No. of Recommendations: 10
(For reference see
https://web.archive.org/web/20170315000958/http://... .)
Author: VTAlumni Three stars, 500 posts Old School Fool
Number: of 284296
Subject: SI Post Discovery Results (updated with 2010)
Date: 2/21/2011 12:58 PM
No. of Recommendations: 70
Updating my findings from post 223171 last year:
What is Post-Discovery?
These are the results of the screens from when they were first posted (or discovered on this board) to the present. This represents the returns of the screens in real time as they've been published to these boards. Meaning any of us would have actually gotten these results investing in them (of course, trading costs and friction is something you can't account for here, but must assume diminishes these to some degree depending on how much you were to trade with).
Why include only Post-Discovery in these results?
Like everyone else, I use the backtesters to select blends and strategies that have worked well in the past decade or two, but my end check on all screens after filtering down is to check their performance after they were discovered. Screens that pass the test of continuing to beat the market into the future (from discovery) are the ones I feel most comfortable with. This is because I want to avoid selecting screens that were optimized for their backtest returns. When looking back at a 10-15 year period one can sometimes find patterns in the data by tweaking the filters over and over till they see numbers they like. This can often lead to screens that don't perform well into the future as they revert back to the mean.
Where is the data from and how is it gathered?
Every week I copy/paste the SI picks to my program that sorts and stores them. It then uses a cron job to continuously search Yahoo Finance for the stock information and stores it into my database. Since I am collecting the data in real time it is very rare that a data point is missing, so there isn't really a problem with survivorship biases. I found that I end up missing maybe 1 in 200 picks and it can account for an overall degree of error of about +/- 0.1% CAGR. Missing data points are just ignored. So if you selected ranks 1-10 and caught one missing it would treat that week as if it were only selecting 9 stocks. Sometimes screens will only give 5 picks or less, so in these cases the system uses as many as it can find. Weeks where no picks are selected for a screen are treated as moving to cash (for both the screen and index its compared to).
What are the cycles you are referring to?
Cycles are the different starting points of the month for a screen. This analysis assumes a 4 week hold period for all stock picks. However, SI picks are released every week. So you have 4 different starting points. Think of it this way, if you wanted to see how a screen did in 2009 you'd get completely different results if you did a 4 week hold and started on the Jan 5th picks as opposed to the Jan 19th ones. There is some overlap in picks, but the overall CAGR can vary widely between the two in just a year. Over the long hall they tend to average out. I've found it is best to include all of the cycles (the 4 different start weeks) and they analyze and compare them all. If there is a large difference between the results then know you have a pretty unstable screen selection strategy. I tend to focus on the average and then look at the worst case also to get a better feel for the screen. It is kind of like creating 4 years worth of statistics on just 1 years worth of data... though with the overlap in picks week to week you have to be careful really looking at it like that. Still it's way better than just looking at a single cycle where you are missing 75% of the picture.
What are the buy/sell times
My system uses Monday morning price for the buy/sell. This tends to give a much more accurate depiction of what the screen will actually do for someone in real life since we get the picks on the weekends and have the option to purchase Monday mornings. Holidays are ignored/skipped so if a a buy/sell date falls on 12/25, it'll move it to 12/26 automatically.
What is the Index field for
Since all these screens started at different times, I rank them based on their performance against an Index (or the market) over the time frame they are tested. So when you see the start/end dates that is the time frame for both the Screen and the Index. In some cases you'll notice that the Index had different returns for screens with the same time frames. This is because the system sometimes will miss a screens returns. They were either not included or not found for a given month. Some screens won't post picks when the market is really down. For these months, both the index and screen are assumed to be in cash... so if some screen had 10 picks selected one month and another didn't post any picks, the Index for one would be assumed to be in cash and not the other.
Why are some rankings missing?
I've removed from the results the screens: Commodity, Eastwood, GS_Mungo_LowVoom, C1m and Goldbug since they all are no longer posted to our weekly screen listings (all disappeared back in 2005-2007). I've included these this time so people can see them [update for 2011].
