Muck's Model

[Brentastic]

For the record, we're pretty much at the level at which if we finish above it, we'll be in cash for the coming week and if we finish below it, we'll be long a +1 for the coming week.

Can you clarify? What's the level? I'm assuming price level? For instance, the Dow is currently down ~100 points as of this writing (price is 11,950). Are you saying if the Dow ends lower (i.e. price drops below 11,950) you will be a +1 (long) for the week and vice versa?

 

TIA

[muck]

First, I don't trade the DJIA. I only trade the S&P500.

 

Second, I don't like being this close to a "go / no-go" decision, but if the S&P500 closes at or lower than 1271.23, we will be long; otherwise in cash.

[Brentastic]

First, I don't trade the DJIA. I only trade the S&P500.

 

Second, I don't like being this close to a "go / no-go" decision, but if the S&P500 closes at or lower than 1271.23, we will be long; otherwise in cash.

Looking like long as of right now.

[Brentastic]

Looking like long as of right now.

Maybe I spoke too soon. Less than 2 pts within of your level. 3 minutes until close.

[muck]

+1 it is

[Brentastic]

+1 it is

So you go +1 even though it closed within 2.5 points of your level? Interesting to see how things shake out this summer. Either we topped on May 2 or another high is in the works.

[muck]

So you go +1 even though it closed within 2.5 points of your level?

 

As opposed to what? Staying in cash even though a buy signal was triggered?

 

Emotional overrides on quantitative strategies is rarely a good idea. I am long for the coming week.

[Brentastic]

As opposed to what? Staying in cash even though a buy signal was triggered?

 

Emotional overrides on quantitative strategies is rarely a good idea. I am long for the coming week.

I 100% Agree on the bolded and didn't mean to imply otherwise. I guess I'm just sorta shocked that you had a price level so specific that a close above or below that level, no matter how small the variance, would trigger an action. It's also interesting that your signal was inverse of the day's action. I can see the logic on why, just interesting.

[muck]

I 100% Agree on the bolded and didn't mean to imply otherwise. I guess I'm just sorta shocked that you had a price level so specific that a close above or below that level, no matter how small the variance, would trigger an action.

 

You can't have a quantitative strategy without threshholds for implementing and closing positions.

 

I'm not sure I understand your dual position of (i) being surprised at having a specific threshhold, and (ii) being a fan of eliminating emotional decision-making overrides in quantitative strategies?

 

It's also interesting that your signal was inverse of the day's action. I can see the logic on why, just interesting.

 

While I've never actually counted them, I wouldn't be surprised if I had more than 100 different inputs. When added up and analyzed, this week, the tipping point came from hitting a particular price (or lower). Next week, the tipping point could be a function of what bond yields did. Oh, and remember that I look at weekly stuff ... not so much what happened on any particular day.

Edited June 24, 2011 by muck

[muck]

Assuming we don't have some sort of hugh selloff tomorrow, it looks like last weeks' signal worked out ok for anyone following it.

 

In pretty much every scenario, we'll be in cash this coming week. I'll update this if it's different than that.

[muck]

Just performed a bit of analysis on the signals that I found interesting...

 

Over the 1406 historical weeks I've analyzed (going back to July of 1984), the distribution of weekly returns for the S&P500 is as follows:

 

< -5% = 1.7% of all weeks (i.e., about once a year, you'd see a week where the market drops -5% or more)

-2.5% to -5% = 7.1% of all weeks (i.e., about 3-4 weeks a year, you'd see a week where the market drops -2.5% to -5%)

-1% to -2.5% = 17.1% of all weeks (i.e., about 8-9 weeks a year, you'd see a week where the market drops -1% to -2.5%)

-1% to +1% = 39.3% of all weeks (i.e., about 20-21 weeks a year, you'd see the market stay within 1% of flat)

+1% to +2.5% = 23.5% of all weeks (i.e., about 12-13 weeks a year, you'd see the market go up +1% to +2.5%)

+2.5% to +5% = 9.4% of all weeks (i.e, about 4-5 weeks a year, you'd see the market go up +2.5% to +5%)

