Time is Non-Linear in Prediction Markets

And why every tool you've seen for them imports the wrong assumption.

---

Open a chart of any S&P 500 stock at the 1-hour timeframe. The price moves. Maybe a lot, maybe a little. Even at 3am with no headlines, there's a tape: bid, ask, microstructure noise, market makers breathing. Clock and market run in lockstep.

Open a prediction market on the same timeframe. Half the time you'll see flat lines. A market priced at 23¢ for fourteen hours straight, four trades in total. Then a sudden gap to 47¢ in eight minutes when a wire hits. Then flat again for the rest of the day. That isn't a chart bug. That's the market working exactly as designed.

Stocks trade in continuous time. Prediction markets trade in event time. The difference is not cosmetic. It is one of the most fundamental structural gaps between the two asset classes, and almost every tool you'll see for prediction markets gets it wrong because it inherited the assumption that time is uniform.

Stocks live in continuous time

A stock price is, in the simplest framing, a discounted stream of expected future cash flows. The cash flows are continuous in concept and the price discounts them continuously. Information about those cash flows arrives in a relatively smooth stream: earnings quarterly, guidance regularly, macro data weekly to monthly, sector beta drifting daily, options flow signaling positioning, microstructure liquidity providing tick-level price discovery.

Yes, stocks have gaps. Earnings can move a stock 15% in a single print. FDA decisions can halve a biotech overnight. M&A headlines reprice a name in seconds. But the underlying process is fundamentally one of constant discounting against a continuously evolving information set. Even on a "boring" day, positioning shifts, factor moves accumulate, the curve steepens, somebody finds something interesting in the 10-Q.

This is why traditional charting works for stocks. A 1-hour candle is meaningful because there is roughly 1 hour's worth of trading, fueled by roughly 1 hour's worth of information. A 1-day candle aggregates a day of price discovery against a day of new information. The clock and the information stream are roughly aligned.

You can analyze stocks as time series because they actually behave like time series. Trends, volatility, mean reversion, factor loadings, microstructure: all of these assume that time is a thing the market is moving through at a roughly steady rate.

Prediction markets do not work this way

A prediction market resolves to a binary outcome at some future date. Until it resolves, the price is the market's probability estimate. But the path from open to resolution does not run on a steady-rate process. It runs on the information schedule of whatever the market is about.

A market on whether a specific bill will pass Congress can sit at 23¢ for thirty days while no one in DC does anything relevant to it. Then a procedural vote in committee gets scheduled, a 24-hour rule gets posted, the bill gets pulled from a markup, and across one afternoon the market relocates from 23¢ to 51¢ to 67¢. Thirty days of nothing, then five hours that contain the next thirty days of probability.

A market on whether a hurricane will make landfall above Category 3 will price quietly until the European model runs at 06z. The model run drops, the cone shifts twenty miles east, the market re-prices in twelve minutes. Then it sits flat again until the next run at 12z.

A market on a court ruling will be flat for weeks until the docket shows a filing. A market on a Senate confirmation will be flat until a senator publicly commits. A market on a Fed decision will be flat until a Fed governor speaks at a conference and someone parses one phrase as more dovish than expected.

Calendar time keeps ticking. Information time does not. They are not the same clock, and almost everything interesting about prediction market trading happens in the gap between them.

The chart isn't broken. The clock is

If you look at any prediction market chart, anywhere, you'll see things that look pathological by stock-chart standards. Long flat lines. Volume that disappears for half a day. Price moves of 15¢ that happen in two trades and then bounce back. Liquidity that ghosts in and out. Wide spreads at one moment, tight at the next, with no visible cause on the chart itself.

Every one of these is the market behaving correctly. The flat line is the market saying "no new information, my current estimate stands." The volume desert is the market saying "no one is paying attention right now because nothing is scheduled to happen." The 15¢ jump is the market saying "a piece of information just landed that meaningfully revised my probability." The spread widening is the market saying "I'm not confident in my current price because the next information event is unpredictable."

Imposing a candlestick framework on this is forcing the wrong abstraction. A 1-hour candle on a market in its quiet phase is meaningless aggregation of nothing. A 1-hour candle on a market in its event phase is too coarse to capture the actual price discovery, which happens in seconds.

