- ago
I have noticed over the last few days that the positions from the previous day sometimes do not match the positions on the following day. In other words, I opened 5 positions yesterday because WL gave me 5 signals. Today I find two additional positions in the trade history (positions) that should have been opened yesterday. However, these were definitely not included in the signals yesterday. I noticed this for the first time on September 26/27. At that time I thought it was due to the “Wealth-Data Alert for 26 Sept 2024” (there is a post in the forum about this). Since the problem occurred again today, I think it's a different problem. I had switched to Wealth Data in mid/late September as only they show the correct open rates (Norgate occasionally has different open rates). When using Norgate, however, I had not noticed the problems mentioned with deviating positions. Could this be due to subsequent editing of the WealthData? If so, when are these normally revised? I often place my market on open orders 9-10 hours before the market opens. Does anyone have any idea where else this problem could be coming from?

0
288
19 Replies

Reply

Bookmark

Sort
Glitch8
 ( 11.74% )
- ago
#1
It’s likely due to NSF positions and trade randomization on each backtest run. If it’s a market order system you can use transaction weight to avoid it, and be sure to place the entry orders with the highest weights.
0
Cone8
 ( 6.71% )
- ago
#2
QUOTE:
I opened 5 positions yesterday because WL gave me 5 signals. Today I find two additional positions in the trade history (positions) that should have been opened yesterday. However, these were definitely not included in the signals yesterday.
I'd agree with Glitch except for what you said here. NSF Positions wouldn't have caused the Entry Signals to have changed, only the backtest positions that they opened.

That said, you're going to have to come up with more evidence because I'd expect this would occur only for one of these scenarios (maybe #3 is likely???):

1. the data (or candidates) changed, or
2. the strategy or its parameters changed, or
3. you've specified Max Entry Signals > 0 but have not assigned Transaction Weight in the Strategy, or
4. less likely, your strategy uses IIR-type indicators (e.g., RSI, EMA, and many others) that had not stabilized. See User Guide> Indicators > Stability of Indicators
0
- ago
#3
“Retain NSF Positions” is activated. I also use a transaction weight. If I only run the backtest until yesterday, the new two positions also appear as a signal. Only yesterday this was definitely not the case. I suspect that something must have changed in the data feed. At least I have no other explanation for this. The backtest from yesterday cannot be reproduced either, as the 2 signals that were not shown yesterday now always appear.
0
Cone8
 ( 6.71% )
- ago
#4
1. Which Provider is it using?
2. What are the top two checked Historical Providers?
3. Did you make a change to Historical Providers?
4. At what time of day did you run the strategy the first time?
5. In which tool is the Strategy running?
6. Is Preferences > Data > Allow Data Corrections from Historical Providers checked?
0
- ago
#5
@Cone Your answer overlapped with mine. Then the cause could lie in no. 4. I'll have to read through that first. I suspected the data feed, as that was the only change I made. But now I have another starting point. I'll take a look at it.
0
- ago
#6
1. Which provider is it using?
I had previously used Norgatedata. The problem occurred for the first time when I was using Wealth Data.
2. What are the top two checked Historical Providers?
WealthData, Q-Data, Yahoo and Norgatedata are checked in the Data Manager under “Historical Providers”.
3. Did you make a change to Historical Providers?
no
4. At what time of day did you run the strategy the first time?
About 9-10 hours before market open (around 0630 CET)
5. In which tool is the strategy running?
As Building Block
6. Is Preferences > Data > Allow Data Corrections from Historical Providers checked?
Yes, this is checked.
0
Cone8
 ( 6.71% )
- ago
#7
5. In which tool is the strategy running?
The tool would be Strategy Window, S. Monitor, Streaming Window, etc.

It's not likely to get there for U.S. stocks with Wealth-Data and Q-Data above it, but you should put Norgate above Yahoo! for sure.

Wealth-Data filter corrections run 1 to 2 hours after the market closes. If you run the Strategy before that, the data (primarily the High and Low values) can change.

Norgate data can change for corrections at any time. This was a rare event, but I noticed that Norgate applied TDG's $75 special dividend more than a week too early for TDG, corrected that by unapplying it after several days, and then applied it when it really occurred - ex-date Oct. 4th.

In fact, this kind of event (a special dividend) is "data change". They don't happen that frequently, but for the Wealth-Data DataSets there could be a handful each month.

At the risk of "cliffs" in your data (like one for $75), you can specify that Norgate Data does not adjust for special events like that, and make sure your strategy runs using that data. Generally, this kind of a adjustment won't change your strategy signals, but for edge cases it could.
0
- ago
#8
I have now deactivated Yahoo data (I don't use it anyway).
I don't understand question no.5. I do not use any tool. I have a building block strategy, open it with a double click and then run a backtest. It is not a MetaStrategy either.


...I think I know what you mean. I don't use a tool.
0
Cone8
 ( 6.71% )
- ago
#9
The Strategy [Development] Window with the Design Surface for the block strategy is the tool. You're running it there and not in the Strategy Monitor, for example.

You could also drag it into a Streaming Window, which wouldn't make sense for an EOD strategy, but you can do it.

And then there's the Signal's Publisher... I could go on.
0
- ago
#10
Thank you, now I have understood what you mean. I use the Strategy window.

I have now read through the explanations of the IIR-type indicators. I think I can rule this out as a source of error. The backtest runs for 5 years and I do not use long lookback periods for the indicators I use. The longest is 80 days.

I also noticed this problem for the first time after using WealthData. I have not had such a case with Norgate so far (August and September 2024). Due to the problem of correct open rates, I only wanted to use WealthData permanently. So my first assumption was that it might have something to do with the data.
0
Cone8
 ( 6.71% )
- ago
#11
Let's see if you can find a smoking gun.
1. Copy and paste your signals into a spreadsheet each day.
2. Also each day, right click and "Copy All" of your Strategy Settings too.

I'd also recommend you check this to make sure you're exclusively using the same data and that it doesn't change Providers.




0
- ago
#12
Thank you very much Cone. I will do that. How can I save the strategy settings with a right click? In the BuildingBlocks, nothing is executed/no menu opens when I right-click.
0
Cone8
 ( 6.71% )
- ago
#13
In Strategy Settings, right click inside one of the frames - like Position Sizing and Copy All settings.
0
- ago
#14
Thank you very much! I suspect “Obtain data from selected DataSet only” could also be the cause. I hadn't ticked that so far. I have now checked the box and will keep an eye on it. I hope I will find the culprit soon :-)
0
- ago
#15
Since activating the checkbox “Obtain data from selected DataSet only”, I have not been able to detect any deviations. So it really seems to have been due to this. My question now is how WealthLab handles the data without this checkbox? Does this mean (if the checkbox is unchecked) that although WealthData is selected, other data sources are used?
0
- ago
#16
QUOTE:
Does this mean (if the checkbox is unchecked) that although WealthData is selected, other data sources are used?

Of course they're used by design, in the order of appearance on the Historical Providers tab. Activated providers only. (i.e. checked entries)
0
- ago
#17
I don't quite understand the meaning behind it. If I have selected WealthData, why do I still have to use other providers? Which specific rates are used for the backtest (if the checkbox is unchecked)? Are average values formed from all of them or how does the data selection work if the checkbox is unchecked?
0
- ago
#18
Strike F1 while on the "Historical Providers" tab of the Data Manager, it's answered in "The Data Provider List". When a symbol can't be found it's requested from another source, unless you check that option from Post #14. To me its purpose looks pretty unambiguous.
0
- ago
#19
Thank you for the explanation :-)
0

Reply

Bookmark

Sort