Read the explanation
The saved simulator assigns each feed a freshness penalty, latency penalty and poisoning penalty, then multiplies their sum by its fixed feed weight. Latency penalty is milliseconds divided by one thousand, times zero point eight. Illustrative latencies two hundred fifty and five hundred milliseconds therefore contribute zero point two and zero point four before feed weighting. Bars show one hundred twenty and two hundred forty pixels at six hundred pixels per penalty unit. Freshness instead uses zero point four times freshness divided by ten, raised to one point three five. Poisoning percentage contributes one point five times its entered value. The saved feed weights are one half, three tenths and one fifth, summing to one. Feed names are fixture labels; this page does not establish actual measurements from those providers. These penalty formulas are authored assumptions rather than empirically validated estimates of agent beliefs. To isolate the poisoning formula, consider a hypothetical single feed of weight one with zero freshness and latency. Poisoning five gives belief drift seven point five. Poisoning ten gives drift fifteen. The displayed bad trade percentage formula adds two point two times drift and one point eight times weighted poisoning, producing twenty five point five and fifty one. Raw execution risk adds zero point zero two five times drift and zero point zero three times poisoning. It becomes zero point three three seven five and zero point six seven five. Bars show one hundred thirty five and two hundred seventy pixels at four hundred pixels per raw risk unit. The returned values are rounded to two decimals, while the danger comparison uses the unrounded risk and is greater than or equal to the chosen threshold. Drift caps at forty five, bad trade percentage at one hundred and risk at one. This coupled score is a fixed simulation formula, not a calibrated probability or security guarantee. The expected drawdown field multiplies the displayed bad trade formula, before its output rounding, by ten percent of the configured portfolio. At an illustrative portfolio of one hundred thousand dollars, a twenty percent assumed probability yields two thousand dollars, while forty percent yields four thousand. Bars compare one hundred twenty five and two hundred fifty pixels at point zero six two five pixels per dollar. The ten percent loss severity is fixed in the source; the calculation does not simulate positions, exposure, realized returns or a distribution of market outcomes. Confidence is another derived display, maximum zero of one hundred minus probability minus drift. It is not measured model confidence. In this offline copy the missing D3 dependency interrupts graph initialization before the first update and resize handler registration; math also requires its external dependency. The explainer therefore describes the saved formulas and their assumptions, without claiming live feed access, healthy graph operation, trading performance or investment advice.