The results:
Numbers from left to right are:
Avg (this is the average CAGR of the 4 cycles for that screen)
Idx (this is the return of the Index selected, here I've used S&P500 but I can change this to anything if people are more interested in something else)
High (this is the best CAGR out of the 4 cycles)
Low (this is the worst CAGR out of the 4 cycles)
Avg-Idx (this is the average CAGR of the screen minus the Index CAGR; basically how well the screen did compared to the Index)
Start (this is the first date of screen picks)
Finish (this is the buy of the last date of screen picks... results actually carry over to this date + 4 weeks for its sell)
Post-Discovery Results for SiPRO Screens Ranks 1-25
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
1 Net-Nets_Graham 35.86 3.35 48.46 27.07 32.51 20040816 20110117
2 Commodity 43.00 11.55 45.15 39.07 31.45 20040816 20070305
3 Eastwood 41.07 11.55 43.70 36.84 29.51 20040816 20070305
4 GS_PCF 22.07 -4.31 27.62 17.69 26.38 20070723 20110117
5 Gentle_Screamer 27.68 3.39 29.44 25.00 24.29 20040816 20110117
6 Dipstick 48.01 23.73 56.10 39.57 24.27 20090112 20110117
7 GS_Mungo 21.69 1.09 29.51 14.45 20.60 20051205 20110117
8 GS_Mungo_Voom 21.64 1.09 30.89 15.05 20.55 20051205 20110117
9 78RPM 22.42 3.35 26.67 19.87 19.07 20040816 20110117
10 Advanced 16.46 1.13 23.31 9.56 15.33 20060710 20110117
11 RS-100 15.52 1.11 21.31 11.91 14.41 20060710 20110117
12 Melange 17.36 3.35 19.14 15.98 14.01 20040816 20110117
13 WK_Voom 14.28 0.40 18.91 7.31 13.88 20060424 20110117
14 Up_5% 17.07 3.35 19.27 11.78 13.71 20040816 20110117
15 PIH_Naked 16.88 3.35 20.73 12.90 13.52 20040816 20110117
16 High_Relative_V 16.99 3.48 20.29 14.77 13.51 20040816 20110117
17 BI 14.35 1.19 22.43 4.02 13.16 20060710 20110117
18 Up5X3 13.52 0.44 17.82 11.47 13.08 20060501 20110117
19 GSX 13.40 0.41 24.19 6.06 12.99 20060424 20110117
20 Value_at_the_To 16.24 3.35 18.45 14.17 12.89 20040816 20110117
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
21 ARS 16.28 3.39 17.25 13.92 12.88 20040816 20110117
22 Silver_Parachut 15.41 3.35 17.15 12.96 12.05 20040816 20110117
23 P/S
Love_You 13.85 1.90 17.93 8.89 11.95 20050110 20110117
24 GS_Mungo_LowVoo 11.20 -0.17 30.20 -14.71 11.37 20051205 20060626
25 C1m 16.06 5.68 31.20 -0.19 10.38 20050801 20051128
26 3pt_Relative_Va 11.48 1.15 18.60 5.73 10.34 20060710 20110117
27 POI 12.23 2.28 13.87 11.08 9.95 20050124 20110117
28 Low_Mult 9.61 0.42 20.36 0.34 9.19 20060424 20110117
29 CANSLIM-26 12.58 3.39 18.68 4.95 9.18 20040816 20110117
30 Shrinkage 11.73 3.35 12.33 11.07 8.38 20040816 20110117
31 LowPSvol 5.91 -2.29 10.44 2.56 8.20 20071119 20110117
32 Small_Value 11.49 3.35 13.54 8.46 8.14 20040816 20110117
33 Microcap_Moment 11.14 3.35 14.59 6.53 7.79 20040816 20110117
34 POG 10.69 3.35 13.50 6.89 7.34 20040816 20110117
35 GSX2 6.95 0.46 12.55 1.71 6.49 20060501 20110117
36 FCF-26 9.50 3.35 10.52 8.33 6.15 20040816 20110117
37 Fried_500 6.74 1.07 8.99 4.93 5.66 20051205 20110117
38 Turnarounds 8.98 3.35 11.28 7.25 5.63 20040816 20110117
39 LowPS+ 6.58 1.15 13.53 -2.73 5.44 20060710 20110117
40 HighOnVolume 5.29 0.00 24.74 -3.59 5.29 20110103 20110117
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
41 Incoming_Cash 2.72 -1.50 5.19 -0.48 4.22 20061225 20110117
42 OptiMan 5.58 1.75 6.34 3.94 3.83 20041213 20110117
43 Z26saTA 4.10 0.39 19.19 -3.31 3.71 20060424 20110117
44 Money_Flow -0.16 -3.76 8.70 -20.24 3.60 20071231 20110117
45 Steady_Growth 6.90 3.39 8.02 4.67 3.50 20040816 20110117
46 Quality_Earning 4.86 2.01 7.55 2.38 2.85 20050221 20110117
