Statistical Optimization (continued) As discussed in post #13, the randomness of the strategy results in inaccurate parameter optimization. Some of the optimization results will be lucky winners, while others might be better in practice but happen ...

The Myth of Averaging

Do you mean that with a small account there is not enough room to take some draw down? Yes, that is indeed a worry. Using 0.01 lots, each point movement the AUDCAD is worth $0.75. Suppose we have a very conservative leverage of 1:30, then we can ...

The Myth of Averaging

I am not sure what you mean by "because you can't". In MetaTrader 5 you can run one single backtest using multiple symbols at once. So you could create an EA that trades 20 symbols simultaneously in the strategy tester. But like I said, this only ...

The Myth of Averaging

Statistical Optimization Approach Looking at the optimization results above, it is remarkable how wide the range of SL and TP parameters varies in the top results. This suggests the results vary wildly due to the random nature of the strategy. In ...

The Myth of Averaging

Level 4 - SL/TP By now you can guess what is coming next... image From the optimization results: image we choose the highlighted entry which has relatively much trades and a nice tight stop loss. Note that all settings above had a positive profit ...

The Myth of Averaging

Level 3 - Equity Curves Let's look a bit closer to what we got so far at level 3. The worst backtest was -1020, the mean of the distribution is at about -200, and the best backtest gave +861 profit. Their equity curves look like this: image ...

The Myth of Averaging

Good observation! When I sort the profits in the backtest results in ascending order in a spreadsheet, I can see only 11 of them hit the -10k: image so we have 1.1% chance of completely ruining our 10k account. The lot size is already at minimum ...

The Myth of Averaging

In this thread I will focus on "averaging down", but you are right that it might be interesting to do a similar thread on Martingale! Diversifying over multiple symbols will certainly yield a smoother equity curve as the peaks and valleys of the ...

The Myth of Averaging

By adding a second recovery trade if the market goes against us, we improved the mean outcome of our strategy by almost 3000! The best case strategies give a nice profit, but admittedly the worst case also became worse. However, the mean is the most ...

The Myth of Averaging

Level 2 - SL/TP Now we are adding a second level. Once the first trade hit the stop loss, we now leave open the first trade and add a second trade in the same direction as the first trade. For both trades we now set new stop loss and take profit ...

The Myth of Averaging

Level 1 - EntryHour Now that we chosen a SL/TP to work with, we can search for the best time of day to enter the market. We let the MT5 strategy tester do an optimization run trying out 1 to 23. Note that we skip 0 because at midnight the spread is ...

The Myth of Averaging

Level 1 - SL/TP Let's get started! For testing an averaging/grid strategy it makes sense to choose a symbol that is fairly ranging in nature. For example AUDCAD is often used with grids. Since this EA is going to make a trade every day, it makes ...

The Myth of Averaging

LOL. Yes, I'd like a $50,000 daily airdrop, please

Most volatile EU/GU/XAU/BTC trading

Python is easy to learn and has several excellent ML libraries, so that is an excellent choice for most purposes!

Diary of a Pro Quant

Ah, I see. That is different from both methods I tried. Since we are looking at seasonal effects on a daily basis, it is very important the matching is exact. However, here it matters what seasonal event we are trying to find. Certain holidays are ...

Diary of a Pro Quant

Ah, perhaps that explains the difference between mine and your cumulative returns graphs. Come to think of it, if you enter/exit at 23:45 PM when spreads are still ok you will probably get almost the same results since volatility is usually very low ...

Diary of a Pro Quant

Below are the 2%, 16%, 50%, 84% and 98% relative returns per day of year (aka -2 to +2 sigma): image I also made a cumulative plot to mimic @ForexStats graph, but it looks very different: image @ForexStats: How did you match days? I tried these ...

Diary of a Pro Quant

I couldn't resist and coded the above out of curiosity. Here is a walk-forward simulation from 1977 to 2023 comparing daily trading based on median return versus a buy-and-hold since 1977: image Unfortunately, the model does not seem to provide an ...

Diary of a Pro Quant

Subscribed! Looking forward for new fresh ideas. Some thoughts on your seasonal cycle idea above: * These kinds of a statistics are very sensitive to outliers. One day with an exceptionally large profit/loss can skew your stats. Instead of the mean ...

Diary of a Pro Quant