Why do I see "Trends" in Randomly Generated Data?

[MAESTRO]

Quote from Thunderdog:

Oh? And why not? Recall what MAESTRO wrote: "...It only works well (no losing days at all) if you are involved in hundreds non-correlated games at the same time."

Clearly, if he is to be engaged in hundreds of "games" at the same time, he cannot be substantially committed to any one. Therefore, his risk per trade must be quite small in relation to his account equity. Further, if he must engage in so many trades/games at once in order not to have a losing day, then this suggests that the reliability is in the number of trades rather than the individual trades themselves. (Sound familiar, surf?)

Note, however, that MAESTRO's approach is markedly different than your won. You take very few trades and hold on to them irrespective of how much they go against you, adding to your position as they do.

You got it!

 

[Thunderdog]

Quote from MAESTRO:

Generally speaking you are right...

Thank you. Surf, you listening?

 

[marketsurfer]

Quote from Thunderdog:

Thank you. Surf, you listening?

yes, thank you for the clarification.

surf

 

[Thunderdog]

MAESTRO, as an aside, could you ballpark the reliability of your entry signals? Approximately what percentage of your trades are winning trades? I know that this is just one element of a comprehensive trading plan, but I'm curious.

 

[MAESTRO]

Quote from Thunderdog:

MAESTRO, as an aside, could you ballpark the reliability of your entry signals? Approximately what percentage of your trades are winning trades? I know that this is just one element of a comprehensive trading plan, but I'm curious.

64.8% wins 35.2% losses. Average win/loss is 3.2/1 money wise. Around 2.8 trades per game a day multiplied by 200 games on average per day. There are 4.8 loosing days per year. One loosing day = 3.76 average winning days (in terms of dollars).

 

[Corey]

Ironically I was just reading a piece the other day on %profitable vs %loss and $gain vs $loss.

What is the better strategy ... the one that runs 55% wins vs 45% losses with average win of $500 and loss of $100 ... or the strategy of that runs 80% wins vs 20% losses with an average win of $350 and loss of $250?

Neither ... they both have the same expectancy of $230
:eek:

(mind you, this is a very short sighted example that discounts variance and draw downs ... but you get the idea)...

 

[ramora]

Quote from MAESTRO:

One loosing day = 3.76 average winning days (in terms of dollars).

For the last 2 years what were the reasons for the down days? Are these out of sample events such as unexpected interest rate changes? Or, events that could have changed your models? Is there a way to reduce the number of down days in the future? Has the avg number of down days remained constant?

Thank you for your posts.

 

[Thunderdog]

Quote from MAESTRO:

64.8% wins 35.2% losses. Average win/loss is 3.2/1 money wise. Around 2.8 trades per game a day multiplied by 200 games on average per day. There are 4.8 loosing days per year. One loosing day = 3.76 average winning days (in terms of dollars).

Impressive. Thank you.

A few other questions just came to this idle mind. What is a "game?" I thought it meant trade. Is a game an expected move? Also, do you ever average down? Average up? Scale in? Scale out?

 

[Thunderdog]

Quote from Corey:

Ironically I was just reading a piece the other day on %profitable vs %loss and $gain vs $loss.

What is the better strategy ... the one that runs 55% wins vs 45% losses with average win of $500 and loss of $100 ... or the strategy of that runs 80% wins vs 20% losses with an average win of $350 and loss of $250?

Neither ... they both have the same expectancy of $230
:eek:

(mind you, this is a very short sighted example that discounts variance and draw downs ... but you get the idea)...

Expectancy is an important number. But for someone with my easily offended delicate sensibilities, variance is key.

 

[MAESTRO]

Quote from ramora:

For the last 2 years what were the reasons for the down days? Are these out of sample events such as unexpected rate changes? Or, events that could have changed your models? Is there a way to reduce the number of down days in the future? Has the avg number of down days remained constant?

Thank you for your posts.

Most of our problems are not related to the algorithm but rather to the execution of orders. Rejects, no fills, lag in data, computer failure, exchange lag etc. The only types of days that really cause us problems are the days when there is no movement at all! Also, quality of data and the connection speed are the major factors. Sometimes the distribution patterns change drastically so we need to change the order generating parameters etc. But those things happen rarely. The number one reason for having sown days is Integrity of Data and Executions!