Author: kentstutzman

Subject: 1 week hold screen for low 5 last week (CAGR 93)

Date: 12/6/2000

Recommendations: 22

Hello all, In short, the best 5 stock screen yet using Sux`s data is DH70RS13 with a CAGR of 44. If you buy the worst performing 5 T1 stocks last week, hold for a week, then repeat the process, you get a CAGR of 78. The best day of the week for trading was Friday which produced a CAGR of 93. Though this does not include commissions and bid-ask spreads, the return is still pretty inticing. Even if you do not want to do a regular weekly trading screen, this could be one possibility for those creating stop-loss methods and wondering what to do with the money for the rest of the period until you are ready to trade again. This post builds off the 12 week average future return thread I just posted. The considerable outperformance of the DH90 screen for the next 1 week inticed me to look into a new screen that holds for 1 week. Lets see if we can tweak that 1 week return a little. Holding the lowest 10 performing stocks last week (L1-10) returned an average of 1.25% next week. Holding the lowest 5 performing stocks (L1-5) returned 1.60% the following week. Dropping the lowest 1 and holding the next 5 (L2-6) yields 1.72%. Dropping the lowest 2 and holding the next 5 (L3-7) yields 1.62%. Here are the 1, 3, 5 and 10 stock holds for several low screens. L2-11 is sorted by the lowest last week return, dropping the lowest 1 stock, and then start buying from position 2 up. L2-11RS13 starts with that same pool of 10 stocks but sorts those 10 by RS13 and starts buying stocks from the highest RS13 downward. These are all 1 week holds with the purchase price being the closing price the same day as the RS1 week range. For example, the previous week used was Tuesday close to Tuesday close with that same Tuesday close as the purchase price. (I ran for delayed purchase prices later). The below returns are actually the average of the 5 starting days for the 5 days of the week. % % % % GSDs 1 3 5 10 1 3 5 10 Test 5 9 13 12 57 35 28 24 RS13 46 49 45 47 54 37 32 25 L1-30RS13 46 69 78 68 78 32 29 24 L1-10 61 77 79 66 48 35 32 24 L2-11 60 81 72 63 59 38 32 25 L3-12 77 79 69 61 62 40 31 24 L4-13 73 64 61 55 62 40 30 24 L5-14 57 56 55 51 69 33 26 22 L6-15 71 76 72 68 58 35 27 24 L1-10RS13 70 73 68 66 60 36 29 24 L2-11RS13 67 71 66 63 55 33 27 25 L3-12RS13 65 70 64 61 52 36 29 24 L4-13RS13 64 63 58 55 61 35 28 24 L5-14RS13 53 58 54 51 60 37 28 22 L6-15RS13 Clearly no one will run out and start an RS13 1 week hold with a 5 stock CAGR of 13. Returns for L1-30RS13 (alias DH70RS13) are comparable to its 1 month hold cousin but with 3 times the number of trades, this one is not desirable for 1 week holds. As the averages indicated, the best screen for a 5 stock 1 week hold was L2-11 with a CAGR of 79. Actually for 5 stocks, you are buying L2-6. Taking the lowest performing stocks last week and sorting by RS13 (L1-10RS13) did not perform as well as the simpler idea of just sorting by low performance last week. A weekly screen would have about 40 more trades per year per stock held than a 1 month hold screen. With a .25% bid-ask spread and commission penalty, that is 11% less per year relative to a monthly screen. Even with a .5% penalty reducing the result by around 22%, that is still a CAGR of 57. The average CAGR over all days was 79 but there is a significant difference in which trade day is used. For 5 stock holds, Friday close price is the best (CAGR 93) and Wednesday close is the worst (CAGR 66). L2-11 with no lag time between purchase xxx % % % % GSD Day 1 3 5 10 1 3 5 10 Test Fri 45 90 93 70 55 24 25 23 L2-11 Mon 48 79 85 69 42 39 36 27 L2-11 Tue 85 81 75 67 30 33 29 21 L2-11 Wed 83 74 66 56 62 43 39 26 L2-11 Thu 43 60 74 69 49 35 31 25 L2-11 Avg 61 77 79 66 48 35 32 24 L2-11 Are these numbers flukes? At 5 stocks per week, 52 weeks a year for 1984 - 1998, that is 3900 stock purchases which is a large sample size. Incidently, at 5$ per trade, that is a total commission price of $39,000. Ouch! This is not a screen idea for small amounts because the commissions would eat you alive. So you think that my test is not ideal because it uses the previous week closing price as the purchase price and it is impossible to calculate the desired stocks and buy at the closing price. (Recall that the Friday data above has a previous 1 week RS from Friday close to Friday close with the same Friday close purchase.) Well that may be true but here are the results if you wait a day and buy on the close for the following day. For example, the Tuesday row means to calculate the previous week from Tuesday close to Tuesday close, buy on Wednesday close and hold for a week. While this one-day delay is our usual backtesting method, as the next tables show, the returns go down dramatically. The 5 stock hold averaged over all days went from a CAGR of 79 to 65. L2-11 with 1 day lag time before purchase. xxx % % % % GSD Day 1 3 5 10 1 3 5 10 Fri 51 85 88 69 59 30 34 30 Mon 20 47 62 60 36 37 32 27 Tue 45 56 53 53 49 36 33 26 Wed 63 59 56 55 83 43 40 26 Thu 45 53 64 61 47 33 30 22 Avg 45 60 65 60 55 36 34 26 Why do the CAGRs go down so much? The average next week return for L2-6 for all starting days if you buy on the exact closing price is 1.2101. If you buy on the next day`s close, it is .9223. That one day really does matter. So how could you implement this? Though it is only tested for closing prices, it would probably work for Friday noon to Friday noon and purchase as soon as possible on Friday. VL 1 week return wouldn`t work for this purpose because they wait until Friday to report Wednesday close data. You would have to calculate it on your own. So how does weekly trading work for the different months? This next table has L1-5 weekly returns for every purchase day in the month (1650 records in each month). The next column simulates a monthly return by multiplying the weekly return by 4.5. The next 2 columns are average monthly returns for each day in the first week of each month for RS13 and DH70RS13 (375 records for each month). L1-5 weekly screen beats the other two screens in 11 out of 12 months and has an average monthly return more than the other 2 screens combined. Notice the consistent returns for L1-5 throughout all months. Another possible use of this weekly trading is to switch to it during the summer when RS13 really underperforms. xx L1-5 L1-5 RS13 DH30RS13 xx Week * 4.5 Month Month 1 1.560 7.020 7.265 5.735 2 1.268 5.706 2.978 3.648 3 1.072 4.824 1.514 1.893 4 1.436 6.462 4.448 2.963 5 1.169 5.261 3.304 4.416 6 1.139 5.126 0.587 2.455 7 0.862 3.879 0.418 1.539 8 1.263 5.684 -0.281 2.148 9 0.366 1.647 -0.398 2.718 10 1.323 5.954 0.510 2.049 11 1.346 6.057 1.414 1.841 12 1.728 7.776 3.500 2.704 Avg 1.211 5.450 2.105 2.842 Here are the average 1 week returns for each position where 1 is the lowest last week performer. There is a clear inverse relationship between last week and next week performance. It is also apparent why any combination of positions 1-6 provide the best CAGR. ## Average future 1 week return 1 1.108 2 1.191 3 1.162 4 1.320 5 1.275 6 1.102 7 0.811 8 1.006 9 0.836 10 0.897 11 0.908 12 0.815 13 0.965 14 0.496 15 0.708 16 0.678 17 0.677 18 0.458 19 0.775 20 0.729 21 0.531 22 0.524 23 0.721 24 0.485 25 0.551 26 0.611 27 0.563 28 0.368 29 0.403 30 0.530 31 0.446 32 0.424 33 0.385 34 0.301 35 0.342 36 0.485 37 0.487 38 0.466 39 0.460 40 0.402 41 0.321 42 0.379 43 0.455 44 0.345 45 0.387 46 0.493 47 0.409 48 0.336 49 0.406 50 0.365 51 0.272 52 0.383 53 0.234 54 0.288 55 0.242 56 0.313 57 0.444 58 0.274 59 0.294 60 0.398 61 0.267 62 0.233 63 0.215 64 0.094 65 0.252 66 0.284 67 0.159 68 0.264 69 0.253 70 0.244 71 0.194 72 0.226 73 0.333 74 0.242 75 0.027 76 0.167 77 0.237 78 0.141 79 0.144 80 -0.003 81 0.142 82 0.109 83 0.076 84 0.085 85 0.211 86 0.100 87 0.017 88 0.111 89 -0.059 90 0.082 91 -0.155 92 -0.163 93 0.016 94 -0.218 95 -0.522 96 -0.281 97 -0.381 98 -0.274 99 -0.601 100 -0.188 Well that`s it for now. I welcome any thoughts you may have. Kent