Actually that buffer you talk about is only not there if you section your performance off on a yearly basis. However, this is arbitrary. I am mentioning this, since this might have a positive effect on your psychology when you leverage in the beginning. The buffer has been there ever since you started your blog 3 years ago.
Given that you need to compound so frequently, would it not be sensible to 'start on a clean slate' every week?
Nope. One week is too myopic a time-frame to work towards. The fear of losing for the week would cause me not to take any risk, and with that have a flat curve.
Weekly Update for week 2/50 ended 1/24/2009 Moderately positive week, up 9.6K (3.2%). Made it Tues/Wed, lost it Thur, and gained some back on Friday. The market is definitely begging for direction. Code: Opening Balance: 301,530 Net gain for the week 9,626 ------------------------------------------------ Net Balance: 311,156 Number of Trades 23 Number of Profitable Trades 12 Since Inception of Thread 01/10/2009 - 01/24/2009 Opening Balance: 320,064 Net gain (Less Margin Interest) -8,908 (Down 2.8%) ------------------------------------------------ Net Balance 311,156 Number of Trades 46 Number of Profitable Trades 23 Top/Bottom Discretionary Trades for the week TICKER ENTRY DATE/TIME EXIT DATE/TIME QTY PURCHASE AMT SOLD AMT GAIN/LOSS TYPE SPY 2009-01-21-14-32-13 2009-01-21-15-59-20 8000 661960 672000 10016 LONG --------------------------------------------------------------------------- STT 2009-01-20-11-09-02 2009-01-20-14-37-56 6000 111465 105271 -6208 LONG
Continuing on the development of the sizing formula. The first variable is cratio. The idea is to shrink my position size when there is little or no buffer, and to size up to the optimal as the buffer increases. With that in mind, I have found the formula below good: f = (cratio^2)/(1+cratio^2) * g(histf(strategy), recentf(strategy)) --stock f = (cratio^2)/(1+cratio^2) * g(histf(strategy), recentf(strategy), daystoexp) --options The value of (cratio^2)/(1+cratio^2) is given below for a number of values of cratio: Code: cratio (cratio^2)/(1+cratio^2) 0.5 0.20 1.0 0.50 2.0 0.80 10.0 0.99 What this means is that at the beginning, with no buffer (cratio=1), I should take 50% of my optimal size (if all other parameters say so). Should I suffer a drawdown of 50% from the starting balance (for any reason), cratio would become 0.50, and my size will only be 20% of normal optimal size. On the other hand, should the account be up 900% at any point in time (cratio=10), I could practically use up the optimal fraction. The above obviously implies that my rate of growth should be smaller at the beginning, but grow more than exponentially as time goes on (provided the edge remains!).
Weekly Update for week 3/50 ended 1/31/2009 Nice week, up 25K (8%). Very good win rate (80%). Lots of trades on SPY as well, with a tendency to exceed my intended size because of averaging up/down: biggest was 8K on Monday, while I lost almost as much on the same SPY earlier. The market see-sawed and ended almost even (open to close). Closed the first position at a loss (8K). Had the good sense to come back immediately in the same direction (LONG) as it appeared I was just shaken out. Glad AMZN came up on Friday, and dished out some gains in stock and options. Code: Opening Balance: 311,156 Net gain for the week 25,279 ------------------------------------------------ Net Balance: 336,435 Number of Trades 24 Number of Profitable Trades 19 Since Inception of Thread 01/10/2009 - 01/31/2009 Opening Balance: 320,064 Net gain (Less Margin Interest) 16,371 (Up 5.1%) ------------------------------------------------ Net Balance 336,435 Number of Trades 70 Number of Profitable Trades 42 Top/Bottom Discretionary Trades for the week TICKER ENTRY DATE/TIME EXIT DATE/TIME QTY PURCHASE AMT SOLD AMT GAIN/LOSS TYPE SPY 2009-01-26-14-05-13 2009-01-26-15-33-30 8000 664230 672500 8245 LONG ZQNBK 2009-01-30-09-36-00 2009-01-30-09-42-05 10000 49956 56690 6563 AMZN CALL --------------------------------------------- CAT 2009-01-26-11-54-47 2009-01-26-13-52-15 8000 257800 254495 -3327 LONG SPY 2009-01-26-12-25-51 2009-01-26-13-51-46 9000 760689 752943 -7771 LONG