[Darwin] UKC by Morpheus33

[AriaS]

Have just seen that the Sharpe Ratio in May was 1.32. No wonder, since all the 27 trades were winners.

The strategy is now public on Myfxbook.

Btw, Myfxbook has malfunction. If your Sharpe Ratio shows 0.0, change the trading period by even one day (through “custom analysis”) and the Sharpe Ratio shows up.

 

[AriaS]

Stress Test

Made a successful stress test for the new back testing algorithm. Tested out-of-sample the months of interest rate changes, including both surprise and predicted. The result is good: no entrances during the news. The only problem, which did not occur in the stress tests so far, but might occur is either closure in loss, in the direction of the spike, or if the closing algorithm decides to stay and closes with a bigger loss. This can be a problem only during surprise. Maybe I should consider manually closing the trade, before the rate announcement, if I see that there is no wide consensus of economists.

 

[AriaS]

Received 2nd 30k allocation.
Decided to add a bit of contrarian sentiment to the strategy, but only to filter out some unwanted entrances. Back tested it, but didn't reach an unambiguous conclusion.

 

[AriaS]

I made a mistake that almost destroyed the account Yesterday. It survived though!

I had the equity breakeven turned on, but I didn't take into account that it needed to be turned off when I scaled the account x2, or I should have coded percent breakeven. Yesterday the floating loss was stopped out by the equity breakeven and this caused the Darwin to lose 3.32%. Luckily it is still less than made in 2 previous months combined, so my own rule is not violated.

Darwin:

Strategy:

 

[AriaS]

Got listed in the "Promising" list.​

 

[AriaS]

Recovered from September's drawdown.
The goal this month: to keep the 60k allocation next month.

 

[AriaS]

Update: need only +1.05% this month to get the allocation to 90k. The Loss Aversion score is very low now. I guess that's because the algorithm keeps holding some losing trades for a very long time, since the dynamic SL does not get triggered. Anyway, I trust its logic, since it always closes with the minimum possible loss.

P.S. Someone asked me why I deleted the account from Myfxbook. Because for some reason when added it gets listed as a demo. I tried deleting and re-adding several times and sent a message to support. No reply, and no fix. Anyway I think that having it on Darwinex is more than enough.

 

[AriaS]

Not satisfied with the result, but since no rule of mine has been broken, I will keep on. At least I know that if it weren't for the equity stop accident (in the middle) it would be much better. The gain would be 55% (48% by equity).

 

[AriaS]

In the last 3 days, ChatGPT helped me build a new backtesting algorithm that is much better than the one I was using before. The key improvement is this: instead of hunting for some perfect setups that don't have drawdowns, the algorithm welcomes them and explicitly evaluates how the setup recovers from them. And also the total Out-of-Sample period grew from 3 months to 27 months, which gives much more realistic picture of behavior across different regimes.

The process now looks like this:

1. Optimization (3 months by Recovery Factor).
2. Recent Out-of-Sample (9 months). The most important phase. Here I strictly filter setups by RF grades and recovery behavior (frequency and duration). The latter is analyzed by GPT. Grades: >=2.0: excellent; 1.5-2.0: good; 1.2-1.5: weak; =<1.2: reject.
3. Long OOS (full year before the recent OOS). This phase is used to understand robustness and regime sensitivity: >=1.3: robust; 1.0-1.3: regime-sensitive; =<1.0: fragile. A weak result here does not automatically reject the setup, but it signals higher risk and affects position sizing.
4. Stress tests (Covid, Ukraine invasion): the purpose here is survival only. The setup is rejected only if recovery logic breaks and DD becomes unbounded.
5. Repeat steps 1-4 every 1-2 months.

Now it looks like professional Quant and hopefully this will improve the result that I am currently not so happy about: RF = 1.6 in 8.5 months.

 

[AriaS]

Compared the strategy to a hedge fund

Here is the how to:

If you want to evaluate your strategy properly, use a strong LLM to calculate all the relevant metrics. Trust me - if it can calculate time dilation from the Schwarzschild metric near a black hole, it can compute your Sharpe ratio.

Risk-adjusted metrics and benchmark choice depend on the type of your strategy. For example, if it is market-neutral FX (like mine), it makes no sense to benchmark it like a long-only stock portfolio. The LLM needs to understand whether you are running trend-following, mean reversion, market-neutral etc. Show it and explain everythig.

Next, show it your daily returns - that gives much more statistical power. Monthly works too, but less accurate.

Make it calculate the fllowing:

• CAGR
• Sharpe ratio
• Sortino ratio
• Calmar ratio
• Beta
• Absolute alpha
• CAPM alpha
• Alpha t-stat

Important: the larger the data sample, the stronger the statistical confidence. Early metrics can look amazing simply because of a small sample size.

Below are my current figures. In my case, any comparison to a typical hedge fund is statistically limited, because my live sample is only 10 months old. For a meaningful comparison, you would ideally want 5+ years old.