AiTrading

• AI aided polymarket/kalshi betting
• Integrated dashboard
• Proper PnL tracking and analysis ⇒ self correcting formula
• Leverage LLMs and our own large scale research via scrapers
• Continuous analysis

TODO

• Define datapoints for each flow
• Define flows
• Risk mgmt solution (low prio)
• max/min
• Proper auditing
• Kill switch

Flow

• A flow is a scheduled/ad-hoc sequence of steps that either produce or derive a datapoint. Or open/close/update a trade position.
• Triggers can come from other flows, manual triggers, cron jobs, news updates etc
• Flows should have templates sort of like interfaces
• Generally speaking, most data point flows should produce a pure output.

Formulas

$$\text{Risk:Reward}=\frac{\text{Potential Loss (Risk)}}{\text{Potential Profit (Reward)}}$$ $$\text{Value at Risk}=W_0\times (1-\exp ({\mu }_{c}T+{\sigma }_c\sqrt{T}{N}^{-1}(p_L)))$$ $$\text{Advanced Trend Formula}=\left(\sqrt[n]{\frac{{\text{Value}}_{\text{end}}}{{\text{Value}}_{\text{start}}}}-1\right)\times 100\%$$ $$nt=\overrightarrow{GT}-\Vert \left(\begin{array}{cc}Negative& Positive\end{array}\right)\Vert$$ $$\overrightarrow{BuyIncentive}=\overrightarrow{nt}$$

Flow execution

erDiagram

flCtx[FlowCtx] {

Type Name "Resolve As"

input dailyVolume "FlowResult"

input openPosition "Position"

property trigger "resource"

}

DATABASE ||--|| FLOW-INPUT-A : persist

DATABASE ||--|| FLOW-INPUT-B : persist

FLOW-ENGINE ||--o{ flCtx : inputs

SCHEDULE ||--|{ FLOW-ENGINE : trigger

AD-HOC ||--|{ FLOW-ENGINE : trigger

FLOW-INPUT-A }|..|| flCtx : reads

FLOW-INPUT-B }|..|| flCtx : reads

flCtx }|--o{ FlowRun : produce

FlowRun ||--|{ "FlowResult [K1]: V" : success

FlowRun ||--|| FlowFailure : fail

FlowRun }o..o{ Position : position

FlowRun ||--|{ "FlowResult [K2]: V" : success

"FlowResult [K2]: V" }|--|| FlowExecution : produce

"FlowResult [K1]: V" }|--|| FlowExecution : produce

Position }|--|| FlowExecution : produce

FlowFailure }|--|| FlowExecution : produce