What to consider when training your Algorithm Intelligence Strategy

Everything you need to know about training your Algorithm Intelligence Strategy.

Training

You don't need to train your Algorithm Intelligence before using it for the first time. Without training, 100% of buy and sell signals will be used to open and close positions. Over time, as the AI validates signals based on your configured settings, it will become more selective and find fewer — but more accurate — opportunities.

Validation happens per trend and always uses the weighted average. When using only Signal Strategies, we do not recommend using the "Corrected score" setting.

How often do you need to retrain?

Once trained, you generally don't need to retrain. However, retrain in the following cases:

• You added a new source — only train the new source. Training all sources again would double the results.
• You changed your AI config or reset the AI — retrain all trading pairs. For example, if you changed the profit percentage target.
• You copied the AI for use in another bot with different markets — retrain for those markets.

Paper Trading vs real funds bot

When using the same AI in both a Paper Trading bot and a live bot, results may differ. Both bots check the same market within a minute and use the current signal — but unless they check at the exact same moment, the signals will differ. This can actually be an advantage: using both bots allows you to check markets more often and faster.

Results

Go to the Editor section to see your Strategies ranked by score — highest first. "Total signals" for a Trend Strategy represent trend changes, not actual trade signals. Strategies without a background colour are performing well; yellow indicates lower performance; red indicates poor performance.

Click on a Strategy to see detailed performance — including results by market condition, average score, and best-performing trading pairs. Note: swing trades can take up to a week or longer to complete. An AI with strategy scores between 40%–60% can still be profitable depending on your AI Config settings.

On the right side of the screen, you'll see average sell signals delivered per market condition. Each of the six scores shows how well signals hit their targets in a given trend. For example, a buy score of 83.81% in an uptrend means that in 849 out of 1,013 cases, the buy signal reached the configured profit target within the validation period. A sell score of 73.22% in a downtrend means that in about 990 out of 1,352 signals, the price dropped by the configured percentage within the validation period.

In the Result section, you'll see the average trend, buy, and sell score based on your configured criteria. It's possible to see no buy or sell signals in a certain trend type — this may mean your strategy is built for a specific market condition, or your Trend Strategy has a strict approach to identifying that trend.

Additional information

Training one trading pair takes approximately 2 hours, depending on the number of sources in your AI. After training, select your trading pairs under "Coins and amounts" in your Baseconfig.

Reset Deletes all existing training data so the AI starts collecting fresh data.

Delete Permanently deletes the AI. This action cannot be undone.