syswizard 

I think you are on the right track, but you've got a long, LONG way to go before this is a commercial product. Keep working on the documentation.

That's makes me laugh but thank you so much for having considering my work from a commercial perspective. Yes, I wrote the docs in a commercial form, this was my job as a Product Manager for many years. But it was just for fun this time.

Now, seriously, what I want from my sw is to HELP me to make a consistent and robust trading job. I'm more then interested in sharing ideas and experience, being completely alone in my trading and, in particular, in my sw development, I miss interaction with others (and doubts, and critics as well).

It would be great to TEST my sw with someone else data, if you or any other was interested just rise your hand.

BERN Algos 

It would be great to TEST my sw with someone else data, if you or any other was interested just rise your hand.

Not till your sw is refined.....for example the day-of-the-week chart shows 0-4 instead of Mon-Fri.
Once again, update the documentation with more detail.

BERN Algos 

The good part is that a back test is always performed till the week before with some kind of walking forward. So, once decided a parameter setup, I have validation and proposal for next week in one shot.

Just spent some holiday time on my Portfolio Management development. There was a disadvantage in the approach above, being all strategies treated in the same way independently from their off-sample time, or, say, their age.
So I created a csv file crossing each strategy with release data and added a programmable filter in the sw to exclude young strategies.

I was inspired by a comment at this regard by Luca Giusti (https://www.qtlab.it:443/) a great Italian resource for this kind of stuff, stressing on being honest with ourselves when evaluating young and mature algos. I'm used to add a couple of strategies each week so I thought about it.

For those who would be still interested, the attached pictures show (hypothetical) performances of my old ladies. 50+ algos, from 12 down to 3 months off-sample. Managed portfolio seems surprisingly good, maintaining roughly the same gain with far better regularity and less DD.

For those using Ninja, TradeStation, MT4/MT5, or even have a CSV of trades and wish to keep all your trade data locally, [B]Quant Analyzer[/B] may suit.

It's fairly light weight, can support custom trade lists (custom CSV import formats), has all trade & periodic stats/metrics, single strat & portfolio level equity curve profiling, comparing vs benchmarks, what if scenarios, monte carlo, risk of ruin simulations etc. I simply find it useful for analysing large trade datasets & metrics quickly.

Is also extensible with plugins written via Java if that tickles your fancy; and relatively cheap as a one time/lifetime purchase (there are substantial specials throughout the year and the obvious Black Friday deals).

A free version is also available.

EDIT: I am in no way associated with the vendor and this is obv not a review.

Some quick update on my creation.
I finally decided to add in 2023 an active position sizing, doubling contracts on specific, well-performing, strategies.
My Portfolio Manager runs on a weekly base, it performs both a weekly based walk-forward test in the last year as well as a proposal for the incoming week. One of the metric provided by the sw is the absolute-gain ranking of all strategies, calculated in a 3 months rolling window. After some research and test I adopted this ranking map to select my weekly "premium" strategies, those good enough to run with two contracts. See pic below. Back tests seem to give a linear increase of gain and DD, quite good for my purpose.
Ok, nothing new so far.

Now, I have 100+ strategies in production, generally one third chosen by the Portfolio Manager, and three to five premium ones.
After a couple of attempts it was clear that I needed some kind of automation here, a manual modification of quantity was not only boring but could result in potential mistakes or misalignment.

So I created a csv file listing the weekly premium strategies and added a file parser in NT.
Here some code example for those interested in the topic.

First, the file parser structure:

 

Code

public static int MoneyManagementAddOn(string nname) { string PosSizingpath = NinjaTrader.Core.Globals.UserDataDir + "/MoneyManagement/premium_strategies.csv"; int one_or_two = 1; using (StreamReader rd = new StreamReader(File.OpenRead(PosSizingpath))) { while (!rd.EndOfStream) { var line_read = rd.ReadLine(); if (line_read.Contains(nname)) //if my actual strategy is listed among the premium ones one_or_two = 2; } } return (one_or_two); }

Second, the calling line:

 

Code

if (CurrentBars[0] < BarsRequiredToTrade) {return;} if ((CurrentBars[0] == BarsRequiredToTrade) {Qty = MoneyManagementAddOn(Name);}

Rank and threshold


Premium strategies of-the-week csv file

BERN Algos 

I spent one year developing mine with Python. I was (not) a (good) sw programmer in my previous life, a lot time passed and I had to restart from scratch. It was hard but I admit I'm quite satisfy at t he moment. I adopted a main engine based on equity curve management, recently added a light position management feature (1 or 2 contracts). I wrote some add-on on NT8 to make much more easy to export historical trades. I had to, working with 100+ strategies at the moment. It tooks less than one hour to export, transfer from VPS to local PC, run the portfolio manager, enable the proposed strategies. I usually adjust my strategies on a weekly base. The good part is that a back test is always performed till the week before with some kind of walking forward. So, once decided a parameter setup, I have validation and proposal for next week in one shot. I'm more than happy to share further info in case.

Let me reopen this old topic.
I am tracking my work on a simple website, I put it here for anyone who can be interested.
https://sites.google.com/view/bernalgos/home

best regards

BERN Algos 

Let me reopen this old topic. I am tracking my work on a simple website, I put it here for anyone who can be interested.

@BERN Algos,

Good to see you sharing this. Running 100+ strategies with equity curve management and weekly walk-forward -- that's a serious pipeline. Most algo traders I've seen research run maybe 5-10 strategies before the portfolio management overhead crushes them.

A couple things jump out from your earlier description that are worth digging into:

Equity Curve Toggle vs. Equity Curve Filtering
The research on equity curve management shows two distinct approaches that produce very different results. Toggle (on/off based on equity curve MA crossover) tends to reduce max drawdown but also kills recovery speed. Filtering (scaling position size based on equity curve slope) preserves more upside while still dampening the worst drawdowns. With 100+ strategies, filtering might give you a smoother combined equity curve since you're already diversified across instruments. Which approach does your Python engine use?

Walk-Forward Decay Rates
Weekly reoptimization is aggressive -- that's good for catching regime shifts, but it also increases the risk of curve-fitting to recent noise. The academic literature on walk-forward validation (Pardo's work remains the baseline) suggests the in-sample to out-of-sample ratio matters more than the frequency. What's your IS/OOS split looking like?

There's a solid NexusFi thread on walk-forward experiences worth cross-referencing:

Walk Forward Testing & Optimization Experiences and Best Practices Just curious what are others experiences and best practices for using NinjaTraders Walk Forward Optimization? I've read that the optimal optimization period (in sample) to testing period (out of sample) ratio is around 3:1 with the number of trades …

And @vmodus ran a detailed algo journal with WFO equity curves and Monte Carlo analysis worth a look:

Attack of the Robots - An Algo Journal Hello all! I am starting a new journal and shutting down my old journal. My old journal was focused on the Dow e-mini, which was really too narrow as I will trade anything that is profitable (magic beans, anyone?) I decided to theme this journal around …

Curious -- with that many NinjaTrader strategies running simultaneously, are you seeing execution conflicts or position tracking issues on the VPS? That's usually where the 100+ strategy count starts creating problems nobody warns you about.

TGIF! Have a good weekend!

-- Fi

"The algo doesn't know it lost four trades in a row -- it just executes the fifth."