| Pages:
1
2
3
4
..
17 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
thats good idea, and or put it in red or something.
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
To be able to select beforehand what kind of mathematical symbols GSB is allowed to use to build strategies.
*
/
+
-
When a strategy only contains * and or / the 3 weight factors and 1 entrylevel can be replaced by only 1 factor.
This simplifies the strategy code significantely.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
- has been removed a few builds ago. Its redundant due to wright being + or -.
with * we are going to simplify it so there is only one weight, not two.
Im not confinced its a good idea to clip GSB's wings so much, but maybe a weight in the fitness for complexity.
If we have a *, complexity goes down. This is going to get more important when we have more secondary filters / truisms.
|
|
|
rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
a trailing stoploss is badly needed.
|
|
|
boosted
Junior Member
Posts: 73
Registered: 16-6-2017
Member Is Offline
Mood: No Mood
|
|
I don't have access to the demo anymore, but here is a list of wishes I like to see in GSB (I will add others later as I remember them):
1) Incorporate options to back-test stocks and ETF's (i.e. Starting Equity, Share size (limits and range)
- This is a big one for me since I mostly trade stocks and ETF's
2) Fill limit order when limit price is exceeded.
- This option can be done in TS AFTER GSB build, but better if included in GSB so we know the results in GSB are solid and fills are not
based on unrealistic hi and
low fills
3) More order types (i.e. Enter on limit stops, Exit at target price, trailing stops, etc.)
4) Add PROM fitness function for robustness. One more can't hurt. Net Profit is nice, but robustness fitness criteria never hurts.
5) Allow 3rd party indicators to be used.
6) Price pattern recognition.
7) Avg. Entry and Avg. Exit Efficiency "truism", condition or fitness.
- This way the build process can put weight into finding best entry and/or exit efficiency along with other build factors.
8) Allow custom fitness equations.
9) Input for min. to max. entries per day.
10) Trade Entry and Exit After time inputs.
11) Report Commission and slippage somewhere in trade back-testing results.
-Always nice to know for sure when looking at the report the $ amount of commission and slippage being reported for the back-tested
run, like in TS.
I am making this wish list from memory since my demo time has run out. The last Beta I used was 28.1. I will add more wish list items as I can
remember or think of them.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Boosted,
thanks for the comments.
1 & 5 are on the short term to do list. TS<> took over 2 weeks of programmers time, so other things didnt get done in the time expected.
2&3) Currently there are no limit orders used in GSB, but it will happen in time.
4,11) This can be done.
9) Not hard too do, though I rarely get systems lots per day.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Boosted,
for stocks was there another fitness you needed? ie % profit per year etc.
We are working on the stocks very soon, so want to add any fitness at the same time as prom.
Currently you can use all sorts of combination of fitness.
it np*pf*aver trade* Pearsons^10 /drawdown etc.
|
|
|
rws
Member
Posts: 114
Registered: 12-6-2017
Member Is Offline
Mood: No Mood
|
|
I found that when I add /drawdown to the fitness formula, building
systems get many times slower. Not 10 times but almost 100 times
there should be something possible like 1/(1-pearsons) in the fitness formula.
Quote: Originally posted by admin  | Hi Boosted,
for stocks was there another fitness you needed? ie % profit per year etc.
We are working on the stocks very soon, so want to add any fitness at the same time as prom.
Currently you can use all sorts of combination of fitness.
it np*pf*aver trade* Pearsons^10 /drawdown etc.
|
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
There is pearsons option in the fitness calculations, but my own testing in ewfo showed it was not good to use it.
NP * ave trade and NP * pf were generally the best.
I will test the speed of dd metric, but again dont think its the best fitness for building systems
|
|
|
boosted
Junior Member
Posts: 73
Registered: 16-6-2017
Member Is Offline
Mood: No Mood
|
|
I can't think of any particular fitness functions off hand that I would specifically want added for stocks. PROM would be the only one right now
that I think of that would be a nice addition.
|
|
|
boosted
Junior Member
Posts: 73
Registered: 16-6-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
I'm an intraday mean reversion discretionary trader and using limit orders is important to catching key reversals in price when coupled with using
some multiple of ATR added or subtracted to highest (h, x bars) and lowest (low, x bars). Multiples of these kinds of mean reversion trades can be
caught intraday, therefore I had limit orders added to wish list. I mostly trade stocks and ETF's but I know they are useful to any strategy,
including futures.
Not only that, but strategy limit orders (assuming they are some pre-calculation of ATR or other important measure) allow for manual trading since
these orders show up in Strategy Orders tab ahead of when most limit orders will trigger. This allows a manual tracking and trading of the strategy
in real time if need be or if one wanted to.
Anyways, limit orders are something I look forward too, since they have many uses's (stops, entries, exits etc.) in strategy building.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I agree that limits and stop entries need to happen. Its a medium term goal to have them. I trade a number of mean reversion systems myself.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by rws | I found that when I add /drawdown to the fitness formula, building
systems get many times slower. Not 10 times but almost 100 times
Highly likely this is a bug. Will look at getting it fixed. My GSB does it to. version 28.5 |
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
This is fixed in 28.7
|
|
|
emsjoflo
Junior Member
Posts: 39
Registered: 7-7-2017
Member Is Offline
Mood: No Mood
|
|
Ninjatrader
I would like to be able to export code to Ninjatrader 8. Just a wish.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Export to mt5, amibroker is on the todo list, but GSB must be a lot mature, as all changes must flow to all the other programs.
What I mean by that is the current exported TS needs to not change after new builds. Every change is then going to have to be remade in
all the other programs. Its great that new features are being made all the time and im not yet sure when thats going to slow down.
