Signal Ranking

Timing is everything, and some opportunities are riper than others. How do you find which of the signals to act on?

Model training customized per token

Custom AI models take in the massive array of social, price, and chain signals (around 20,000 columns) per token and generate predictions in the form of probability distribution functions (PDFs) that map the likelihood of different ROIs over a forecast range.

Probability Distribution Functions.

Buy signal indicators are calculated per risk profile

The probability distribution functions output by the model are combined with a user's risk profile then used to calculate a custom Indicator based on that risk profile.

Let's go through an example using one of the predefined risk profiles:

conservative_risk_profile = {
    1: [0.1, -0.1],
    2: [0.2, -0.2],
    3: [0.2, -0.2],
    7: [0.2, -0.2]
}

Which, in words translates to:

• Day 1: Aim for 10% gain, accept 10% loss.

• Day 2: Aim for 20% gain, accept 20% loss.

• Day 3: Aim for 20% gain, accept 20% loss.

• Day 7: Aim for 20% gain, accept 20% loss.

The likelihood of these price changes across 1, 2, 3, and 7 days are calculated using the probability distribution functions above and are combined to form an aggregate prediction, or Indicator (shown in red below)

$PORTAL signal chart pre- and post-launch. Blue = Price, Red = Indicator

When the Indicator is positive, it's a buy signal. The more positive the Indicator, the stronger the buy signal. Indicators are calculated for each token in each Watchlist for each user according to each set of risk parameters everyday.

A positive Indicator is the default trigger to construct a position. Signals are ranked according to the Indicator value.