but these are minor refinements.

maler

Very Nice Work maler! A rarity here on ET. Someone posted some actual (verifiable and objective) work
and shared it. A new poster to boot.

Although I did not fully verify (yet), it looks sound; 2 things stand out.
1) Ret col sign is backwards (although
the error didn't translate across calcs-- I see you subtracted on Pnl calc-ok).
2) Did you look at the non-hedged rtns?:eek: Not as smooth, but sharpe is almost equal.

Thanks to Talon and maler for something I don't see very often here.
Seeds of a profitable system that is current and actual work to verify it in an objective manner.

Gann, you also get kudos for dipping your toes in the abyss.

 

[talontrading]

How is it the same? You "obtained" the system in the very first post of the thread and since you were helpless you "obtained" a link with tickers and trade dates a few pages later.

As for candor, I think I was perfectly candid and honest from the beginning. The tone (look it up if you're unclear on the difference between tone/candor) of this thread changed when you destroyed it with your hostile look at me attitude.

You were simply off topic and out of place.

I have decided what to do with this thread. It's done... I have no desire to deal with BoWo (and I agree I have ended up looking like a real idiot here), but I will write an article or series of articles sometime soon and deal with these issues. I have written for all the major magazines... will probably be SFO this time so keep an eye out.

Quote from bwolinsky:

Read some hershey threads, and see how the candor changed when I obtained his system. Just the same here.

 

[talontrading]

Ok... I'll give it a shot... one of our pairs models dynamically fits a GMM model to 5 minute data and sizes the two legs switching between either a GARCH process model or a simple nonparametric regression depending on a goodness of fit test. in addition, the model makes a final trade or no trade decision based on analysis of cointegration on a higher timeframe (which is a modified 60 minute timeframe with overnight gaps adjusted for.)

both the development and actual deployment version of this were coded in C++ and ran on its own Unix box.

this actually traded (automated) for a while but we found it did not perform any better than the much more simple system we currently use.

i did see you patting yourself on the back in one of your other threads for using an "indicator called R squared" in your pair model.

I dunno... can you do that in WealthLab? I'm thinking not.

Quote from bwolinsky:

Provide examples of calculations you do that you think WL can't...still waiting....DING... THERE AREN'T ANY!

 

[talontrading]

i didn't know that background information joe. thanks for the contribution.

Quote from u21c3f6:

Back to the subject matter.

As for "surprise" earnings, this has worked well for many years. I first came across this in a book titled "The Science of Winning" by Burton Fabricand (a physicist) published in 1979 (was he the first?). The book had two parts, one on horse race wagering and one on the stock market. I recommend the book not so much because I think it will make you rich (I do think it will help you to be profitable) but because (at least in my case) it gets you to think about things differently. I actually bought the book because at the time I was more interested in horse race wagering than in the stock market! (and frankly I still am more interested in sports wagering because it is more exciting)! :eek:

The way the author looked at this was he compared the earnings estimates with the actual earnings. He broke the earnings into 20% higher, 10% higher, flat, 10% lower and 20% lower. He showed over his testing period that overall the higher earnings outperformed the lower earnings when compared to the DJIA. If the DJIA was up, the 20% and 10% stocks were up higher than the DJIA and the other stocks were not up as much or were lower. When the DJIA was down for the period, the higher % stocks were not down as much and the lower % stocks performed worse than the DJIA. He basically bought the "surprise" earnings stocks and continued to hold if earnings continued to be reported above estimate. If future earnings were revised down or reported down, he sold.

He briefly touched on shorting and options (which were not anywhere near as available then as today). The obvious approach was to short stocks that had "surprise" earnings lower and to buy the appropriate option for the "surprise" earnings higher or lower stocks. Bottom line is that he used this method to significantly outperform the DJIA from that time to at least his last edition of his book in 2002.

Joe.

 

[talontrading]

somehow... in all the layers upon layers of bullshit... the fact that a guy took the time to do a decent backtest got lost.

thank you.

at first glace those results look better than our actual trading results, but like i said i know i cost us some money screwing around with options... the difference may simply be that... of course position sizing is also a big part of it but i tried to guess-timate to get close to what you would have done.

thank you so much for your work.

Quote from whoisjohngalt:

I was suggested that ET people are bad with numbers so... here's a trade-by-trade equity curve...

