If you had to choose ONE, what defines a good EA? - page 2

New comment
 
Enrique Dangeroux #:
(quoting Simon Gniadkowski) Take a look at this chart,  it shows the relation between Win rate (WR)  and Risk:Reward (R:R), in this case the spread was 0 and uses a simulated coin toss with an even number of Long and Short trades taken at random with no attempt to predict the direction of the market.  You will see for a 50:50 R:R scenario the WR is 50%
It seems to me that a coin toss is an oversimplified analogy to a financial market. While a coin is inherently binary (heads or tails), a trade is more dynamic. This is to say that the P/L of a coin toss is settled upon the very instant that the coin lays flat. In contrast thereto and in the markets, each long and short trade runs to varying distances away from trade entry price. Only at the time of trade exit, the P/L is settled─and a TP price twice as far away from the trade's entry price as that trade's SL price is inherently half as likely to be struck versus that SL price (remember, we are assuming no statistical edge in this coin toss entry logic "analogy"). All that it takes to create a losing test is consecutive more-likely-than-not stop-outs─which only reinforces the importance of maximum drawdown.
 
Ryan L Johnson #:
It seems to me that a coin toss is an oversimplified analogy to a financial market. While a coin is inherently binary (heads or tails), a trade is more dynamic. This is to say that the P/L of a coin toss is settled upon the very instant that the coin lays flat. In contrast thereto and in the markets, each long and short trade runs to varying distances away from trade entry price. Only at the time of trade exit, the P/L is settled─and a TP price twice as far away from the trade's entry price as that trade's SL price is inherently half as likely to be struck versus that SL price (remember, we are assuming no statistical edge in this coin toss entry logic "analogy"). All that it takes to create a losing test is consecutive more-likely-than-not stop-outs─which only reinforces the importance of maximum drawdown.

Drawdown is not the point.The point of the already stated simplified model is the proven relationship between RR and win rate. The edge if any of the system influences the curve making it worse or better.

Low max drawdown and high recovery factor will always outperform high win rate strategies!

Lower drawdown is always better, no to refute here. You can have a trading system with a tight stop == low win rate and accumulate more balance drawdwown and thus a lower recovery factor vs a system with a wider stop == higher win rate and as a result a higer recovery factor. 


 
Enrique Dangeroux #:
The point of the already stated simplified model is the proven relationship between RR and win rate.

Got it. Thanks.

So then in the chart, a 0.50 Risk:Reward is really just a 1:2 Risk:Reward expressed as a decimal, and a corresponding Win Rate of roughly 30% makes sense.

As a side note, I don't know many traders who are using more Risk than Reward. Perhaps an expanded chart of only 0.01 to 1.00 would be more relevant.

 
Eusebiu Dascalu:
  • High Win Rate
    14% (4)
  • Low Max Drawdown
    43% (12)
  • High Profit Factor
    18% (5)
  • High Recovery Factor
    25% (7)
Low max drawdown 
 
For me, the single most important factor isn't the 'Monthly Return'—it's the Recovery Factor combined with a strictly limited Drawdown. >
Many EAs look great on paper but fail during high volatility. In my development of the V18 PRO MASTER for Gold (XAUUSD), I focused on ensuring the Max Drawdown stays under 1%. If an EA can't protect the capital during a 'black swan' event, the profits don't matter. A good EA should prioritize survival over greed.
 
Samuel Mkandawire #:
Low max drawdown 
me too
 
Thank you all for your opinions. It matters a lot for making decisions.
 

If I had to choose one, I’d go with low max drawdown.

Win rate and profit factor can look good, but if drawdown isn’t controlled, the system won’t survive long enough to realize those gains. I’ve seen many EAs with strong backtests struggle in real conditions mainly because of poor drawdown control.

That’s also the reason I design my own EAs around a low drawdown approach — capital protection first, then performance.

12
New comment