> +5% = 1.8% of all weeks (i.e, about once a year, you'd see a week where the market goes up more than 5%)

 

**********************************

 

Interesting analysis:

 

For weeks scoring "-1" --

< -5% = 5.5% of all weeks

-2.5% to -5% = 15.9% of all weeks

-1% to -2.5% = 22.0% of all weeks

-1% to +1% = 36.0% of all weeks

+1% to +2.5% = 12.2% of all weeks

+2.5% to +5% = 7.9% of all weeks

> +5% = 0.6% of all weeks

 

For weeks scoring "0" --

< -5% = 1.2% of all weeks

-2.5% to -5% = 6.5% of all weeks

-1% to -2.5% = 18.7% of all weeks

-1% to +1% = 43.3% of all weeks

+1% to +2.5% = 22.1% of all weeks

+2.5% to +5% = 6.7% of all weeks

> +5% = 1.5% of all weeks

 

For weeks scoring "+1" --

< -5% = 1.7% of all weeks

-2.5% to -5% = 6.4% of all weeks

-1% to -2.5% = 14.6% of all weeks

-1% to +1% = 34.7% of all weeks

+1% to +2.5% = 27.7% of all weeks

+2.5% to +5% = 12.2% of all weeks

> +5% = 2.6% of all weeks

 

For weeks scoring "+2" --

< -5% = 0.0% of all weeks

-2.5% to -5% = 1.8% of all weeks

-1% to -2.5% = 14.3% of all weeks

-1% to +1% = 33.9% of all weeks

+1% to +2.5% = 35.7% of all weeks

+2.5% to +5% = 10.7% of all weeks

> +5% = 3.6% of all weeks

 

For weeks scoring "+3" --

< -5% = 0.0% of all weeks

-2.5% to -5% = 3.5% of all weeks

-1% to -2.5% = 9.6% of all weeks

-1% to +1% = 34.8% of all weeks

+1% to +2.5% = 30.4% of all weeks

+2.5% to +5% = 19.1% of all weeks

> +5% = 2.6% of all weeks

 

***************************

 

Said another way:

 

If the model said to go short, the market dropped in excess of -1% 43.3% of the time (compared to 26.0% of the time for all weeks, regardless of signal). So, you'd have been more than 50% more likely to have a losing week on weeks the signal is "-1" than on average.

 

If the model said to go short, the market dropped in excess of -2.5% 21.4% of the time (compared to 8.8% of the time for all weeks, regardless of signal). So, you'd have been nearly 2.5x more likely to see a really big move down on weeks the signal is "-1" than on average.

 

If the model said to buy (i.e., +1, +2 or +3), the market went up at least 1% 45.5% of the time (compared to 34.8% of the time for all weeks, regardless of signal). So, you'd have been almost 1/3 more likely to see a nice upswing in the market on those weeks than otherwise.

 

If the model said to buy, the market went up in excess of +2.5% 16.3% of the time (compared to 11.2% of the time for all weeks, regardless of signal). So, you'd have been almost 50% more likely to see a really big upswing in the market on those weeks than otherwise.

 

And if the model said to buy on leverage (i.e., +2 or +3), the results were even more pronounced (i.e., up at least 1% 51.5% of the time (vs. 34.8% in all weeks) and to be up at least 2.5% 19.3% of the time (vs. 11.2% in all weeks)).

Edited July 1, 2011 by muck

[muck]

We're in cash for next week...

[muck]

A preliminary heads' up...

 

The signal is probably "cash" again next week.

[muck]

Preliminary heads' up ... probably long at the "+1" level next week.

 

..however...

 

...there is a chance we'll be a "+3", depending on tomorrows' trading...

[Brentastic]

[Ursa Majoris]

Preliminary heads' up ... probably long at the "+1" level next week.

 

..however...

 

...there is a chance we'll be a "+3", depending on tomorrows' trading...

Huh. And here am I finally thinking of throwing in the towel and heading for cash across all funds.

[muck]

Huh. And here am I finally thinking of throwing in the towel and heading for cash across all funds.

 

Two things to remember:

 

1) my models are not right 100% of the time.

