| StockFetcher Forums · General Discussion · Range from the highest point and arbitrage of market deficiences | << >>Post Follow-up |
| guspenskiy13 976 posts msg #120317 - Ignore guspenskiy13 |
5/31/2014 6:43:46 PM A long time ago in this thread: http://www.stockfetcher.com/forums/Filter-Exchange/TRO-S-CROCK-POT-2-0/115856/50 I found a very interesting idea. I will quote it here, so it is easier for everyone... Kevin: "....just take a position in both DUST and NUGT at the open and use 3% trailing stop loss on both. The loser stops out quickly while the other runs. Likely to make a net gain on the pair. And it could be done every day. On stop buy at +X% from the open and then put in a trailing stop - nice, dude. Need to work on this one tonight. What I was thinking was that you go place a long bet on whichever stock goes up more than X% from the open, knowing that since they are inversely related only one will do this. Or just buy the one that gains 3% using a set of buy stops. That was the idea from Bartune1, which saves you money." Long story short and how I incorporated this: DUST/NUGT are triple leveraged ETF's with ADR(30)% = 5.2% as of today. They run a lot. You go long of them both - with a decided trailing stop. While one will get stopped out, another one will make money. You make more money on one then you lose on another - that's the idea. Kevin called this "arbitraging market deficiencies". 
This is the screen of NUGT, some explanation - there are two days #1 and #2. Two different occurrences. Day #1 is doji-like on the daily, volatile and directionless. Day #1 is going to bring us a little loss - mostly commissions. Day #2 has direction, momentum and range. While we will lose on NUGT ({OPEN_BUY_PRICE} / {HOD} * {TRAILING_STOP}), we will win more on DUST. Simple. This simple list of actions gives on average 0.4-0.5 percent a day. Not too bad. I started to think about the process - buy both, with trailing on both - one sure loss and one sure win. But if we know that one will be a sure loss - we could possibly exclude that, right? At least that would get rid from commissions... So what IS IT? What action starts the "money-making process" for us in this case? We start making money when one pair exceeds 3% trailing stop and gets closed, while another runs. It is important to understand that this moment has nothing to do with the price percent gain of the instruments, because a) percent gain includes gaps b) trailing stop is calculated from the highest high that was reached after we have sent our trailing stop order. Scenario: NUGT DUST equal dollar amount bought at the open. Trailing stop 3%. GDX goes negative 0.5% but then reverses and closes the day at 2% gain. While we gain 6% on NUGT....we don't lose 2% on DUST...because of the trailing stop. DUST was up 1.5% intra-day and the HOD is the point from where our trailing stop is calculated, which makes us lose 1.545% (100 - 101.5 * 0.97 = 1.545). Overall for the day we would gain 4.455% (6% - 1.545%) So....how do we get rid of the commissions in this case...and buy only one instrument...? To do this we calculate the +3% price from the LOD for the stock. And in the previous scenario we would just buy NUGT when it crosses 3% price level from the LOD (1.545% from the open). And gain the same 4.455% on the day - excluding the commissions and slippage from DUST. To illustrate the difference and the concept using an example of past Friday: 
this is how we would do it with 3% trailing stop on both position. because of the reversion we would probably acquire a loss at the end of the day. 
this is how we would manage DUST using the same concept with 3% buy level from the LOD. A maximum available gain of ~1.7%, so the trailing in this case needs to be optimized and backtested. As you can see, DUST did reverse on Friday....quite violently. If we bought both NUGT/DUST with 3% trailing stop, we would acquire overall loss. If we bought only DUST with an optimized stop-loss, we could gain some money. But again...they reversed...so could we use the same +3% price buy level on NUGT?? 
