HOW TO DESIGN A SYSTEM (NOT JUST A FILTER) - StockFetcher.com Stock Screener

StockFetcher Forums · Filter Exchange · HOW TO DESIGN A SYSTEM (NOT JUST A FILTER)

<< 1 ... 36 37 38 39 40 ... 43 >>Post Follow-up

jaybird0730 1 posts msg #112670 - Ignore jaybird0730	4/5/2013 2:09:47 PM Kevin, I wanted to add my voice to the chorus of thankful reviewers who have read this system and your excellent rationale describing the business of setting one up. In addition to the clear explanation and rationale for this system, this system contains an element that most other systems lack: an �edge�, the intrinsic statistical advantage that followed over time, allows the system (and the trader) to win. I�ve back tested the system with various start and stop dates for multiyear periods (using minimum 5 yrs in each test), and SF back testing consistently outputs about a 60/30 win loss ratio, with about 30% CAGR. This is based on your original system parameters: 5 open positions max, and using only $20k per trade, with a max hold time of 20 days (I know you had modified this to max hold = 15 later in the thread.) In essence the system is simple, and this is key in my opinion. For any two comparative systems with equal historical returns, the simpler system is naturally the one to follow, as it contains less risk (fewer chances to make a calculation mistake on trigger). I don�t think people should conflate the mathematical language of Bollinger Bands and Z Scores as the reason for any perceived complexity in the system. It appears you�ve simply chosen to express a relationship you found in the market using the precise dexterity of mathematics. The fact that you've modeled the Monte Carlo input variables and have also published these results and rationale is an added bonus that may be underappreciated. While this does place a burden on those of us without a strong math skill set, it should not obstruct the core truth: that the system itself is fairly straightforward. I�m currently reading some of your other work that has been posted in SF and I look forward to benefiting from that posted. If these could be propagated to the masses effectively and implemented, the potential world of financial good it might do is awe-inspiring. To that end, I would encourage you regarding enhancement plans on the website you�ve established. I would further consider that the content you�ve generated perhaps be compiled into a book form as well so that it is easily referenced in a self-contained location. Perhaps SF might assist in this regard by devoting a section to system building, using your examples as the feedstock content material. The threaded discussions in this forum are valuable in providing a dynamic aspect to the conversation, but having content in consolidated form would be of persistent value and for maximum good to those us of individuals seeking to benefit, and I sense that you are motivated in this regard as well. Respectfully, J
Kevin_in_GA 4,599 posts msg #112674 - Ignore Kevin_in_GA	4/5/2013 7:11:59 PM J - thanks. The core idea here is also used in my Pangolin-Z system over at Collective2, but improved upon quite significantly such that the profit seen over the last five years is higher, as is the percent of trades profitable and the Sharpe Ratio.
amtmail 34 posts msg #113158 - Ignore amtmail	5/6/2013 4:14:30 PM Thank You Kevin do you feel that this filter still working now ? if not ,do you have any modifications or changes for it
Kevin_in_GA 4,599 posts msg #113159 - Ignore Kevin_in_GA	5/6/2013 5:11:58 PM It still works well - here is a backtest (5 positions max) since 1/2/2013: There were 55 total stocks entered. Of those, 50 or 90.91% were complete and 5 or 9.09% were open. Of the 50 completed trades, 35 trades or 70.00% resulted in a net gain. Your average net change for completed trades was: 0.86%. The average draw down of your approach was: -2.67%. The average max profit of your approach was: 2.54% The Reward/Risk ratio for this approach is: 2.17 Annualized Return on Investment (ROI): 40.45%, the ROI of ^SPX was: 35.76%. Equity Summary Starting Account Value: $100,000.00 Fees per trade: $8.95 Ending Account Value: $109,240.19. Cash: $-0.03 Market value of holdings at end: $109,240.22 (paid: $107,379.02) Realized gain/loss: $7,871.24 ($157.42 per trade.) Total Commissions Paid: $939.75 Unrealized gain/loss: $1,861.20. But here is the modified version (Pangolin Z): There were 58 total stocks entered. Of those, 53 or 91.38% were complete and 5 or 8.62% were open. Of the 53 completed trades, 40 trades or 75.47% resulted in a net gain. Your average net change for completed trades was: 1.86%. The average draw down of your approach was: -2.56%. The average max profit of your approach was: 2.56% The Reward/Risk ratio for this approach is: 4.16 Annualized Return on Investment (ROI): 82.74%, the ROI of ^SPX was: 35.76%. Equity Summary Starting Account Value: $100,000.00 Fees per trade: $8.95 Ending Account Value: $120,819.16. Cash: $0.22 Market value of holdings at end: $120,818.94 (paid: $120,316.46) Realized gain/loss: $20,835.78 ($393.13 per trade.) Total Commissions Paid: $993.45 Unrealized gain/loss: $502.48 More then 2.5 times the profit, higher win %, and much better Reward/Risk ratio.
fortyfour 189 posts msg #113161 - Ignore fortyfour modified	5/6/2013 5:55:51 PM Kevin, Does Pangolin Z drawdown more in the same drawdown periods as your original. Maybe you already answered that with " higher Sharpe ratio too" but just want to be sure. Thanks.
Kevin_in_GA 4,599 posts msg #113194 - Ignore Kevin_in_GA	5/7/2013 10:24:37 PM Original Z-score system - Performance for period 1/2/2007 through 4/1/2013 Max Drawdown - 17.15% Sharpe Ratio - 0.9103 Pangolin Z - Performance for period 1/2/2007 through 4/1/2013 Max Drawdown - 11.77% Sharpe Ratio - 1.748 This period encompasses the severe 50+% drawdown seen in the S&P500.
tomm1111 202 posts msg #114749 - Ignore tomm1111 modified	8/4/2013 10:16:10 PM Kevin, I've visited your systems on the collective2 web page. Impressive indeed. I have a question. 1. When you optimize this system, specifically what was the in-sample and out-of-sample data time frames used? I�m potential subscriber and I'm trying to understand curve fit risk. Thanks, tomm1111
Kevin_in_GA 4,599 posts msg #114750 - Ignore Kevin_in_GA	8/4/2013 10:39:24 PM Kevin, I've visited your systems on the collective2 web page. Impressive indeed. I have a question. 1. When you optimize this system, specifically what was the in-sample and out-of-sample data time frames used? I�m potential subscriber and I'm trying to understand curve fit risk. Thanks, tomm1111 ++++++++++++ 1. The in-sample data was 1/1/2008 until 12/31/2012 (or thereabouts). The goal was to include a period of protracted drawdown into the design - this is especially important for a long-only system like Pangolin Z. No out-of-sample data was used. I started trading it live on C2 on 3/26/2013, and so far it is up 17%. I would suggest looking at www.statisticalinvesting.com to see the backtest data and results.
tomm1111 202 posts msg #114751 - Ignore tomm1111 modified	8/4/2013 11:10:47 PM Kevin, Thanks for the explanation. Your website answered the question I had. tomm1111
tonyctl 7 posts msg #114907 - Ignore tonyctl	8/16/2013 3:57:11 AM New to SF here and would like to know when I am backtesting this system and have Max. trades per day set to 2, will the backtest automatically select the first 2 stocks with the lowest Z scores to purchase? Also, does max. trades per day mean max number of purchases in one day or combined buy and sell orders?

StockFetcher Forums · Filter Exchange · HOW TO DESIGN A SYSTEM (NOT JUST A FILTER)

<< 1 ... 36 37 38 39 40 ... 43 >>Post Follow-up