2) my models tend to get "in" then "out" then "in" then "out" then "short" then "in" then "short" then "out" ... etc ... so, the signal is only for the coming week, and not for the coming month / quarter / year.

 

As we get closer to the close today, I'll let everyone know if it's going to be a "+3" or not. For now, assume it'll be a "+1" as that's more likely than not.

[TimC]

Huh. And here am I finally thinking of throwing in the towel and heading for cash across all funds.

 

I did that to mine last night. Gonna wait a month or two and see how things sort it out. I don't want to take another 50% nosedive like I did last time. Just in case the 2 parties can't get together and master the obvious.

[Brentastic]

Two things to remember:

 

1) my models are not right 100% of the time.

2) my models tend to get "in" then "out" then "in" then "out" then "short" then "in" then "short" then "out" ... etc ... so, the signal is only for the coming week, and not for the coming month / quarter / year.

 

As we get closer to the close today, I'll let everyone know if it's going to be a "+3" or not. For now, assume it'll be a "+1" as that's more likely than not.

I'm not sure if your model accounts for options expiration Fridays but that's today, fyi. My point is that today's market move is probably more based on options expiration, meaning more manipulation and volatility.

[muck]

I'm not sure if your model accounts for options expiration Fridays but that's today, fyi. My point is that today's market move is probably more based on options expiration, meaning more manipulation and volatility.

 

It makes no special adjustment for certain calendar events (holidays, expiration date, etc).

[muck]

Unless the market rallies approx. 1.4% from these levels before today's close, we'll be at "+1" for next week. If it does rally that much or more, we'll be "+3".

 

Good luck to you all.

[muck]

Barring some very unusual activity this afternoon, we'll be selling our long position and going to cash at the close. The market is up a shade more than 2% this week.

 

*********************

 

Summary of the past 15 weeks:

 

4/8 -- "+1" -- next week = -0.6% return for the S&P500 (model lost -0.6%)

4/15 -- "+3" -- next week = +1.3% for the S&P500 (model made +4.0%)

4/22 -- "0" (i.e., cash) -- next week = +2.0% for the S&P500 (model made 0%)

4/29 -- "0" -- next week = -1.7% for the S&P500 (model made 0%)

5/6 -- "+1" -- next week = -0.2% for the S&P500 (model lost -0.2%)

5/13 -- "0" -- next week = -0.3% return for the S&P500 (model made 0%)

5/20 -- "+3" -- next week = -0.2% for the S&P500 (model lost -0.6%)

5/27 -- "+3" -- next week = -2.3% for the S&P500 (model lost -6.9%)

6/3 -- "0" -- next week = -2.2% for the S&P500 (model made 0%)

6/10 -- "0" -- next week = +0.0% for the S&P500 (model made 0%)

6/17 -- "0" -- next week = -0.2% for the S&P500 (model made 0%)

6/24 -- "+1" -- next week = +5.6% for the S&P500 (model made +5.6%)

7/1 -- "0" -- next week = +0.3% for the S&P500 (model made 0%)

7/8 -- "0" -- next week = -2.1% for the S&P500 (model made 0%)

7/15 -- "+1" -- next week = +2.1% for the S&P500 (as of the time of this post) -- (model made +2.1%)

 

S&P500 over period = +0.9%

Model over period = +3.0%

 

In weeks that we've traded, the model has made money in three of seven weeks (but made +3.0% overall in those seven weeks).

In weeks the model was in cash, the S&P500 was down in five of eight weeks (and down a total of -4.3% in those eight weeks).

[muck]

FYI, given all the nuttiness in Congress, I thought I'd let everyone know that probably, the signal for next week will be "+1". Probably.

 

...lots can change between now and the close tomorrow afternoon...

[muck]

Update: Barring something unexpected, we'll be a "+1" for next week. If the market rallies about 1% from here, we'll be a "+3" for next week. In any event, we're going to be long.

[Brentastic]

You just know once they get this debt ceiling raised the markets will rally. It's such a B.S. dog & pony show they're putting on. They won't default and everyone knows it. And when it gets passed, the market manipulators will rally this thing to a new high (probably).