The Humble Shorter

[remy1111]

Recently, I’ve been working on setting up a BTC short bot together with ChatGPT, and I have to say, that works surprisingly well.

The bot only opens a deal when several indicators line up and suggest that the market looks strong in the moment, but is still structurally bearish. For example, RSI above 45, the price sitting in the upper part of the Bollinger Bands, a bearish market structure on the 15‑minute chart, and a SuperTrend that continues to point downward.

Take profit is mainly based on SuperTrend, with an additional AVP target of 5%. I’ve also added a kind of “fake” stop‑loss. It’s set to Webhook, but there’s no trigger behind it, the important part is that it still uses the move‑stop‑loss function.You’ll notice it triggers quite often, but that’s exactly the point: it helps ensure the bot closes trades in profit.

The DCA settings can handle a decent upward move, and because the entries are fairly sharp, the bot doesn’t need to cover too much before the position starts working in its favor.

Looking at the backtest, the most striking thing is how steady the bot behaves over a long period. The test ran from December 9, 2021 to February 8, 2026 on the 1‑minute timeframe, which gives a pretty reliable picture of how the strategy performs across different market conditions. The max drawdown stayed at ‑11.55%, which is very reasonable for such a long test. Most trades finish within a day (average 1D 10H), although there are three trades that stretched to around 65 days. One of those happened right around the bearmarket low in late 2022. If I ever get the feeling/indication that we’re approaching a similar bottom again, I’ll set a ‘minimum price to open deal’ so the bot doesn’t start shorting at the very bottom of a bear market. The other two long‑running trades still had one or two DCA orders left, so there was enough room to manage them safely.

All in all, the test results give a good impression of how the bot performs: controlled, consistent and without extreme outliers. The disadvantage is that it does not generate extreme profits, but every profit counts.

Backtest Data

Links and Settings

• Backtest URL: Gainium app

• Bot URL: Gainium app

• JSON settings:

  BTC Short.json (21.4 KB)

[remy1111]

I’m also working on a long bot, but I’ll come back to that later.

Once Kraken becomes available, I’m planning to run this one with 2x or 3x leverage as well (after properly backtesting it first), because I have a feeling it will perform well there too.

[aressanch]

Thanks for sharing! Gona give this one a try on paper.

[remy1111]

I’ve made a small adjustment, I set the Move Stoploss Unrealized Profit Trigger to 1.0 and the Stop Loss Value to 0.6.

I have noticed that it closes a little too quickly on the stop loss, while on the higher TF BTC is still falling.
In the coming days/weeks, I will see if I can find a different/better exit strategy. Yesterday, I was busy working with Supertrend on other timeframes, but that wasn’t always better. I have also tried a few options with additional TP indicators, but then it often performs (much!) worse.

The backtests below show the difference between the old version and the new version. The new version stays in deals slightly longer, resulting in fewer deals, but generating more profit.

[remy1111]

Another interesting consideration is increasing the number of deals and possibly adding a cooldown period after starting a deal.

[remy1111]

@aressanch If I go for the 2 deal option (with DCAs included) will that work with Currency reference set to % BTC Total, or should I choose % BTC Free? Or will neither of them work?

[aressanch]

I find that when the bot opens more than 1 deal the % options are less intuitive, no matter is you use free or total (doesn’t matter here because free and total base will be the same in a short). The issue is that when the second short opens, it will be 10% of the remainder 90%, so 9% of your original quantity. If you don’t mind this, you can set it to either, otherwise you can set the base order in BTC so it’s exactly 10% of the original.