Index  Recent Threads  Who's Online  Search 

New Thread 

judgetrade

Here are the corrected statistics with variable slippage ... aaa.png (202534 bytes) (Download count: 272) 

Edit 1 times,
last edit by
judgetrade
at Nov 18, 2019 2:17:07 AM

mgerstein

Hi Andreras, I do not understand your (ab)/abs(b) formulation, sorry but my math is terrible. Could you give me the exact formula I should use? That would be great, Thank you!!! This is a formulation we use throughout p123 when one of the items in a %change calculation might be negative . . . the crossingzero problem. Suppose I want to compute the % change in EPS fro, b (1.00) to a (1.20). Most folks would do it as a/b or 1.20/1.00 which produces 1.2. The make it appear as a %, we'd actually do ((a/b)1)*100. But suppose the prior (b) figure was minus 1.00. In that case, the standard formulation would give us 1.2/1 which would ultimatley wind up looking like a negative 20% growth rate, the exact opposite of what we want. The alternative crossingzero formulation would be this: If a and b are 1.20 and 1.00 respectively, then its (1.201.00)/1.00 which is .20/ To express as a % its ((ab)/abs(b))*100. Suppose the b figure is 1.00. Now, it's 1.20  1.00/abs(1.00) which is .20/1.00 = .20. It also works if the numerator figure is negative, or if both figures are negative. “What you have here is a variation of that I’ve been noticing — investing based on the behavior/emotions of machines and those who program them.” That is kind of the Idea of all my models. As Quants, at least this is my definition, we hack emotions of others especially in turnaround situations where the second derivative (e.g. rate of change) happens and market participants need to catch up (and try to be faster then the others). E.g. I do not care so much about earnings, but for example the very recent rate of change earnings. That's fine. And as you've seen in your work, it can be quite profitable. My quibble is only with the labels. That sort of thing can't be called value or quality because it isn't. Among traditionally labeled styles, the one that works best is momentum, or a different label can be created. When we work only for ourselves, the label is irrelevant. It only becomes an issue when we want to communicate what we're doing to others who may or may not see the formulations we're using. I tested longer term “standard Qualtiy stuff” (like 5 Year earnings consistency), but got ridd of it a bc. Rate of change (“fundamental momentum”) seems to give better results. When you are looking for rate of change  something that can, as you know, be very well worth looking for  we stepped outside the realm of Quality and moved over to growthmomentum. The thing is, I think I can explain that by emotions of market participants emotions (my assumption could be wrong), that chase that fundamental momentum (which is kind of backed by academic papers), so I give it a try. I know that is risky, bc. I do not have a 100% or even 50% knowledge about market participant emotions, but my “intuition” tells me, it’s worth the risk. I would still define it as quality, but in rate of change terms, but you are right its kind my own definition. Again, we're back to labels. Quality can be but is not necessarily about "better results." Its about a mixture of results (return) and risk. and moving up in Quality can be and often is justified in the face of lower returns but also lower risk. As I said, personal definitions are fine when we keep our work private. But when we share our ideas with others, we need to speak in terms of commonlyaccepted definitions lest we cause confusion and misunderstanding and wind up with the proverbial Tower of Babel. And its especially important in areas like designer models. I've long complained that most models are promoted/marketed horribly. You're very strong in momentum, traditional and alternative (i.e fundamental momentum) and working with related behavioral phenomenon. Own it. Articulate it, Make it yours. That's how sellers of investmentrelated intellectual property market/promote themselves. Marc Gerstein Director of Research, Chaikin Analytics Blogs: https://actiquant.com, https://portfoliowise.com/portfoliowiseblog/ , https://www.chaikinanalytics.com/blog/ Twitter: @MHGerstein I predict the future, as soon as it becomes the past 


kumar

Geo, Warran Buffet says looking at the price is gambling not investment, ETN SVXY lost 90% in one day. Why we not able to extract performance every year by Graham and Buffet value investing. We have all required factors and formula for value investing. Buffet says take the market sell off to buy stocks not to sell. Any successful strategy to buy 52 weeks low and sell on 52 weeks high. In the chart and past data it looks simple and straightforward and how to achieve same in p123 and in out of sample. That is question of thought. Thanks Kumar 478E996EA29C41E1A56AD9ED207FF4DB.png (788643 bytes) (Download count: 228) Warren Buffett — 'Risk comes from not knowing what you're doing' 'Should you find yourself in a chronically leaking boat, energy devoted to changing vessels is likely to be more productive than energy devoted to patching leaks' 

Edit 2 times,
last edit by
kumar
at Nov 18, 2019 2:11:19 PM

geov

Andreas, The max D/D of your model = 25.46%. From my experience you can only get such a good D/D with market timing. Please confirm, and if possible let us know what you use. Thanks, Georg 


geov

Geo, Warran Buffet says looking at the price is gambling not investment, ETN SVXY lost 90% in one day. Kumar Kumar that is the reason why the model always holds 2 positions. Also ProShares Short VIX ShortTerm Futures ETF seeks daily investment results, before fees and expenses, that correspond to onehalf the inverse (0.5x) of the daily performance of the S&P 500 VIX ShortTerm Futures Index. It used to be 1.0x. So the risk is less now. If they had implemented this from the start one would "only" have lost 45% in one day. 


