Going live with my portfolio

Hi all, attached you find the status of my first live portfolio system after a few hundred trades. It's a portfolio consisting of 10 different versions of cycle-, trend-, cluster-, countertrend- and season algos combined with 20 assets. It uses 1h to 4h bars in my OANDA account. Each algo was tests for some weeks succesfully on a demo account. The portfolio was backtested with a sharp ratio of ~3. Although only a few weeks online it created some profit (initial capital 4k).

- Thank you JCL for providing such a cool tool and all the knowledge sharing! Also thanks to everybody who answered all my newbie questions :-)
- During my developmet work i was never able to reproduce winrates like the benchmark Z-systems have. I always had <40%. Meaning, my success is depending on a few big wins, that can be seen in my attached statistics. Is that something to worry about?
- I have algo performing very well so far and others generating minor losses. When (and based on what) do you typically decide when to take out an algo out of your portfolio or minimize it's margin?
- Any comments or suggestions how to improve?

If anybody is interested in my MATLAB tool that takes trade.csv and generates this nice pictures.... feel free to ask.

Stefan, after 7 Weeks of live trading, and with no knowledge of your off-sample back test statistics, this could be just a very initial assessment.

Impressive, that you could set up a broad ensemble of strategies and assets. I would not be worried, that you cannot reach the declared returns of Z Strategies yet.
But, with an ensemble of 10 algos/20 assets, perhaps you should expect a smoother balance curve. I mean the algos balancing each other for minimizing the drawdown.

But again, it is too early to say. Thanks for sharing, good luck.

Thanks for your comments. Maybe starting in these times is a little bit more confusing than normal as the market has changed so drastically in the last weeks.

But I am still in the winning zone and observing / optimizing

- During my developmet work i was never able to reproduce winrates like the benchmark Z-systems have. I always had <40%. Meaning, my success is depending on a few big wins, that can be seen in my attached statistics. Is that something to worry about?

> % of wins is just one metric. Maybe your strategies are leaning towards benefiting from a trending market. If so then low % of wins would be expected.

Do your countertrend algos have a higher win %?

- I have algo performing very well so far and others generating minor losses. When (and based on what) do you typically decide when to take out an algo out of your portfolio or minimize it's margin?

> Are you using Equity Curve Trading? https://manual.zorro-project.com/trademode.htm
"For this purpose the strategy monitors the equity curves of any component of a portfolio strategy. When unusual losses are recognized, real trading with that component is suspended. For detecting whether it's worth to resume trading, the subsequent trades are then entered in phantom mode until their returns indicate that the market is getting profitable again. In this way the strategy switches in and out of phantom mode dependent on the market situation."

e

Here is a piece of code for phantom trading.

// Equity curve (phantom) trading
void equityCurve()
{
#ifdef PHANTOMTIME
vars Equities = series(LowPass(series(ProfitClosed+ProfitOpen),8));
var EquitySmooth = LowPass(Equities,PHANTOMTIME);
if(falling(Equities) && Equities[0] < EquitySmooth && LossStreakLong+LossStreakShort > 2) {
setf(TradeMode,TR_PHANTOM);
} else {
resf(TradeMode,TR_PHANTOM);
}
#endif
return;
}

Thanks for your input! Indeed my countertend has best win rates of all >50 %.

I experimented a lot with equity curve trading, but did know with what goal for optimization of filter parameters. Of course profit factor can be optimized, but in reality I did not see a big benefit. Up to now, I am observing each component I a Matlab generated heat map and manually switching them off with with the factors file.

Thanks! I experimented as well with algos like that.... But did not get a substantial benefit. Did you?

My more specific question regarding Z system would now be, how Z system developers can judge what component is actually generating profit and which not..... Somewhere in the Z system manual on the web page, a table gives this indication.... Based on what?

It is done by phantom trading.
More or less as above, plus some tmf tricks, correlated loss streaks turn the position around.

That is, why it is so important to not touch the Z-Systems.

Above are assumptions, but somewhere there is the trick.

So is equity curve trading something like some suggest to only paper trade if the current equity level is below the moving average of the equity curve? I have read some money management suggestions if your equity begins falling below the moving average, you only start trading again live, once your equity (using paper trading) has gone back above the moving average.

Phantomtrading depends on you strategy design, it can improve or reduce the results.
If you are trading statistical effects like go long EUR/USD by the beginning of December, and short EUR/USD by the beginning of January, phantom trading will not really help you.