This is why prediction market traders who came from equities sometimes feel like the markets are "broken" or "thin." They aren't broken. They're running on a different time substrate. The substrate is event time, and clock-time visualizations don't render it.

Five clocks running at once

What makes prediction market time non-linear isn't just that information arrives in bursts. It's that several different time dimensions run simultaneously, often at very different rates, and the effective time the market is in at any moment is the composition of all of them.

Calendar time is the obvious one. Wall-clock time until the market resolves. This is the only time most tools display.

Information time is how much of the relevant fact set remains unknown. A market on a Senate vote that already had 51 senators publicly commit has very little information time left, regardless of how many calendar days remain. A market on a Fed decision two days away with no recent Fed speeches has a lot of information time left, even though calendar time is almost out.

Attention time is whether the market is currently being looked at. An obscure court-ruling market might have full information available but no one watching, so the price drifts only when the rare informed participant notices. The price isn't wrong because the information isn't there; it's wrong because the attention isn't there. Attention can arrive suddenly, often triggered by a tweet or a news pickup, and reprice a market in minutes despite no new underlying information.

Procedural time is the schedule of the institution the market is tracking. Congress runs on rule packages, motion calendars, suspension lists, conference committees. Courts run on dockets, scheduling orders, briefing schedules. Weather runs on model run cycles. Economic data runs on agency release calendars. Procedural time is largely deterministic given the institution, but it's invisible if you're only looking at calendar time and price.

Liquidity time is when the order book actually has depth. A market with $50 of size on the bid and ask isn't really tradable in the same sense as a market with $5,000 on each side, even at the same nominal price. The 20¢ on the screen might be a ghost price that vanishes the moment you cross. Liquidity time correlates with attention time but not perfectly. Sophisticated market makers post in quiet periods. Retail floods in during loud ones.

In equities, all five of these get compressed by deep, continuous liquidity and a constant flow of information. They become invisible because the substrate is dense enough to mask their unevenness. In prediction markets, the substrate is sparse, and the unevenness is right at the surface.

Where alpha actually lives

If you trade prediction markets long enough, you notice something. Most of your alpha doesn't come from "I have a better forecast of the underlying outcome." It comes from "I know what time it is."

The classic mispricing is treating calendar time as real time. A bill is 180 days from its deadline; the market prices it accordingly. But the bill missed its committee vehicle, leadership didn't include it in the next package, and procedurally it's effectively dead. The calendar says half a year. The procedure clock says it's over. The market is pricing the calendar. The alpha is in the procedure.

The opposite mispricing also happens. A market is one day from resolution; the price is high because "it's almost over." But the key fact, the weather model or the court filing or the agency announcement, hasn't dropped yet. Calendar time is almost out. Information time is still wide open. The market is pricing the calendar. The alpha is in the information.

A different version: a market has been flat for weeks, looks dead, looks fully priced. Then a small piece of new procedural information lands, a hearing scheduled, a docket entry filed, a senator opening their schedule. Almost no one notices. The price doesn't move. But the procedure clock just advanced. Anyone reading the docket knows the market should be 8¢ higher. It will be, in 36 hours, when the news pickup happens. The alpha is in the attention lag.

The general principle: most participants treat all five clocks as one clock and assume the one clock is calendar time. The trader who has the right multi-clock model can see when those clocks have desynchronized, and act before the rest of the market resyncs.

What this means for product

If event time is the substrate, then a tool for prediction markets shouldn't just display calendar time. It should display whatever clocks are currently dominant for the specific market in question.

This is the core idea behind what SimpleFunctions has been building. Query Gov, Query Econ, world delta, indices, monitors, contextual next-actions: these aren't just more data. They are an attempt to make the actual relevant time of the market visible, market by market.

The right framing:

SimpleFunctions maps clock time, source time, and market time into one state object.

For a Congress market, the relevant clocks are calendar (deadline), procedural (committee schedule, floor calendar, suspension list, conference status), information (key amendments still pending), attention (news cycle position). Query Gov isn't a search engine over bills. It's a way to expose the procedural clock for any bill, instantly, so a price can be read against the right time.

