Tutorial - Probability Flow Calculator

Left click anywhere in the draw area to summon the node creation dialog.

A node generally represents an event or state.

We want to analyse the potential consequences of the US 2024 presidential election, so we create a node about the election taking place.

If a node has no incoming edges, it is a root node, indicated by brown-ish border.

Root nodes are assumed to have a probability of 1 to happen. They are independent events which we consider to always take place. Every probability flow must stem from one or more root nodes. A root node doesn't have to be specific, it can simply be e.g. "the current state of the world". But we cannot define a root node with a sub-1 probability.

In our case, we can be fairly certain that the election will take place, so using a root node for that is fine.

Any root node is converted into a standard node if any incoming edges are connected to it. Similarly any standard node will convert into a root node if all its incoming edges are removed.

Click an existing node to begin drawing an edge. These are directional connections joining to different outcomes.

When drawing an edge, if you click the empty canvas, a node creation dialog will open up. If you instead click another existing node, edge creation dialog will open:

In the edge creation dialog, you can define the probability of an outcome for your origin node. Note that the total outgoing probabilities of a node shouldn't exceed 1, since the outcomes are considered mutually exclusive. An oversubscription is indicated by the source nodes' description turning red.

We will discuss mutual exclusivity and other attributes of probabilities more in the advanced section.

After pressing enter or clicking "Save Edge", the edge is drawn. You can edit the probability of an edge at any time by clicking the probability value in the drawn edge. You can also delete the edge in the edit.

We will draw an edge and create a new node: Trump wins. We will give the edge a 49% probability, in accordance with the polls in the beginning of October 2024. We will then draw another edge from the same source node: Kamala wins. We give that edge a 51% probability.

Most nodes will be standard nodes. Any node which has incoming connections is a standard node. A standard node has three attributes:

• Description: Short text describing the event or state
• Probability: A float value between 0 and 1 describing the probability of that event taking place.
• Node type: Node type determines the calculation done to the incoming probabilities to determine the node probability. Default is OR.

Border color helps distinguish between node types. Usually you will use mainly OR nodes. OR will calculate the probability that one or more of their incoming events will take place. AND nodes calculate the probability that all of their incoming events take place.

Since we have only one incoming edge to each of our two standard nodes, Trump Wins and Kamala Wins, it is irrelevant whether we pick OR or AND. We will go with the default OR for both. The node type can be changed by clicking the node's AND or OR, which opens a dialog.

The probability value of a standard node is determined by its inbound probabilities and the node type, i.e. the calculation taking place. It can't be directly adjusted.

Next we use the previously described mechanisms to add three more nodes and connect the two mutually exclusive outcomes.

Now we have defined three possible outcomes for Ukraine support given the outcome of the presidential election. We estimated the probabilities for the edges individually, assessing that Kamala has a higher chance to increase support to Ukraine than Trump. (Note, the edge probabilities may not render very clearly here). We give Kamala probabilities of 0.2, 0.3 and 0.5 to "reduced", "same" and "increases", respectively. We give the opposites to Trump.

So, we see that all three outcomes are fairly equally probable according to this analysis and the probabilities we use. Increase in support is slightly more likely than the alternatives.