|
John W
 Elite
    Posts: 654
Joined: 10/11/2012
Location: Sydney, NSW, Australia
User Profile |
Is it possible to expand the V2 Account > Edit Settings > "Alphabetical" to":
"Alphabetical Ascending"
"Alphabetical Descending"
and "Alphabetical Random".
Why? I listed my past theoretical trades alphabetically and noticed that OV is less likely to take trades for symbols lower down in the alphabet.
IF my surmise is true for others also, then it's likely that symbols lower in the alphabet are not being substantially traded by the total sum of OV users, symbols higher in the alphabet are being substantially traded.
Allowing the user to chose from Symbols that are Alphabetically Ascending, Descending or Random allows the sum total of all OV trades more likelihood to split across ALL symbols, reducing slippage and fill risks for all.
John

[Edited by John W on 2/27/2014 9:40 PM]
Attached file : V2 Edit Settings.png (150KB - 395 downloads)
|
|
Jim Dean
 Elite
   Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
I agree with John that some amount of randomization would be useful.
Here is another way to do it - offer an option to rank (sort) the symbols according to the selected criteria (including alpha up/dn, price up/dn, vol, p*v) … then BIN them into five (if small list) or ten (if big list) equal group sizes … then SHUFFLE the order randomly within each bin. Use the customers ID number as a seed for the shuffle so all are different.
Result will have the effect of the desired sort but will also contain enough differences so that "piling on" of identical trades between diff customers will be reduced.
Btw, offering ascending sort order for V or P*V … seems to be a recipe for disaster if someone does not ALSO provide a fairly high volume filter in the Accounts settings - please add a check of some sort to the OVest entry logic about that, if you keep those two ascending order options in there. Who wants to trade a notably ILLiquid stock?
[Edited by Jim Dean on 2/28/2014 5:05 AM]
|
|
GordonG
 Regular
   Posts: 76
Joined: 10/11/2012
Location: Sydney, Australia
User Profile |
Have been agitating for a few of these features as well via "next quantum leap". Have wanted to see monthly stats on accounts for ages and wrote my own to get around it via automated data extraction - covered elsewhere. Random selecxtion would surely be preferable to skewing to top of alphabetical list, but remember "randomness" has to be repeatable for backtesting. My alternative suggestion which I'm getting more keen on is to have a way to at least set a priority over the top of these options. So I'd want candidates to be taken from one or more strategies before being taken say from another less successful (statistically) strategy. Perhaps that's an indicator to a more immediate option, perhaps select between candidates by which strategy it comes from, taking the statistically most successful above another, and then if from same strategy take "random". So a candiate ranking/priority setting is needed or a way to effect the same. Another option I'd like to think worth trying is to use ARM metrics to help select the best candidate. Surely we can at least attempt a few more "smarter" solutions... whether they improve things or not we'll only know once we try them.
|
|
Diamondjag
 Veteran
 Posts: 123
Joined: 10/11/2012
Location: Brighton, Co.
User Profile |
I think this is the same issue that Steve Mayo is trying to get Nirvana to address....some type of sorting that has some real meaning to results. Borrowed from another thread.....
"These options all still suffer from a bias effect (ie. biasing selection toward low-alpha, high/low price, high/low volume stocks.) Any chance of getting a sorting routine that uses some measure of predicted performance, such as the advisor rating for the signaled trade? Even better, I though you guys were working on something akin to DSS???
If you can't sort using some historic signal performance measure, in terms of simply eliminating the selection bias, adding the option of a truly random trade selection might be a good option to add to this list. That combined with the "Allow settings to reduce" option would help make the equity curve and backtest metrics more statistically valid." Steve Mayo
|
|
Jim Dean
 Elite
   Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
This is a tough one ... there are two "conflicting" needs here:
1. make wise selections when not enough buying power to take them all
2. make sure that a bunch of OV users don't "pile on" to one obscure symbol
I seem to recall that Ed said that they've got #2 covered through some other means, by "randomly staggering" the order times for different users ... not sure that he said those words but that's the gist that I recall.
If, somehow, #2 is dealt with, then the usefulness of randomization drops off significantly ...
|