A close is not a day.

On Polymarket the price is the probability. A YES share that pays $1 if the event happens trades at $0.63 when the book thinks the event is 63% likely. The natural picture of a changing estimate is a single stroke: last traded probability, connected through time.

That stroke is not a lie. It is a projection. A line has room for one number per timestamp. The open of the interval, the most confident the book got, and the least confident it got are not hidden behind a setting. A line has nowhere to put them.

Equity and crypto traders already know this distinction. They would not accept a Bitcoin chart that plotted only the daily close and called it a day. Prediction-market tools still ship that chart as the default, because the venue itself publishes a close-only series and most overlays just restyle it.

A smooth line reads as a consensus. It has no wicks, so it has no visible disagreement. A day the book ripped from 40% to 70% and closed back at 42% looks, on a line, like a two-point day. The eye files it under “nothing happened.” The book did not agree.

That is the specific way a probability line misleads. It does not invent a number. It discards the range, and range is the thing you use to decide whether a move is real, whether a level was tested, and whether the people still in the market have already been shaken out.

Smoothing makes the same problem worse. A moving average drawn on top of a close-only series is two layers of compression: first the day is reduced to one print, then those prints are averaged. What remains is a curve that is pleasant to look at and almost useless for timing. Predm does not bake a moving average into the series it rebuilds. If you want one, you add it yourself, on top of candles that still have their wicks.

On 12 June 2026 this market closed at 46.5% after a previous close of 46.5%. Close to close that is no move at all. Inside the day it traded as low as 36.5% and as high as 52.5%: a 16.0-point swing in what the book believed.

Across the snapshot the close-to-close move came in under half the day’s true range on 23 of 43 days. Most of the time, most of what happened lived in the wick. A line chart of this market is a chart of the minority of the action.

A prediction market ends at 0 or 1. That fact is sometimes used to defend the line: if the only thing that pays is the final print, why keep the path? Because you do not trade the resolution. You trade the days before it. The path is the whole of the tradeable history.

A market that spends two weeks oscillating between 35% and 55% and then rips to 90% on a headline is a different market from one that grinded from 40% to 90% in a straight line. Both can close at 90% the night before resolution. The line, on that last evening, looks the same. The people who got chopped in the range, and the people who never had a reason to fade the grind, are not the same traders.

The destination is the settlement. The path is the trade. Prediction-market OHLC is the longer version of that sentence.

A close-only series is most dangerous on a quiet market. One print at 14:00 and one print at 18:00 produce a tidy segment. Between them the book may have been empty, or it may have printed a 20-point wick on a single aggressive fill that nobody else was there to fade. The line draws the same shape either way.

A candle cannot invent liquidity that was not there, but it can refuse to pretend. An interval with one trade is a doji sitting on a pin, or it is no candle at all. Predm does not draw a candle on a market that cannot fill an open, a high, a low and a close that mean something. A line has no equivalent honesty: it always has a last price.

Start with candlesticks rebuilt from the raw trade feed, not a restyled close. Then:

Range, not just direction. High minus low is the day’s disagreement, measured in points of probability. A 3-point body inside a 20-point range is a market that tried something and could not hold it.

More than one timeframe. A line on the daily and a line on the hourly are two compressed stories. Multi-timeframe signals exist because a day can be up while the hour is breaking, and that disagreement is information.

Who is on the other side. A 70% close held by a crowd is not the same 70% as a close held by three wallets. Holder concentration is how you tell those two markets apart. The line cannot.

A line is not fake data. It is incomplete data. The close is real. It is just not the day.

Candles are not a trading system. Rebuilding OHLC does not tell you which way a market resolves. It tells you what the book actually did, at a resolution a line cannot carry.

Two venues are live. Predm reads Polymarket and Hyperliquid. Aster, Binance, Kraken and Kalshi are marked coming soon on the homepage and no surface here serves their data yet.

Does Polymarket itself show candlesticks?

The venue’s own chart is a probability line. Third-party tools that overlay a line on that series are restyling the same one-number-per-point history. A candle needs the open, high and low of each interval, which have to be rebuilt from trades.

Is a reconstructed candle the same as a TradingView candle?

Same four fields, different underlying. A crypto candle is a price. A prediction-market candle is a probability. The axis is 0–100 and the market resolves to 0 or 1. Polymarket vs crypto charts is the comparison.

Can I just add a moving average to the line?

You can, and you will still be averaging closes. The wick information is already gone. If you want an average, put it on candles so the range is still visible underneath.

When is a line good enough?

For a headline number — “this market is at 63%” — a line is fine. For anything that depends on whether a level was tested, whether the day was violent, or whether the book has already been through a range, it is not.