I think...yes! |
| guspenskiy13 976 posts msg #120319 - Ignore guspenskiy13 |
5/31/2014 9:00:29 PM To summarize, when the security exceeds more than half of its average day range(30) from LOD - the move is strong and the odds are in your favor. Try it and look at different stocks, etfs... I used 3% for NUGT/DUST - they are quite volatile with ADR(30) 5.2%. I think this is just another way to look at the price action. Maybe it's not new. But it gives you a clear entry point IMO. To be backtested soon. |
| Kevin_in_GA 4,599 posts msg #120320 - Ignore Kevin_in_GA |
5/31/2014 10:46:20 PM when the security exceeds more than half of its average day range(30) from LOD - the move is strong and the odds are in your favor. +++++++ Have you calculated these odds, or is that speculation? What I mean is you are making a statement based on very limited data on a few trades of one pair of ETFs. Conceptually I see what you are trying to do, but I would want to see a statistical analysis of at least 50-100 executions of this strategy to see how it actually might behave. |
| guspenskiy13 976 posts msg #120331 - Ignore guspenskiy13 |
6/1/2014 6:02:46 PM Kevin, just got started with Stratasearch. Can't wait to test it. When I've read your initial idea about this - I manually backtested tick by tick NUGT and DUST buy at the open - sell after 2 pm strategy with trailing stop of 3.5%. The return was 23.7% over nine weeks with an average of 2.63% a week, with re-investing. This needs to be optimized and automatically back-tested of course. ____ I really tried to understand my own thinking of this concept - and I guess everything is just based on the ADR. Those 3x gold etf's have 5.2% ADR(30) - so when we enter at 3%, we have a hypothetical "room" for 2.2% gain - on average. There are different types of days in the stock market - if we want to generalize into two categories, that could be "directionless days" and "days with strong direction". With NUGT/DUST, they could be volatile and gain/lose 1-2% on the day going "red to green and backwards" all day, but sometimes they "shoot" for 6-8-10% and more. I actually analyzed 2 years of price data - for how many times each percentage gain at the EOD was: >3% 44.6% >4% 35% >5% 28% >6% 21.5% >7% 14.5% >8% 11% >9% 8% >10% 6% When I am going to back-test this and optimize - I want to pick such percentage level, where the stats to reach higher prices are the best. To be honest, sometimes I get the feeling that all I wrote here is bullshit and nothing new, technically we are just buying at a predefined price level and then hold, basing our decision on the statistical probability of the security to run higher from this price level....which is pretty much TRO's MTC method. Look at the following stats for NUGT "run" over 3 percent: 3% reached: 42 3.25%: 41 3.5%: 39 4%: 31 Can we just enter NUGT at 3% with a stop-limit sell at 4% and a stop-limit sell at 2% OCO order? It gives us a pretty decent risk/reward ratio, doesn't it? 4% profit target has a historical probability of 73.8%....so the chance of a loss would be 26.2%... Risk and reward is the same, so we are pretty much dividing 0.738 by 0.262....and get the RR ratio of 2.8....Not too bad... _______ I got a little bit off-topic, so again... what's the difference of this method compared to MTC? Everything seems the same, we buy at pre-defined price and we use the statistical probability for further gain.... The difference is.... that TRO's MTC method is...."STATIC".... it always calculates the buy/sell levels from the open...whether on dollar amount or on the percentage amount...those levels don't change throughout the day....they remain the same.... My idea of the buying is ..."DYNAMIC".... well...it's dynamic in the way that we use the LOWEST point to calculate the BUY trigger.... And it is changing as long we get new LOWS...or new HIGHS...if we are looking to sell... As I said earlier....I don't understand the concept I am describing fully....but it seems some-what logical.... To finish this post....there is something....creepy....from last Friday....Same example with 3% buy LOD point.... 
An eye-balled percent level by me.... but it is RIGHT AT the 61.8% FIBONACCI RETRACEMENT LEVEL....how crazy is that.... Maybe we should use some fibonacci math in this concept....gotta say hello to Leo Fib.... |
| Kevin_in_GA 4,599 posts msg #120335 - Ignore Kevin_in_GA |
6/1/2014 7:34:53 PM Remember that StrataSearch uses only daily bars, so this type of intraday backtesting will not work - say the stock drops all day from the open, and closes 6% below the open price. SS will look at that in a backtest and say you made a 3% gain if you "bought at 3% above the LOD", even though you never would have entered the trade as you are planning. |
| guspenskiy13 976 posts msg #120336 - Ignore guspenskiy13 modified |
6/1/2014 8:06:23 PM Just figured that out. Do you have any suggestions for a backtester with intra-day data? Tradestation? P.S. Off-topic, but could you post the code for VXX/SPY ratio for SS? I would highly appreciate that. quote: "But you should see they day trading system I have been testing over the last few months ... now that equity curve would have you breathing heavy." Also...I've been going through your old posts...did you share this system anywhere? Thanks! |
| StockFetcher Forums · General Discussion · Range from the highest point and arbitrage of market deficiences | << >>Post Follow-up |