Kelly for traders

[kut2k2]

Quote from maxpi:

kut2, thanks for the nicely informative thread.

You're welcome

 

[kut2k2]

Quote from u21c3f6:

"I have no idea where this version of Kelly comes from but to each his own. "


It's the same formula just repackaged. It is basically your edge divided by your odds (this form is mostly used for "gambling" propositions IMO).

k = (bp - q)/b

where

f is the fraction of the current bankroll to wager;
b is the odds received on the wager;
p is the probability of winning;
q is the probability of losing, which is 1 − p.

The adjustment I made based on wager size is my own "creation".

Joe.

OK but as I tried (unsuccessfully ) to point out in Example 2, simplifying even a 3-outcome situation to a 2-outcome situation can lead to miscalculations. This is a prime reason why so many traders have tried and failed to successfully use Kelly in their trade management: trying to force-fit a multi-outcome situation into a 2-outcome formula.

This is also a great example of why William Eckhardt is railing against summary statistics, e.g., average win and average loss. It's like trying to describe the Taj Majal from a single exterior snapshot.

 

[Lucrum]

Quote from kut2k2:


This is also a great example of why William Eckhardt is railing against summary statistics, e.g., average win and average loss. It's like trying to describe the Taj Majal from a single exterior snapshot.

Out of curiosity what does he suggest tracking?

 

[kut2k2]

Quote from TL Trader:

Out of curiosity what does he suggest tracking?

"... we use only the most robust and assumption free statistical tests, We have an aversion to summary statistics that obliterate important structural elements. For assessing systems, we use a technique called bootstrapping so that the complete distribution of past outcomes can make itself felt in decisions; the distribution is not simply viewed in terms of its mean and variance which can give a distorted picture."

http://elitetrader.com/vb/showthread.php?s=&threadid=75851

 

[maxpi]

Quote from kut2k2:



This is also a great example of why William Eckhardt is railing against summary statistics, e.g., average win and average loss. It's like trying to describe the Taj Majal from a single exterior snapshot.

If the distributions of the average wins and losses is tight it should not be a big problem, even if they are maybe not so tight really. Many traders use fixed targets and exits, how far can the frigging averages deviate under those conditions?

Humans tend to go wrong by assuming that the answer to a question has to be complex. They really will run away from a simple solution to a morass of complexity.

 

[kut2k2]

Quote from maxpi:

If the distributions of the average wins and losses is tight it should not be a big problem, even if they are maybe not so tight really. Many traders use fixed targets and exits, how far can the frigging averages deviate under those conditions?

Humans tend to go wrong by assuming that the answer to a question has to be complex. They really will run away from a simple solution to a morass of complexity.

I submit that calculating expected return over expected squared return is at least as simple as first calculating average win, then average loss, then win-rate before finally plugging them all into the 2-outcome formula.

In a spreadsheet, you can do SUM(...)/SUMSQ(...) in a single step.

 

[trading1]

Quote from kut2k2:

I submit that calculating expected return over expected squared return is at least as simple as first calculating average win, then average loss,...

Yes. What is the relevance of "expected return over expected squared return", it may be good, though where does it apply?

 

[trading1]

Quote from maxpi:

If the distributions of the average wins and losses is tight it should not be a big problem, even if they are maybe not so tight really. Many traders use fixed targets and exits, how far can the frigging averages deviate under those conditions:

Even with the best plans many traders are in for a shock when distributions don't work out as expected.

I believe that packaging the trades into a time group, like one months trades, and considering that as a single unit is the best way of using Kelly. That way the averages are unlikely to deviate so far. Kelly sort of assumes a fixed win / loss ratio but in fact a win / loss ratio will vary over time and the best way to obtain a more fixed ratio (or lower the deviations of the averages) is to group the trades, considering a group of trades as a single unit.

 

[kut2k2]

Quote from trading1:

Quote from kut2k2:

I submit that calculating expected return over expected squared return is at least as simple as first calculating average win, then average loss,...

Yes. What is the relevance of "expected return over expected squared return", it may be good, though where does it apply?

I explained that in the original post. Which part is unclear to you?

 

[joesan]

Quote from kut2k2:

I explained that in the original post. Which part is unclear to you?

what is f in the original post?

Accroding to Ralph Vince ,f should be the percentage of the account to be risked in every trade, rather than the percentage of the account to be put into a trade .