MetaStrategy-different-results-vs-individual-strategies

Hi,

Have you correctly assigned Transaction.Weight to trades in each of the individual strategies?
https://www.wealth-lab.com/Support/ApiReference/Transaction

P.S. Please use PNG instead of JPG for screenshots to avoid the blurry look.

QUOTE:
deactivated NSF

Deactivating NSF doesn't avoid the NSF issue. It only removes those trades from the backtest and therefore their Positions and eventual exit signals are not processed. This has the effect of speeding up large backtests / optimizations.

With Retain NSF Positions ticked, do you have NSF Positions in any given backtest? If you do, results will vary if you don't properly assign T.Weight.

Many thanks for your response.
I double-checked the Transaction Weight, indeed in one of the strategies hasn't had one. I entered it as well as ran the strategies with and without NSF ticked, no difference in the results anymore! Thanks!
However, I still get different results in the meta strategies vs. individual strategies.
Are there any other tweaks?

QUOTE:
However, I still get different results in the meta strategies vs. individual strategies.

We'll need to set up a test case, can you help us with sample code that demonstrates the anomaly or the name of an un-modified public Strategy?

Thanks Eugene, I replicated the issue taking pre-defined strategies (Knife-Juggler & RSI2) adding transaction weight. Still there is a difference in running those as a meta-strategy. To illustrate I got some screenshot (NSF ticked):

My response wasn't complete before and may have misled the reader that backtests are the same with and with NSF checked - maybe, but probably not. The reason is explained in the Help guide > Strategy > Strategy Settings > Advanced Strategy Settings:

Unchecked
Backtests discard NSF Positions and they are not added as Open Positions. A typical strategy using single position logic will continue to process the entry logic block. The main effects are:

1. Strategies have the opportunity to trigger and enter Positions immediately following trades rejected due to buying power.
2. Large simulations may run noticeably faster.

For the purpose of the discussion, let's accept that the results will be different with and without Retain NSF Positions checked. If we accept that, then I think currently MetaStrategies does not consider the Advanced NSF Setting and will always Retain NSF Positions.

If I'm not mistaken, the profit shown for your Knife Juggler-TW indicates that you ran it with NSF Unchecked. My result - after refreshing Wealth-Data - is:
NSF Unchecked: $1.990M profit
NSF Checked: $1.447M profit

Because Wealth-Data applies corrections (mainly for special dividends) from time to time, to get the same results as me, you need to completely refresh the data.

QUOTE:
If we accept that, then I think currently MetaStrategies does not consider the Advanced NSF Setting and will always Retain NSF Positions.

Yes, MetaStrategies are forced to always Retain NSF Positions.

QUOTE:
To illustrate I got some screenshot (NSF ticked):

Christoph, have you enabled "Retain NSF..." in each strategy and saved them prior to including individual strategies in the Metastrategy?

If you're running MetaStrategy Components with Percent of Equity sizing, I would never expect the same results as the single run. Because the positions are sized dynamically, based on the COMBINED equity curve of the MetaStrategy. This is surely different than the equity curve that's generated when a Component strategy runs alone.

That said, we would expect to get the same trades with $ sizing. After testing it, I found 4 trades (out of 5700+) that were in the S. Window, but not in the Meta S. And, 2 trades that were in the Meta S. but not in the S. Window.

We'll work to improve those discrepancies.

Update:
User error on that test - there were NSF Positions. After correcting that, the only discrepancies were the 2 trades in the Meta S. that were not in the S. Window. Both occurred on the same day/bar. Looking into that.

We're continuing to nail this down. The 2 trades that were "injected" by the MetaStrategy were 2 of 3 symbols that left the Dow 30 index that day. The trades should not have occurred and that's why they were "missing" in the S. Window backtests.

Remember, when backtesting with Wealth-Data DataSets, you're getting a backtest using the actual components at the time. If a backtest owns a symbol that leaves the DataSet's index, Wealth-Lab will close that Position automatically. The discrepancy with MetaStrategy deals with that, and I expect it to be fixed in the next build.

QUOTE:
If you're running MetaStrategy Components with Percent of Equity sizing, I would never expect the same results as the single run. Because the positions are sized dynamically, based on the COMBINED equity curve of the MetaStrategy.

Yes, that makes perfectly sense. However, I thought, once I set rebalance to never, I should get the same results.
I re-ran the strategies with fixed position size and still I get different results

Let's see how it works for you once build 21 is released.

It seems that the issue with different results between the meta strategy and the individual strategy is not yet solved (Build 49).
I tried to use a single strategy in the meta strategy builder. As there is only one strategy the result of the meta strategy and the individual strategy should be the same (whatever strategy you are using). But as you can see in the screenshot there are different values shown on the screen.

I don't know if it is a bug or a feature: MetaStrategy is not reacting to the checkbox "Retain NSF Positions".

If "Retain NSF Positions" is checked or not in the imported strategy, if it is checked or not in the strategy settings of the MetaStrategy, the result is always the same: Wealth-Lab calculates the strategies always as if "Retain NSF Positions" is checked.

Is this behaviour intended or is it going to be fixed?

QUOTE:
Is this behaviour intended or is it going to be fixed?

You will find this question answered straight in the Help > Strategies > MetaStrategy.

QUOTE:
It seems that the issue with different results between the meta strategy and the individual strategy is not yet solved (Build 49).

Is this behavior consistent for this strategy, some or all strategies?

Also what happens if you choose a fixed dollar size?