geov

Although I prefer ETF models, it is not too difficult to design a smallmedcap model with annualized return > 50% and reasonable turnover and drawdown. Here is performance from Jan2004 with variable slippage, without and with market timing. The buy rules are based to a large degree on Yuval's recently posted screen rules. SmallMed Caps no mkt timing.png (151152 bytes) (Download count: 211) SmallMed Caps with mkt timing.png (152971 bytes) (Download count: 212) 

Edit 2 times,
last edit by
geov
at Nov 18, 2019 6:04:00 PM

Testious

Hi Andreras, I do not understand your (ab)/abs(b) formulation, sorry but my math is terrible. Could you give me the exact formula I should use? That would be great, Thank you!!! This is a formulation we use throughout p123 when one of the items in a %change calculation might be negative . . . the crossingzero problem. Suppose I want to compute the % change in EPS fro, b (1.00) to a (1.20). Most folks would do it as a/b or 1.20/1.00 which produces 1.2. The make it appear as a %, we'd actually do ((a/b)1)*100. But suppose the prior (b) figure was minus 1.00. In that case, the standard formulation would give us 1.2/1 which would ultimatley wind up looking like a negative 20% growth rate, the exact opposite of what we want. The alternative crossingzero formulation would be this: If a and b are 1.20 and 1.00 respectively, then its (1.201.00)/1.00 which is .20/ To express as a % its ((ab)/abs(b))*100. Suppose the b figure is 1.00. Now, it's 1.20  1.00/abs(1.00) which is .20/1.00 = .20. It also works if the numerator figure is negative, or if both figures are negative. “What you have here is a variation of that I’ve been noticing — investing based on the behavior/emotions of machines and those who program them.” That is kind of the Idea of all my models. As Quants, at least this is my definition, we hack emotions of others especially in turnaround situations where the second derivative (e.g. rate of change) happens and market participants need to catch up (and try to be faster then the others). E.g. I do not care so much about earnings, but for example the very recent rate of change earnings. That's fine. And as you've seen in your work, it can be quite profitable. My quibble is only with the labels. That sort of thing can't be called value or quality because it isn't. Among traditionally labeled styles, the one that works best is momentum, or a different label can be created. When we work only for ourselves, the label is irrelevant. It only becomes an issue when we want to communicate what we're doing to others who may or may not see the formulations we're using. I tested longer term “standard Qualtiy stuff” (like 5 Year earnings consistency), but got ridd of it a bc. Rate of change (“fundamental momentum”) seems to give better results. When you are looking for rate of change  something that can, as you know, be very well worth looking for  we stepped outside the realm of Quality and moved over to growthmomentum. The thing is, I think I can explain that by emotions of market participants emotions (my assumption could be wrong), that chase that fundamental momentum (which is kind of backed by academic papers), so I give it a try. I know that is risky, bc. I do not have a 100% or even 50% knowledge about market participant emotions, but my “intuition” tells me, it’s worth the risk. I would still define it as quality, but in rate of change terms, but you are right its kind my own definition. Again, we're back to labels. Quality can be but is not necessarily about "better results." Its about a mixture of results (return) and risk. and moving up in Quality can be and often is justified in the face of lower returns but also lower risk. As I said, personal definitions are fine when we keep our work private. But when we share our ideas with others, we need to speak in terms of commonlyaccepted definitions lest we cause confusion and misunderstanding and wind up with the proverbial Tower of Babel. And its especially important in areas like designer models. I've long complained that most models are promoted/marketed horribly. You're very strong in momentum, traditional and alternative (i.e fundamental momentum) and working with related behavioral phenomenon. Own it. Articulate it, Make it yours. That's how sellers of investmentrelated intellectual property market/promote themselves. Mark, you said:"Suppose the b figure is 1.00. Now, it's 1.20  1.00/abs(1.00) which is .20/1.00 = .20." No,in this case the result is (1.20  1.00)/abs(1)=2.2/abs(1)=2.2 (220%).........not 0.2. Am I wrong? Anyway, despite what you said, which is terribly right, the formula a/b used by Judgtrade genarally (and I do not know why) gives a better return. 

Edit 1 times,
last edit by
Testious
at Nov 21, 2019 3:51:45 PM

judgetrade

Hi Mark, understood! (correct labeling!). Thank you for your hints! Best Regards Andreas 


judgetrade

Hi Georg, yes, market timing is included: Buy rule: (close(0,#spepscy)>ema(75,0,#spepscy)) or (close(0,#bench)>ema(75,0,#bench)) Sell rule: (close(0,#spepscy)<ema(75,0,#spepscy)) and (close(0,#bench)<ema(200,0,#bench)) My take: its very loose, e.g. it will not protect you from a 20% DD, but from a 2008 scenario. It was out of the market in 2015 very shortly, otherwise it hold me in the market basically since 2011 which was fine so far. Also: the SP500 Earnings are estimated by a ton of analysts, the whole market looks at that number, so they might be wrong for prolonged time, but they correct regulary. I invested a ton of time in macro timing the market and I have to say that this simple formula (combined with a good ranking system) is really hard to beat. And from what I have learned its really hard to be protected from DDs that are in the 1025% range, so I simply teached myself to withstand the vola and think more in terms of total return instead of so called "volatility adjusted" return. Also I regulary have a look at your work at (pops up every day at my screen when I open my browser!!!): https://imarketsignals.com/bci/ Best Regards Andreas 

Edit 1 times,
last edit by
judgetrade
at Nov 22, 2019 5:22:21 AM

