Conor Mcnamara / Profile

I would describe automated trading trading as something that does work. But when you understand what trading is about... it's not more ideal than manual trading.
It is like a car driving by itself. Do you trust that a car will automate driving perfectly? Are you happy to sit in a car that drives without your control? This is exactly how I would describe automated trading.

It's common that EAs won't trade on an ideal entry price. Most EAs are simply trading when a signal comes. But I learned that's not sophisticated enough. When you get a signal to trade, you can add more logic and use two flags "buyPrepared" or "sellPrepared". Then do price detection of local extreme price from that shift in time of when the trading signal came.

There are many indicators on the codebase that are just proof of science or for data organization, but they're not suitable for trading signals. For "quant" trading, it's a layered combination of tools which makes something suitable for signals. Combine ADX, ATR, and an oscillator..and this gives a much more reliable trend

this wonder zigzag got a full optimization, but the code is not available
It can be provided for payment, and payment must be in the form of a briefcase of cash

The file will be provided to you on a floppy disk

It would be helpful to have a table of algos which are profitable in automation, and which are not.
Some indicators are profitable sometimes, but can't survive all market conditions, which means the algorithm essentially isn't good enough for automation. It seems that with most of them, there's a problem. Parabolic sar for example will fail in ranging markets. Fractals is often after the fact, and not always giving true signals. Donchian channel and zigzag by itself is constantly repainting. RSI gives false signals if the market starts to trend in one direction.

Multitimeframe systems and systems that combine indicators are promising, but there are downsides to this as well (it will calculate fewer opportunities and miss out on many).

There's only a few algos left to be reliable for automation.

statistically, multi-timeframe trend shows much more positive results for quantitative trend indicators proving that it's not a fad

extreme risk is not a bad thing if the deposit meant nothing to you, and the fact of trading is that higher risk will always give a more fruitful profit over time. But I see so many EAs with no fail-safe features.
For example, let's say the user wants to take really high risk, and the account gained 30%. At least then you could implement a fail-safe where the deposit is protected and the EA can't fall below the original deposit after a certain account gain is reached. I coded these deposit protection features, because it's not logical to allow the expert to blow the account at a late point. It's very easy to code these things, there's no excuse not to have money management

In your EAs, do you create a time window for trading? or do you let the EA trade all hours and all sessions? and if you do create an hourly trading window, which hours do you let the EA trade?

small backtest on the time where real ticks were available, this is trading on H1 chart with the trend following indicator on H12. Possible overfitting, and will not always be smooth. this is why we need either to supervise at times or else automate risk management such as a drawdown controller

some small discovery in mtf trend following indicators if anyone is interested

M5 chart timeframe likes the indicator to look at M15 or M20 for short term scalping

M20 chart timeframe likes the indicator to look at H8 for swing trades

H1 chart timeframe likes the indicator to look at H12 for swing trades

using the two-stage entry concept on M5 chart with the trend following indicator looking at M15 (MTF). Drawdown is 1%

because of the fact that indicator signals can sometimes be false, I developed a concept where the quality of the signal is first tested by the EA with 0.01 micro lot size before it executes a normal sized position (based on a risk % of the account). The PnL of the 0.01 lot size trade is examined... if the position was profitable, it makes another trade with bigger volume, if the position was unprofitable, it could mean that the indicator signal was false. This proved to be good in backtest. Instead of performing surgery on a weak trade, this basically pinpoints bad entries and cancels them

Russian roulette level risk and small take profit

I automated Ron Blacks "Swing line" to do scalping on the H1 timeframe, and on backtest for 2024 it gave very impressive statistics when combined with ADX trend strength. But in 2025, the results are poor surprisingly. I wonder why that is

How do people think they published a safe and profitable EA when it doesn't use stop losses and allows equity drawdown to to continue over 50%? There isn't always a correction in markets so the money will be lost, and there will be a margin call. Shame on people purchasing pricey EAs from idiots

"The road to hell is paved with good intentions"

https://www.mql5.com/en/code/87 this is a clever and elegant indicator which measures volume pressure, it's a unique idea, it compares the live tick volume on each current bar to the average volume of the entire series and compares percentages of volume

I don't need to use AI for coding, but could you argue that if AI isn't helping you, then it is not a fault with AI but rather the fact that your prompt isn't reasonable enough? Maybe I'm way off I don't know

Published code Trendline zigzag in qualitative channel

Good evening