 

[size]

Quote from talontrading:

and I agree I have ended up looking like a real idiot here

Rather than looking like an idiot, you have gained my respect, and I am sure most reading this thread would say the same. Over the past few years I have only checked in to this site every few months or so, but this thread has been holding my interest.

I know that you do not like to optimize, but have you (or anyone else here who did the backtests) tried filtering out those stocks which had earnings reports within the holding period? I think the logic of doing that would fit into your way of thinking, since an earnings report is event which randomly interrupts the expected action, though I suspect the positive and negative impacts of might balance out over time.

 

[talontrading]

Thank you for the kind words, but... you have to admit I did come off looking like an idiot in that exchange.

As for your filtering question, in a word no. I'm sure you could do that... my sense from kind of remembering these trades (this isn't something I spend a lot of time thinking about. I often don't even know we're actually trading these until they're on because it's so streamlined someone else puts them on. no thought involved.) is that it would be a wash. and you're vulnerable to a lot of other bad events like bad comments from the company, news from the ceo... all kinds of BS like that can hurt you as badly as earnings.

What we do is we know we have a "problem" (I put it in quotes because it's so small relative to our capital it could never become a real problem where I lose sleep over it) and we deal with it in other ways. For instance, PCLN we dealt with this way -- one of our trainees has been very focused on learning to daytrade stocks and has been doing this about 6 months. 6 months ago this person didn't know what a bid ask spread was and it's difficult to imagine someone having a background further from finance. Regardless, in 6 months she's built a good degree of consistency and has been adding size.

We came in and I knew that we had a problem with the PCLN in the adds/drops account so I told her that was her stock for the day, we increased her size even more and I laid out what I would look for if I were trading it. She had never traded a stock as violent as I expected this could be so it was kind of a trial by fire. She basically bought heavily into the pullback off the open, scalped around the position she had built, and was fortunate enough to have held a base position for most of the day. Anyway, end of the day she had made back a significant chunk of the loss for the short. I hate it when people start throwing their actual dollars around here (and most of it is probably BS anyway), but I'll tell you here because I am proud of her and proud of our teaching process that the amount would have bought a nice house in most parts of the country.

So that's how we deal with problems like this. Sometimes we make it worse... sometimes a lot worse, but much more often we offset most or all or more than the loss. That's why it's hard to track performance of some of my more discretionary programs. Did this trainee really just have a great day where she made 50X her average day on PCLN? Obviously not, but naively looking at our accounts would make you think that... and there are similar bumps in many of our accounts.

Sorry if I came off arrogant in this forum. I don't actually think I'm a really great trader. I have friends who do most of the kinds of trading I do better than I do... some so much better it's scary. What I am very proud of though is the training and teaching. Time will tell, but I think I have been very successful and hopefully some of my students will be more successful than I have been.

At any rate I've gotten some nice PM's and I hope people got something out of this. The 4 systems I outlined do have an edge and should be usable... the first 2 fairly out of the box (AFTER you do the homework), the intraday SP will take a lot of work but I gave you the basics... the swing trading discussion you can't get anything out of but at least you've seen one approach there.

We could have talked more about pairs but since I am such a dick (no, serious, I am) I dont want to give BoWo anything that he might be able to make use of. He emailed me and informed me that it was impossible to construct intraday pairs trades because... well he said "Pairs on intraday values are inferior to daily systems, because when you hit that magic level, be it overbought or oversold, you instantly take the trade. The problem you have is that it continues to run against you, hence why no pairs trade should ever be put on at any level below daily." I count at least 5 errors in those 2 sentences and I have no desire to help him out... so there will be no further pairs discussion sorry.

Quote from size:

Rather than looking like an idiot, you have gained my respect, and I am sure most reading this thread would say the same. Over the past few years I have only checked in to this site every few months or so, but this thread has been holding my interest.

I know that you do not like to optimize, but have you (or anyone else here who did the backtests) tried filtering out those stocks which had earnings reports within the holding period? I think the logic of doing that would fit into your way of thinking, since an earnings report is event which randomly interrupts the expected action, though I suspect the positive and negative impacts of might balance out over time.

 

[xburbx]

not sure if you left talon but are you checking pm's still?

 

[4XQs]

talon, it's obvious to all the real traders on this site that this is good lessons being dealt out and plenty of valuable contributions from many ET'ers. Just put bwolinsky on ignore and he won't bother any of us anymore.