How Kunkafa works out the chance a market reaches a price

Seventy experts, one question each

Seventy independent views of the same market. When they agree you see it; when they disagree, that tells you something too.

Each expert looks at the same market and answers on its own: how far, and which way. There is no single model with a single opinion. The chance you see for a level is not one voice; it is what the record shows for forecasts like this one, and the experts’ agreement or disagreement is inside it. A level that many experts reach the same way carries a higher chance; a level they split on carries a lower one, and you see that.

We do not publish which methods the experts use or how they were built. What we publish is how their forecasts turned out, every day.

What a forecast is made of

Every forecast is four things, and always four: how far (the change from today’s price, as a percentage), which way (up or down), how long it has (the duration), and the chance it gets there. The side you asked about comes first; the other side sits below it. A forecast that hid one side would be claiming the price cannot go there, which is false.

On the screen the four things are a ladder: one row per level, with its change, its duration and its chance, above a price line that draws the same levels as a band.

Where the chance comes from

The chance beside a level is a rate from the record: of past forecasts like this one, at this strength, how many reached the level. It is measured against what actually happened, not estimated from a formula about volatility. That is the difference between Kunkafa and an expected-move calculator: the calculator tells you a range the options market implies; Kunkafa tells you how often the price really got there.

The experts’ stated chances are checked against outcomes and corrected, so that a level shown at 70% is one that forecasts like it reached about seven times in ten. In the record you can see this for yourself, by strength.

Duration is a ceiling, not a choice

The slider under the market runs from five minutes to seven years, and its handle is a ceiling. “Up to 7 years” shows every duration at once, pooled into one picture; slide it down and the longer durations drop out. The bands under the track, Scalping, Day trading, Swing, Investing, sit where those durations fall, so the scale explains itself.

Everything downstream pools the same way: the ladder, the price band, the record and sharing all cover every duration up to the ceiling.

The record and its rules

Beside every forecast sits “How similar forecasts performed”: how many forecasts were evaluated, how many were strong enough for your filter, how many finished in the forecast direction, and how many reached the level. It is counted over the last day, 7 days, 30 days, or all time, and it follows four rules:

• Under 100 finished examples, no rate. The record says “Not enough past examples to estimate performance reliably.” instead of showing a number nobody should trust.
• One filter drives everything. The strength you set narrows the list, and the record is counted on the same setting, so a number on screen never disagrees with the control above it.
• Filters narrow the list, never the drawing. The chart always shows every level.
• A rate under 1% says “under 1%”. One percentage format, and every percentage says what it measures.

Each past forecast ends one of five ways: it reached the level, it passed it, the period ended ahead, the period ended behind, or it is still running. The results page publishes the same counts for every market, refreshed each morning.

Training and testing, in scale

The experts were trained on about 10 billion data points of price history across stocks, indices, currencies, commodities and crypto, and each was checked on millions of past moves it had never seen. After training, their stated chances were compared with what happened and corrected, so that the chance you read means what it says.

That is all we publish about the internals, on purpose. What we publish instead, and what you should hold us to, is the record.

What we never do

• No advice. A forecast is a chance, not an instruction. Forecast, not advice. For the rational investor: emotion out, scenarios in.
• Nothing simulated. An empty market says it is empty; a missing value is left blank and named, never filled with a plausible number.
• Nothing metered. Looking at a forecast costs nothing and moves no counter.
• No red and green, no countdowns, no streaks, no celebration. Up is mint, down is fuchsia, and nothing rewards watching the screen.
• No hiding one side. Up and down: the side you asked about first, the other always within reach.

Glossary

Expected move

How far a market’s price could move, up or down, over a duration. Kunkafa shows several levels of expected move per market and duration, each with the chance of reaching it.

Chance of reaching

How often past forecasts like this one actually reached the level, measured against what happened. It is a rate from the record, not an opinion.

Duration ceiling

The duration slider’s handle. “Up to 7 years” shows every duration from five minutes to seven years at once; lowering it removes the longer ones.

Forecast strength

A filter on how strong a forecast has to be to appear in the list and count in the record. It narrows the list, never the drawing.

Seventy experts

The seventy independent judgements behind each market’s forecast, each giving how far and which way on its own. Agreement and disagreement both show in the chance.

Finished in the forecast direction

A past forecast that ended, at the close of its own duration, on the side it forecast.

Reached the level

A past forecast whose price touched the forecast level at some point within its duration, whether or not it stayed there.

Still running

A past forecast whose duration has not ended yet, so it has no outcome.