Trading Lessons/Insights From Coin Flipping

[walterjennings]

Quote from tradingjournals:

Walter: I think it is what he meant. You choose two numbers that you put in two envelops for instance, s/he opens one of the two envelops, and s/he decides whether that number is the largest of the two, or the number in the other envelop is the largest. S/he can have a positive expectancy, because s/he will use the information from the opened envelop to decide (more intelligently) on the number in the unopened envelop.

Based on the wording in the original game.

"Choose N1=50 and show that number, flip a coin to decide if N2 is 51 or 49. It should be impossible for them have a chance other than 50% of guessing if N2>N1. Which will lead to positive expectation with the imbalanced payout."

Beats the game. I am assuming there was an error with how it was described if it is actually possible for it to be unbeatable.

 

[tradingjournals]

Quote from walterjennings:

Based on the wording in the original game.

"Choose N1=50 and show that number, flip a coin to decide if N2 is 51 or 49. It should be impossible for them have a chance other than 50% of guessing if N2>N1. Which will lead to positive expectation with the imbalanced payout."

Beats the game. I am assuming there was an error with how it was described if it is actually possible for it to be unbeatable.

I see your point. I understood it in the sense that the two numbers are randomly generated between 1 and 100 ( uniform distribution). I believe that his point is possibly that the conditional distribution of the second number as it relates to his payoff maximization will change depending on the first number revealed. So s/he would add some additional condition to decide which is to choose. There is an algorithms known I think as as the Odds algorithm that solves problems like these.

 

[masterjaz_99]

Quote from Mike805:

Let's try a few simple tests.

Categorize the trades by entry in one hour increments.

So the stats would look like

630am-730am : PnL, Avg. P, Avg. L, #Trades
730am-830am : PnL, Avg. P, Avg. L, #Trades

Etc...

If its daily then find the VIX closing on entry and the day:

10 >= VIX < 15 : PnL, Avg P/L Etc
15 >= VIX < 20 : PnL, Avg ...

Day of week entry:
Monday : PnL, Avg P/L etc
Tuesday: PnL

etc etc

Mike

Mike,
Great minds think alike. I've actually broken the day down to 15min intervals and there are times of day where the 5dma or other filters work better than others--namely the first 30-45min.

Here is a plot of 5dma for the entire day. Red line is below, blue line is above. Both lines assume a long entry, hence going short blue would flip the line...

More to follow...

M

 

[u21c3f6]

Quote from tradingjournals:

Walter: I think it is what he meant. You choose two numbers that you put in two envelops for instance, s/he opens one of the two envelops, and s/he decides whether that number is the largest of the two, or the number in the other envelop is the largest. S/he can have a positive expectancy, because s/he will use the information from the opened envelop to decide (more intelligently) on the number in the unopened envelop...

This goes back to the selection process I wrote in a previous post.

I would choose two numbers that were either both higher or both lower than 50. Then if one of these numbers was selected at random, the chance of guessing correctly whether the other number was higher or lower than the first number selected has a long-term expectation of 50%.

Joe.

 

[masterjaz_99]

Now we have the 5dma for only the 0930 to 0945 period. A couple things to note: 1) period overall is a poor time to be long--go short! 2) Red line is very volatile, whereas the blue line is stable. I'd pick the blue line and run with it, going short above the 5dma, even though it results in less profits...but not really. Not may favorite filter for this time period.

The chart:

M

EDIT: Don't hold me to this data, I am re-analyzing my tick data, which takes some time and care (i.e. correcting for day light savings...). Period is roughly 01012006 thru 06302009.

 

[Mike805]

Quote from masterjaz_99:

Here is a plot of 5dma for the entire day. Red line is below, blue line is above. Both lines assume a long entry, hence going short blue would flip the line...

Can you clarify this for me? I'm not sure I understand, what do you mean by entire day?

 

[masterjaz_99]

Quote from Mike805:

Can you clarify this for me? I'm not sure I understand, what do you mean by entire day?

Open to close, 0930 EST to 1400 EST.

EDIT: The plot is cumulative points for ~900 trading days (horizontal axis) if you go long the open and exit the close. Red line is cumulative total for going long when prior day's close was below 5dma, blue line is long when prior day's close was above 5dma. Hence you should go short and that line will flip over.

 

[Mike805]

Quote from masterjaz_99:

Open to close, 0930 EST to 1400 EST.

EDIT: The plot is cumulative points for ~900 trading days (horizontal axis) if you go long the open and exit the close. Red line is cumulative total for going long when prior day's close was below 5dma, blue line is long when prior day's close was above 5dma. Hence you should go short and that line will flip over.

Gotchya, thanks.

EDIT: Have you looked at the 10am PST (1pm EST) timeframe as entry?

I'll PM you some interesting stuff that may spark some ideas.

 

[masterjaz_99]

Quote from Mike805:

Gotchya, thanks.

EDIT: Have you looked at the 10am PST (1pm EST) timeframe as entry?

I'll PM you some interesting stuff that may spark some ideas.

Sounds good. I also a general downtrend for the 1300-1400EST time period, although with much chop...Look forward to the PM.

M

 

[masterjaz_99]

Quote from masterjaz_99:

Mike,
Great minds think alike. I've actually broken the day down to 15min intervals and there are times of day where the 5dma or other filters work better than others--namely the first 30-45min.

Here is a plot of 5dma for the entire day. Red line is below, blue line is above. Both lines assume a long entry, hence going short blue would flip the line...


More to follow...

M

Okay, so back to improving a 'coin flip/50-50' method. Today, for example, I bought NQ at 1791.75 and sold at 1819.00.

My current idea is to wait for a move before entering the position, such as a retrace/pop 1pt against the desired direction. Today, I would've wanted a move to 1790.75. It just so happened to be the LOD, so I do not know if I would've gotten a fill. Had I been filled, awesome an extra point of profit, had I not, missed a 30 pt move. Is getting 1pt per trade worth missing a 30 pt move????

Also, what if I waited on the exit as well for a 1 pt favorable move. NQ closed at 1819 (I actually sold the close value according to my feed), but could I have waited for 1820? Today, yes, I could have and got another free 1pt. It's traded to 1823.50 so far AH. But how much would I have to give up in order to secure that 1pt???

These are my thoughts going forward...any other ideas???

M