|Current location||Thread information|
OmniVest User Forums
General Discussion Topics
ATM And The Value Of Ranking
Last Activity 4/15/2019 7:43 PM
4 replies, 1141 viewings
|Printer friendly version|
Location: Sleepy Hollow, IL
I’ve always felt that the ability to rank the order of trades taken based on the performance of a particular strategy in the past would be beneficial.
That may still be true, but ATM ranking involves the ranking trades according to the recent characteristics of the underlying symbol.
This is a new capability, so I spent some time investigating the ATM Ranking function.
I used the following to obtain what I felt would be a large, unbiased data set of trades for the test;
1) Symbol list: Nirvana Optionable Stocks list filtered for C>=5. This gave me slightly less than 4,000 stocks.
2) A variety of strategies, both RTM and trending;
Running the To Do list with this many symbols and strategies took a number of hours, but I now have an excellent data set for this and future testing.
In Port Sim, I used 2X margin and set the trade size to 1% of equity to enable as many trades as possible.
Because of the large number of trades generated, I had to limit the analysis date range to 1/1/2007 to present - which still gave me 58,441 trades to work with.
This is a table of the pertinent results;
1) Base run with 100 trades allowed and no ranking: 58,441 trades.
2) The same settings, but with the # of trades allowed reduced to 4 for all 3 Market States.
3) Identical to #2, but with Ranking enabled using Nirvana’s VTY_PRICE(10,5) descending for all 3 Market States.
The statistics show a definite improvement when enabling the ranking function (#2 vs #3);
A) Similar # of Trades
B) Increased ROI
C) Increased Calmar
D) Lower Avg % Inv
E) Increased ratio of ROI / # Of Trades
This satisfied me that Ranking based on the attributes of the underlying symbol was very beneficial.
I then ran the Port Sim using the list with 4,000 stocks and a selection of Strategies for each Market State that I developed previously with these settings;
1X (no margin)
4% Of Equity for trades
4 Long Trades allowed in each of the 3 Market States
Ranking = VTY_PRICE(10,5) decreasing for all 3 Market States
The difference in the results when enabling the Ranking Function was even more impressive than I expected,
and clearly demonstrates the value of ranking;
Previously, we had no way to measure the effect of the current characteristics of the underlying symbols on the outcome of the trades generated.
Now we do - and the effect is much larger and more positive than I expected.
With ATM, we can take advantage of this and certainly improve the quality of our trade selection.
Once again, I’m thoroughly impressed by the capabilities of ATM - it’s definitely a winner in the collection of tools Nirvana’s provided over the years.
I’m looking forward to Ed’s webinar this afternoon sharing more details on how to use ATM.
[Edited by Mark Holstius on 2/1/2018 11:34 AM]
Attached file : 00 Strategies Test list.jpg (68KB - 797 downloads)
Attached file : 01 Ranking Results.jpg (130KB - 734 downloads)
Attached file : 02 Ranking Comparison.jpg (1248KB - 832 downloads)
Location: L'ville, GA
I strongly agree that the Ranking feature from ATS is the most valuable aspect of ATM, at least for now. I’ve sent Ed a list of potential future enhancements, many of which I think will offer the same degree of benefit.
“Onwards and upwards, gee whiz!”
Location: L'ville, GA
Here are modified versions of Ed's Ranking formulae that I sent him during the webinar.
Mod's assure that TRII and RSI curves are not sloping downwards ... also provides bearish market ranking for shorts ...
TRII(60,40) * iif( LnReg_Slope(TRII(60,40),3) >= 0, 1, 0)
... for Longs or Shorts ...
RSI(14) * iif( LnReg_Slope(RSI(14),3) >= 0, 1, 0)
... that's for Longs-only (ie bullish market) ...
(100-RSI(14)) * iif( LnReg_Slope(RSI(14),5) <= 0, 1, 0)
... that's for Shorts-only (ie bearish market) ...
... haven't had time to test this, but it should be a very useful refinement
[Edited by Jim Dean on 2/1/2018 3:25 PM]
I'm not sure I'm following your notation. Were you intending these formula to be in the Trade Filtering tab of ATM or the Trade Ranking?
Maybe a light bulb just went on -- the formula is saying multiply the TRII value by either 1 or 0 depending upon if the value of the LnReg_Slope of the TRII is >=0. That's a cool way of introducing logic to either include or exclude the value.
If the above is really what this is doing, were you thinking that these might be in addition to the Momentum/Movement (Raging Bull); Pullback/Trend (Bear Market); or Ranker1/Ranker2 (Default)?
I know it was some quick thinking during the webinar that resulted in these, but I'd love to do some testing and post some results.
Location: L'ville, GA
You figured it out. Sorry about the hurried iPhone mistakes during the seminar
The first one is an improvement of Ranker1 (Trii). Valid for long or short. (Maybe change the ,3 to a slightly bigger number (4 or 5) for long. And maybe change the ,3 to 2 for short. Smaller values = less lag.
The second one points out a “fault” (imho) of RSI ranker2 (the supplied one). Needs to be different for long and short. So, here is what I’d suggest:
Replace the default with defaulL and defaultS … assign filters and strats etc for long only and short only respectively … use TRii rule for both, but different RSI rules respectively.
[Edited by Jim Dean on 2/1/2018 6:20 PM]
E-Mail this thread to a friend
(un)/Freeze thread |
|Toggle e-mail notification|