I had this intraday setup I was feeling pretty good about.

    15-min chart.

    Buy when MACD crosses above signal and RSI is above 50.

    Sell when MACD crosses below and RSI below 50.

    1:2 risk-reward. Swing low/high stop.

    On charts it looked clean. Nice entries. Logical.

    So I ran it through AI to see how it actually performs.

    171 trades later…

    Win rate: 35.7%

    Avg win: ₹230

    Avg loss: ₹162

    Net result: -₹3800

    Yeah… not great.

    It wasn’t completely broken, but clearly not strong enough to trust with real size.

    Instead of ditching it immediately, I asked the AI what it would improve. And the suggestions were surprisingly practical and not magical, just structural tweaks.

    It said I should probably add volume confirmation.

    Which makes sense, MACD cross without strong volume is just noise half the time.

    Then it suggested adding a 50 SMA trend filter.

    Only take longs above it, shorts below it.

    Honestly, that alone might filter out a lot of chop.

    It also pointed out my stop-loss logic could be smarter.

    Instead of fixed swing stops every time, maybe trail after 1R so winners can breathe but profits are protected.

    Another thing I hadn’t thought much about – time exits.

    If a trade doesn’t move after 2–3 candles, just exit.

    No point babysitting dead trades.

    It even suggested testing other timeframes. I was only using 15-min.

    Maybe the logic behaves better on 5-min or 30-min. Fair point.

    What I liked most was this:

    It didn’t say “this strategy is bad.”

    It treated it like a rough draft and helped refine it.

    No promises.

    No “AI found the holy grail.”

    Just structured improvements.

    Honestly, seeing -₹3800 on paper probably saved me from losing more live.

    Curious how you guys improve strategies. do you systematically refine them like this, or mostly adjust based on feel?

    https://www.reddit.com/gallery/1r3uhny

    Posted by kairoX_14

    3 Comments

    1. ThanklessWaterHeater on

      If you ask an LLM for advice on a bad idea, the fact that it responds doesn’t change it into a good idea.

    Leave A Reply
    Share via