47 Blue_Skies 20.30 18.17 54.31 3.05 2.13 20100503 20110117
48 Bob 1.77 0.39 3.22 0.36 1.38 20060424 20110117
49 Up5X2 0.00 0.00 0.00 0.00 0.00 20060424
50 LowPSvol5 0.00 0.00 0.00 0.00 0.00 20071112
51 S&P_Peg 3.20 3.35 4.67 2.32 -0.15 20040816 20110117
52 Zweig-26 0.12 3.35 6.09 -4.52 -3.23 20040816 20110117
Post-Discovery Results for SiPRO Screens Ranks 1-10
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
1 Commodity 49.76 11.55 57.10 43.00 38.21 20040816 20070305
2 Net-Nets_Graham 35.23 3.35 56.54 25.54 31.88 20040816 20110117
3 Eastwood 39.01 11.55 48.79 31.90 27.46 20040816 20070305
4 GS_PCF 21.59 -4.31 30.79 16.57 25.91 20070723 20110117
5 Gentle_Screamer 27.64 3.39 35.20 20.99 24.25 20040816 20110117
6 78RPM 25.73 3.35 32.78 17.55 22.38 20040816 20110117
7 Silver_Parachut 25.23 3.35 28.75 18.75 21.88 20040816 20110117
8 Blue_Skies 39.73 18.17 81.41 7.62 21.56 20100503 20110117
9 BI 20.56 1.19 30.36 9.83 19.37 20060710 20110117
10 RS-100 20.43 1.11 31.48 12.75 19.32 20060710 20110117
11 GS_Mungo_Voom 19.94 1.09 29.04 12.55 18.85 20051205 20110117
12 GS_Mungo 19.86 1.09 26.61 14.14 18.77 20051205 20110117
13 High_Relative_V 21.97 3.48 23.07 20.73 18.49 20040816 20110117
14 Dipstick 42.11 23.73 45.17 40.73 18.38 20090112 20110117
15 C1m 22.44 5.68 41.36 -3.93 16.76 20050801 20051128
16 WK_Voom 17.00 0.40 25.64 9.43 16.60 20060424 20110117
17 Advanced 17.37 1.13 29.16 7.27 16.24 20060710 20110117
18 P/SLove_You 17.28 1.90 21.05 10.95 15.38 20050110 20110117
19 Low_Mult 15.75 0.42 28.03 1.52 15.33 20060424 20110117
20 Up_5% 17.96 3.35 22.04 12.65 14.61 20040816 20110117
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
21 Melange 17.51 3.35 24.31 13.07 14.15 20040816 20110117
22 Value_at_the_To 16.93 3.35 20.94 15.52 13.57 20040816 20110117
23 HighOnVolume 13.26 0.00 69.25 -16.22 13.26 20110103 20110117
24 GSX 13.65 0.41 22.65 7.85 13.24 20060424 20110117
25 Turnarounds 16.37 3.35 19.38 14.18 13.02 20040816 20110117
26 Up5X3 13.10 0.44 18.44 7.54 12.66 20060501 20110117
27 Small_Value 15.80 3.35 18.63 11.30 12.45 20040816 20110117
28 PIH_Naked 15.63 3.35 21.30 11.15 12.28 20040816 20110117
29 ARS 15.17 3.39 17.57 12.45 11.78 20040816 20110117
30 3pt_Relative_Va 12.07 1.15 16.95 6.63 10.92 20060710 20110117
31 CANSLIM-26 13.80 3.39 20.70 7.83 10.40 20040816 20110117
32 POI 12.57 2.28 16.14 6.25 10.29 20050124 20110117
33 Microcap_Moment 13.63 3.35 24.36 6.65 10.28 20040816 20110117
34 FCF-26 13.57 3.35 15.67 7.74 10.22 20040816 20110117
35 LowPSvol 5.82 -2.29 13.98 -8.30 8.10 20071119 20110117
36 Quality_Earning 9.88 2.01 14.13 8.03 7.86 20050221 20110117
37 LowPS+ 8.04 1.15 14.36 -2.62 6.90 20060710 20110117
38 Incoming_Cash 5.30 -1.50 12.30 -0.38 6.79 20061225 20110117
39 POG 10.06 3.35 16.09 6.73 6.71 20040816 20110117
40 Fried_500 7.63 1.07 11.78 4.02 6.55 20051205 20110117
Rank Screen(?)(?) Avg Idx High Low Avg-Idx Start Finish
41 Shrinkage 9.88 3.35 11.65 5.89 6.53 20040816 20110117
42 GS_Mungo_LowVoo 6.27 -0.17 10.30 -4.46 6.44 20051205 20060626
43 GSX2 6.17 0.46 8.75 4.93 5.70 20060501 20110117
44 Steady_Growth 6.87 3.39 7.73 5.34 3.48 20040816 20110117
45 OptiMan 4.45 1.75 7.07 2.35 2.70 20041213 20110117
46 Z26saTA 2.60 0.39 17.03 -4.48 2.21 20060424 20110117
47 S&P_Peg 5.14 3.35 9.27 3.04 1.79 20040816 20110117
48 Money_Flow -2.00 -3.76 8.01 -24.16 1.76 20071231 20110117
49 Bob 1.68 0.39 2.54 0.28 1.29 20060424 20110117
50 Up5X2 0.00 0.00 0.00 0.00 0.00 20060424
51 LowPSvol5 0.00 0.00 0.00 0.00 0.00 20071112
52 Zweig-26 0.80 3.35 7.29 -3.18 -2.55 20040816 20110117
Conclusions: appears MI is still working :)
-Eric