Im very open to NT8 but its going to be 4 to 12 months away is my guess.
|
|
|
cyrus68
Member
Posts: 171
Registered: 5-6-2017
Member Is Offline
Mood: No Mood
|
|
Would it be possible to include the following simple statistical measures, as indicators, in future builds?
Sharpe Ratio
Regression Slope
Correlation
Standard Normal. defined as: [X - mean( X,n)] / [StndDev(X, n)]
I will gladly give up some of the indicators, already included in GSB, in favour of the above.
Also, it would it would be nice to be able to apply the above indicators on the Close of secondary data, and use the results as inputs to modelling.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Ive only seen Sharpe Ratio as a performance metric, not as a indicator.
linear regression could be done.
Correlation is between two data series. You would need custom indicator for this.
Standard dev is already done.
All indicators can be done on all data streams.
|
|
|
jptann
Member
Posts: 101
Registered: 13-6-2017
Member Is Offline
Mood: No Mood
|
|
Peter:
In my experience it is beneficial for all instruments to stop trading from June 15 through July 31 each year. I call this the "Hampton break" period
where all the NY guys go to the Hampton's and all of the Chicago guys go to Wisconsin or Michigan lakes. Can we get a filter in GSB that closes down
trading from X date to Y date each year.
Also, for just ES, sometimes January is a bad month. My experience over the last 5 years, is that January "sucks" in my performance so I would like
to remove this month also. So I need a filter set that is a little bit more complicated than what you provide. Like a way to "and" these filters.
Say stop trading from 6/15/xx to 8/1/xx and January/xx. I think the January would be easy. The way you have the days of the week set up. Just need
an inclusive range also.
Thanks in advance
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
If advanced mode is on, there is a filter called dates on the left GUI. This can do all of this but its going to be a manual process to enter in each
period.
You could also do it via notepad once youve put in a few dates. The file name for the settings is shown in the red box below. Its in the
\data\optimizer folder
|
|
|
cyrus68
Member
Posts: 171
Registered: 5-6-2017
Member Is Offline
Mood: No Mood
|
|
Yes the Sharpe Ratio is mostly used as a performance metric. But it can also be used as an indicator. Think of it as a t-statistic. i.e. a measure of
the significance of the slope. A higher t-statistic can be the result of a greater slope relative to its volatility. I can send admin the EL code for
this, if you like.
The standard normal statistic is NOT the standard deviation. It is the deviation from the mean, divided by the standard deviation. I repeat the
formula: [X - mean( X,n)] / [StndDev(X, n)]
This should be very easy to do. Far simpler than the slope.
I wasn't aware that you could apply the internal indicators to secondary data. Great to hear this. I will certainly try it.
|
|
|
QuantDean
Junior Member
Posts: 6
Registered: 9-5-2017
Member Is Offline
Mood: No Mood
|
|
Limit Order support please! 
|
|
|
jptann
Member
Posts: 101
Registered: 13-6-2017
Member Is Offline
Mood: No Mood
|
|
QantDean:
The issue with limit orders is that there is no way that you can be assured that you get a limit order fill. There is no question that strategies
that I develop have significant better performance, but how am I to be assured that the back test and walk forward are truly representative of the
strategy performance going forward into the unknown. This is especially significant in low volatility (like now).
My development philosophy is not to try to get the "best" system but I try to see how any system can fail. Limit orders add a significant error into
development work, at least for me, especially with swing systems. I much rather develop a Day trading or swing system with full slippage and
commissions using only market orders. This is the most conservative case. Then if you want to add a limit order component to the system (say for
adding another contract) you are assured that the basic system does not have any "gotchas". I can't tell you how many limit order systems that I have
developed where the performance is dominated by a few large trades where the system gets the absolute top or bottom of the move. Of course, you may
only have a 10% chance of getting that price and if you miss these few moves, the strategy is not very good.
This is my opinion only and what works for you can be different, but I want to make sure that any of my back testing, sensitivity analysis, and walk
forward studies, I can be assured statistically that I will get a fill.
|
|
|
gmoney
Junior Member
Posts: 14
Registered: 8-5-2017
Member Is Offline
Mood: No Mood
|
|
Can you make the default sort order for columns descending instead of ascending? Virtually every time I sort results I want the highest at the top.
It's minor but takes a step out of every time I sort. Thanks.
|
|
|
boosted
Junior Member
Posts: 73
Registered: 16-6-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by jptann | QantDean:
The issue with limit orders is that there is no way that you can be assured that you get a limit order fill. There is no question that strategies
that I develop have significant better performance, but how am I to be assured that the back test and walk forward are truly representative of the
strategy performance going forward into the unknown. This is especially significant in low volatility (like now).
My development philosophy is not to try to get the "best" system but I try to see how any system can fail. Limit orders add a significant error into
development work, at least for me, especially with swing systems. I much rather develop a Day trading or swing system with full slippage and
commissions using only market orders. This is the most conservative case. Then if you want to add a limit order component to the system (say for
adding another contract) you are assured that the basic system does not have any "gotchas". I can't tell you how many limit order systems that I have
developed where the performance is dominated by a few large trades where the system gets the absolute top or bottom of the move. Of course, you may
only have a 10% chance of getting that price and if you miss these few moves, the strategy is not very good.
This is my opinion only and what works for you can be different, but I want to make sure that any of my back testing, sensitivity analysis, and walk
forward studies, I can be assured statistically that I will get a fill.
|
Simple fix for limit orders. Builder has it and TradeStation has it as an option within Strategy options.
If limit orders are added to GSB it should also have "Fill limit order only if price exceeds limit order price." As long as this is an option fills
should not be any problem unless you are using large share/contract size.
|
|
|
| Pages:
1
2
3
4
..
17 |
![[*]](images/davis/default_icon.gif){kind=icon}