For a market on an economic data release, the relevant clocks are calendar (release date), information (consensus estimate distribution, last print, sub-component drift), attention (whether the print is being warmed up by Fed-speak in advance). Query Econ isn't a FRED mirror. It's a way to expose the data calendar and the distance to the last print, so a market price has the right substrate to be evaluated against.

For a weather market, the dominant clock is model run schedule. GFS, ECMWF, ICON, runs at fixed times throughout the day. The market's effective time advances when a run drops, not when the wall clock ticks. A weather-market dashboard that doesn't show "next model run in 47 minutes" is missing the only time signal that matters.

For an election market, the relevant clocks are poll cadence, filing deadlines, debate schedule, primary date, FEC report dates. Calendar time alone tells you almost nothing about whether the market is in a live information window or a dead waiting period.

For a court or agency market, the dominant clock is docket activity. The market's clock effectively advances when a filing posts, when a scheduling order issues, when an agency posts a rule. Between those events, the calendar moves but the market should not.

A real terminal for prediction markets has to know what kind of clock dominates which market and surface it. "1m / 5m / 1h / 1d" is a question shaped by stock-chart UX, and it's the wrong question. The right question is: what is the next real information node, and how far away is it?

What a prediction market chart should actually show

Take everything above and ask: what should a chart for these markets look like?

Price-over-time is still part of it. But the price axis is the easy part. The hard and useful part is overlaying the time substrate.

Event markers: visible points on the timeline where a real event occurred. Court filing dropped. Committee voted. Model run released. Poll published. The price reaction next to the event tells you whether the market was already discounting it or got caught.

Source update markers: independent of price, when the upstream sources fed new information. The Congress mirror got 47 new bill actions overnight. FRED released CPI. The poll aggregator updated. These markers tell you when the information substrate moved, even if the price didn't.

Stale windows: visual shading on the chart that says "this region is information-dead." No new procedural events, no new releases, no new poll data. Just calendar drift. The chart should make it visually obvious that "nothing happened here" is the correct read, not a data gap.

Information gaps: the inverse of stale windows. Periods where the procedure or information clock advanced rapidly but the market hadn't reacted. The visual signal of "the market is behind."

Next catalyst countdown: not "time to resolution" but "time to the next scheduled event that will move price." For a bill, that's the next committee meeting. For a weather market, the next model run. For a Fed market, the next Fed speaker on the calendar.

Market reaction lag: how fast the market reacted relative to the source update. A consistently fast reaction means the market is being watched. A consistently slow reaction means there's an attention lag worth exploiting.

These overlays don't replace price and volume. They give price and volume meaning by plotting them against the right substrate.

TradingView draws price as a function of time. A SimpleFunctions terminal should draw the non-linear relationship between events, information release, and market reaction. Price sits inside that relationship, not the other way around.

The watcher

This is also why a watcher running continuously on these markets is more useful than a daily research note. The watcher's job isn't to issue calls. It's to keep asking, market by market: did the time layer just change?

Did this market shift from waiting period into event window? Did the procedure clock cross a threshold no one's flagged yet? Has attention finally caught up to a fact that landed three days ago? Is the market still pricing on the previous time layer while the actual layer changed at 14:32 UTC?

A watcher running on event time is far more valuable than a watcher running on calendar time. Most prediction-market trades that have any edge are decided in the moment some clock changes, and the trader who notices the change first, by minutes or by hours, takes the move. By the time it shows up in a daily summary, the alpha is already gone, distributed across whoever watched it happen live.

The deeper claim

Prediction markets do not trade in clock time. They trade in event time.

Most tools fail at prediction markets because they were built for asset classes whose time was already smoothed for them by deep liquidity and continuous information flow. Charting libraries assume time is a constant-rate axis. Dashboard frameworks assume "1 hour" is a meaningful aggregation. News feeds assume relevance is roughly proportional to recency.

Prediction markets violate every one of these assumptions. The flat lines aren't bugs. The jumps aren't outliers. The waiting isn't dead air. The market is keeping its real clock, the one made of committee calendars and weather model runs and docket entries and poll releases, instead of yours.

The opportunity in building for these markets is exactly to make that clock visible. To map calendar time, source time, and market time into one observable state. To turn the non-linearity of prediction-market time from a chart artifact into the primary thing the user sees.

That is what SimpleFunctions is for.